USF Libraries
USF Digital Collections

The tip of the blade

MISSING IMAGE

Material Information

Title:
The tip of the blade self-injury among early adolescents
Physical Description:
Book
Language:
English
Creator:
Alfonso, Moya L
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla.
Publication Date:

Subjects

Subjects / Keywords:
Youth
Segmentation
Self-harm
CHAID
YRBS
Dissertations, Academic -- Measurement and Evaluation -- Doctoral -- USF   ( lcsh )
Genre:
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Summary:
ABSTRACT: This study described self-injury within a general adolescent population. This study involved secondary analysis of data gathered using the middle school Youth Risk Behavior Survey (YRBS) from 1,748 sixth- and eighth-grade students in eight middle schools in a large, southeastern county in Florida. A substantial percentage of students surveyed (28.4%) had tried self-injury. The prevalence of having ever tried self-injury did not vary by race or ethnicity, grade, school attended, or age but did differ by gender. When controlling for all other variables in the multivariate model including suicide, having ever tried self-injury was associated with peer self-injury, inhalant use, belief in possibilities, abnormal eating behaviors, and suicide scale scores. Youth who knew a friend who had self-injured, had used inhalants, had higher levels of abnormal eating behaviors, and higher levels of suicidal tendencies were at increased risk for having tried self-injury.Youth who had high belief in their possibilities were at decreased risk for having tried self-injury. During the past month, most youth had never harmed themselves on purpose. Approximately 15% had harmed themselves one time. Smaller proportions of youth had harmed themselves more frequently, including two or three different times (5%), four or five different times (2%), and six or more different times (3%). The frequency of self-injury did not vary by gender, race or ethnicity, grade, or school attended. Almost half of students surveyed (46.8%) knew a friend who had harmed themselves on purpose. Peer self-injury demonstrated multivariate relationships with gender, having ever been cyberbullied, having ever tried self-injury, grade level, and substance use. Being female, having been cyberbullied, having tried self-injury, being in eighth grade, and higher levels of substance use placed youth at increased risk of knowing a peer who had self-injured.Chi-squared Automatic Interaction Detection (CHAID) was used to identify segments of youth at greatest and least risk of self-injury, frequent self-injury, and knowing a friend who had harmed themselves on purpose (i.e., peer self-injury).
Thesis:
Dissertation (Ph.D.)--University of South Florida, 2007.
Bibliography:
Includes bibliographical references.
System Details:
System requirements: World Wide Web browser and PDF reader.
System Details:
Mode of access: World Wide Web.
Statement of Responsibility:
by Moya L. Alfonso.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 235 pages.
General Note:
Includes vita.

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 001920201
oclc - 187918834
usfldc doi - E14-SFE0002096
usfldc handle - e14.2096
System ID:
SFS0026414:00001


This item is only available as the following downloads:


Full Text
xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam Ka
controlfield tag 001 001920201
003 fts
005 20080107110418.0
006 m||||e|||d||||||||
007 cr mnu|||uuuuu
008 080107s2007 flu sbm 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0002096
035
(OCoLC)187918834
040
FHM
c FHM
049
FHMM
090
LB3051 (ONLINE)
1 100
Alfonso, Moya L.
4 245
The tip of the blade :
b self-injury among early adolescents
h [electronic resource] /
by Moya L. Alfonso.
260
[Tampa, Fla.] :
University of South Florida,
2007.
520
ABSTRACT: This study described self-injury within a general adolescent population. This study involved secondary analysis of data gathered using the middle school Youth Risk Behavior Survey (YRBS) from 1,748 sixth- and eighth-grade students in eight middle schools in a large, southeastern county in Florida. A substantial percentage of students surveyed (28.4%) had tried self-injury. The prevalence of having ever tried self-injury did not vary by race or ethnicity, grade, school attended, or age but did differ by gender. When controlling for all other variables in the multivariate model including suicide, having ever tried self-injury was associated with peer self-injury, inhalant use, belief in possibilities, abnormal eating behaviors, and suicide scale scores. Youth who knew a friend who had self-injured, had used inhalants, had higher levels of abnormal eating behaviors, and higher levels of suicidal tendencies were at increased risk for having tried self-injury.Youth who had high belief in their possibilities were at decreased risk for having tried self-injury. During the past month, most youth had never harmed themselves on purpose. Approximately 15% had harmed themselves one time. Smaller proportions of youth had harmed themselves more frequently, including two or three different times (5%), four or five different times (2%), and six or more different times (3%). The frequency of self-injury did not vary by gender, race or ethnicity, grade, or school attended. Almost half of students surveyed (46.8%) knew a friend who had harmed themselves on purpose. Peer self-injury demonstrated multivariate relationships with gender, having ever been cyberbullied, having ever tried self-injury, grade level, and substance use. Being female, having been cyberbullied, having tried self-injury, being in eighth grade, and higher levels of substance use placed youth at increased risk of knowing a peer who had self-injured.Chi-squared Automatic Interaction Detection (CHAID) was used to identify segments of youth at greatest and least risk of self-injury, frequent self-injury, and knowing a friend who had harmed themselves on purpose (i.e., peer self-injury).
502
Dissertation (Ph.D.)--University of South Florida, 2007.
504
Includes bibliographical references.
516
Text (Electronic dissertation) in PDF format.
538
System requirements: World Wide Web browser and PDF reader.
Mode of access: World Wide Web.
500
Title from PDF of title page.
Document formatted into pages; contains 235 pages.
Includes vita.
590
Advisor: Robert Dedrick, Ph.D.
653
Youth.
Segmentation.
Self-harm.
CHAID.
YRBS.
0 690
Dissertations, Academic
z USF
x Measurement and Evaluation
Doctoral.
773
t USF Electronic Theses and Dissertations.
856
u http://digital.lib.usf.edu/?e14.2096



PAGE 1

The Tip of the Blade: Self-Injury Among Early Adolescents by Moya L. Alfonso A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Educational Measurement and Research College of Education University of South Florida Major Professor: Robert Dedrick, Ph.D. Carol A. Bryant, Ph.D. John Ferron, Ph.D. Nancy Heath, Ph.D. Tony Onwuegbuzie, Ph.D. Date of Approval: June 25, 2007 Keywords: Youth, Segmentation, Self-harm, CHAID, YRBS Copyright 2007, Moya L. Alfonso

PAGE 2

Dedication This work is dedicated to my true family, those folks who have stood by me during these surreal years. To my girls, my loves, my life, what would I do for material without you? You have my heart. Theref ore, when you hurt, I am broken; when you smile, I melt. To my serious, kind-hearte d Lauren, who has hated mommy’s “job” all these years for the time it took away from her. I’m sorry darling. I will try to make it up to you and hope that I at l east made you proud. To my wild, purple-haired, piercingloving daughter, Jackie, thank you for all of the silliness you bring to our lives—well, maybe not all of it! You are worshipped, my love. To my husband, Peter, what can I say? You have known me as a student since we met over 16 years ago. Never has your support wavered. You believed I could finish wh en I did not. To my friend, Tracie, who I lost last year, I miss you every day. Finish ing this without you here was hard, but your belief in me was in my heart. To my friends Terri, Tina, Julie, Leah, Jeanine, and Jen, you have been called upon a great deal this past year, and you have risen to the occasion. I could not have finished this pr ocess without your support.

PAGE 3

Acknowledgements I am grateful to a number of people who supported me during the dissertation process. I am most grateful to my husband, Peter, who supported me through my program of study and gave me the gift of time, which was needed to conduct the research and write the dissertation. Peter spent many weekends driving two teenage girls everywhere and anywhere while mom sat in front of the computer. This did not go unnoticed; neither did the many meals he prepared (or drove to pick up) for the family. I am grateful to my colleagues at the Florida Prevention Research Center, especially Julie Baldwin, Carol Bryant, Kelli McCormack Brown (former faculty), and Robert McDermott, for their good humor, advice, s upport, and guidance duri ng this process. I am grateful to Jennifer Mainey, Safe School Liaison, and Sherri Reynolds, Supervisor Grants, Pupil Support Services for supporting this study and providing information on the study county when needed. I am extremely grat eful to my chair, Robert Dedrick, PhD, for his never-ending patience, kindness, good humor, guidance, and availability. Dr. Dedrick will never know how many times in the past six months I came close to quitting, but he should know how his patience, unders tanding, and willingness to work with me during those weekly meetings made this horrendo us process doable. I am grateful to my committee members for their support, guidance, a nd genuine interest in the topic. I am especially grateful to Carol Bryant, PhD, for her assistance with CHAID and the use of her office, not to mention her ever-p resent ability to make me laugh.

PAGE 4

i Table of Contents List of Tables iv List of Figures viii Abstract ix Chapter One: Introduction 1 Research Problem 3 Conceptual Framework 7 Research Purpose and Questions 8 Research Approach 8 Significance 11 Organization of Remaining Chapters 13 Definitions of Terms 13 Chapter Two: Literature Review 15 Introduction 15 Early Adolescence 15 Theoretical Approaches to Adolescent Risk Behavior 19 Self-injury 20 Definitions of Self-Injury 20 Etiology and Functions of Self-Injury 25 Prevalence and Trends of Se lf-Injury during Adolescence 30 Sociocultural and Gender Variation 32 Self-injury and Adolescent Development 34 Popular Culture and Self-Injury 39 Social Contagi on & Self-Injury 42 Behavioral Correlates of Self-Injury 50 Prevention and Intervention 51 Segmentation 59 Chi-square Automatic Interaction Detection (CHAID) 61 Segmentation Validity 66 Summary 67 Chapter Three: Method 70 Research Approach 70 Accessible Population 73 Instrumentation 75

PAGE 5

ii Measures of Self-injury 77 Data Collection 78 Protection of Human Subjects 79 Analysis Procedures 80 Step 1: Data Entry and Cleaning 80 Step 2: Creation of Study Datasets 82 Step 3: Variable Selection and Modification 82 Step 4: Description of Self -Injury in General Middle School Population 88 Step 5: Exploration of Rela tionships Between Self-Injury and Other Behaviors 89 Step 6: Identification of Meani ngful Segments of Youth Who SelfInjure 91 Step 7: Present Findings 94 Issues to Consider 94 Chapter Four: Results 96 Introduction 96 Research Purpose and Questions 96 Description of Self-Injury in General Middle School Population 97 Prevalence of Self-Injury 97 Frequency of Self-Injury 97 Peer Self-Injury 98 Bivariate Relationships Between Student Demographic Variables and Self-Injury Outcomes 98 Relationships between Self-Injury and Other Variables 99 Multilevel Logistic Regression Analyses 106 CHAID Analyses 109 Relationships between the Frequency of Self-Injury and Other Variables 119 Multilevel Logistic Regression Analyses 131 CHAID Analyses 135 Relationships between Peer Se lf-Injury and Ot her Variables 145 Multilevel Logistic Regression Analyses 151 CHAID Analyses 152 Cognitive Interviewing 156 Summary 157 Chapter Five: Discussion 161 Purposes of the Research 161 Overview of Method 162 Summary of Findings 164 Strengths & Limitations 174 Dissemination 179 Implications for Prevention 179 Implication for Further Research 187

PAGE 6

iii References 189 Appendices 209 Appendix A: 2005 Middle School Youth Risk Behavior Survey 210 Appendix B: Exploratory Factor Analysis Results 232 Appendix C: Relationships among Predictor Variables 233 Appendix D: Summary of Bivari ate and Multivariate Results 235 About the Author End Page

PAGE 7

iv List of Tables Table 1 Sample of Self-injury Meas ures Used with Adolescents and Associated Prevalence Rates 31 Table 2 Description of the Accessi ble Population by School (N=1743, December 2005) 74 Table 3 Comparison of Sample Obtained and Enrollment by School (December 2005) 75 Table 4 Middle School Youth Risk Be havior Survey Item Categories 76 Table 5 Interval-Level Variable De scriptive Statistics (N = 1748) 83 Table 6 Prevalence Information fo r Categorical Study Variables 83 Table 7 Individual Variables Selected for Use and Associated Theoretical Or Conceptual Framework 84 Table 8 Scales Developed for Use and Associated Theoretical or Conceptual Framework 84 Table 9 Scale Definitions and Inte rnal Consistenc y Reliability 85 Table 10 Scale Descriptive Statistics 87 Table 11 Cohen’s Effect Size Interpretation Rules-of-thumb 90 Table 12 Self-injury and Deve lopmental Theory Variables 100 Table 13 Self-injury and Precipitants of Self-Injur y (Chi-square tests of independence) 101 Table 14 Self-injury and Precipitants of Self-Injur y (Independent t-tests) 101 Table 15 Self-injury and Social Contagion (Chi-square tests of independence) 102 Table 16 Self-injury and Social Contagion (Independent t-tests) 102

PAGE 8

v Table 17 Self-injury and Substance Use 105 Table 18 Self-injury and Problem Behaviors (Chi-square tests of independence) 105 Table 19 Self-injury and Problem Be havior Comparisons (Independent ttests) 106 Table 20 Multilevel Logistic Regressi on Analysis of Factors that Predict Having Ever Tried Self-injury (N=1748) 107 Table 21 Effect Size Values for Segmentation of Having Ever Tried SelfInjury – Suicide Included 113 Table 22 Effect Size Values for Segmentation of Having Ever Tried SelfInjury – Suicide Excluded 116 Table 23 Effect Size Values for Segmentation of Having Ever Tried SelfInjury – Suicide Excluded (Transformed Variables) 119 Table 24 Frequency of Self-Inju ry and Development Variables 120 Table 25 Frequency of Self-Injury and Precipitants of Self-Injury 124 Table 26 Frequency of Self-In jury and Social Contagion 126 Table 27 Frequency of Self-In jury and Problem Behaviors 127 Table 28 Frequency of Self-Injury and Su icidal Ideation, Plans, and Attempts 128 Table 29 Frequency of Self-Injury and Having Ever Used Substances 131 Table 30 Multilevel Logistic Regression Analys is of Factors that Predict the Frequency of Self-Injury – Once versus Never (Past 30 Day Frequency) (N=1748) 132 Table 31 Multilevel Logistic Regression Analysis of Factors that Predict the Frequency of Self-Injury – More than Once versus Never (Past 30 Day Frequency) (N=1748) 134 Table 32 Multilevel Logistic Regression Analysis of Factors that Predict the Frequency of Self-Injury – More than Once versus Once (Past 30 Day Frequency) (N=1748) 135

PAGE 9

vi Table 33 Effect Size Values for Segmen tation of Frequency of Self-Injury – Suicide Included 139 Table 34 Effect Size Values for Segmen tation of Frequency of Self-Injury – Suicide Included 141 Table 35 Effect Size Values for Segmen tation of Frequency of Self-Injury – Suicide Excluded 145 Table 36 Prevalence (%) of Knowi ng a Friend Who Had Self-Injured by School Attended 146 Table 37 Developmental Theory Va riables (Independent t-tests) 146 Table 38 Peer Self-Injury and Precipitan ts of Self-Injury (Chi-square tests of independence) (N=1738) 147 Table 39 Peer Self-Injury a nd Precipitants of Self-Inju ry (Independent t-tests) 147 Table 40 Peer Self-Injury and Social Contagion (Independent t-tests) 148 Table 41 Peer Self-Injury and Prob lem Behaviors (Chi-square tests of independence) 150 Table 42 Problem Behavior Comp arisons (Independent t-tests) 150 Table 43 Multilevel Logistic Regressi on Analysis of Factors that Predict Peer Self-Injury (N=1748) 152 Table 44 Effect Size Values for Se gmentation of Peer Self-Injury 156 Table 45 Study Research Questions Procedures, and Key Findings 158

PAGE 10

vii List of Figures Figure 1 Sample tree diagram. 64 Figure 2 Model of research approach. 72 Figure 3 Segmentation of having ever tr ied self-injury with suicide included in the model. 112 Figure 4 Segmentation of having ever tr ied self-injury with suicide excluded from the model. 115 Figure 5 Segmentation of having ever tr ied self-injury with suicide excluded from the model (transformed variables). 118 Figure 6. Frequency of self-injur y by attitudes toward school. 120 Figure 7. Frequency of self-inj ury by belief in possibilities. 121 Figure 8. Frequency of se lf-injury by parent comm unication scale scores. 122 Figure 9. Frequency of self-injury by th e frequency of having been a victim of bullying. 123 Figure 10. Frequency of se lf-injury by frequency of having been a victim of cyberbullying. 124 Figure 11. Frequency of se lf-injury by time spent using computer or video games for fun. 126 Figure 12. Frequency of self-inj ury by abnormal eating scores. 127 Figure 13. Frequency of self-inj ury by suicide scale scores. 128 Figure 14. Frequency of self-i njury by deviancy scores. 129 Figure 15. Frequency of self-inj ury by substance use scores. 130 Figure 16. Segmentation of frequency of self-injury with suicide included in the model. 138

PAGE 11

viii Figure 17. Segmentation of frequency of self-injury with suicide included in the model (transformed variables). 140 Figure 18. Segmentation of frequency of self-injury with suicide excluded from the model. 144 Figure 19. Segmentation of knowledge of peer self-injury with suicide included in the model. 155

PAGE 12

ix The Tip of the Blade: Self-injury During Early Adolescence Moya L. Alfonso ABSTRACT This study described self-injury within a general adolescen t population. This study involved secondary analys is of data gathered using the middle school Youth Risk Behavior Survey (YRBS) from 1,748 sixthand eighth-grade stude nts in eight middle schools in a large, southeastern county in Fl orida. A substantial percentage of students surveyed (28.4%) had tried self-injury. The prevalence of having ever tried self-injury did not vary by race or ethnicity, grade, school attended, or age but did differ by gender. When controlling for all other variables in the multivariate model including suicide, having ever tried self-inj ury was associated with peer self -injury, inhalant use, belief in possibilities, abnormal eating behaviors, a nd suicide scale scores. Youth who knew a friend who had self-injured, had used inhala nts, had higher levels of abnormal eating behaviors, and higher levels of suicidal tendencies were at increased risk for having tried self-injury. Youth who had high belief in th eir possibilities were at decreased risk for having tried self-injury. During the past month, most youth had never harmed themselves on purpose. Approximately 15% had harmed themselves one time. Smaller proportions of youth had harmed themselves more frequently, including two or three different times (5%), four or five different times (2%), and six or more different times

PAGE 13

x (3%). The frequency of self-injury did not va ry by gender, race or ethnicity, grade, or school attended. Almost half of students surveyed (46.8%) knew a friend who had harmed themselves on purpose. Peer self-injury demonstrated multivariate relationships with gender, having ever been cyberbullied, ha ving ever tried self-i njury, grade level, and substance use. Being female, having been c yberbullied, having tried self-injury, being in eighth grade, and higher levels of substa nce use placed youth at increased risk of knowing a peer who had self-injured. Chisquared Automatic Interaction Detection (CHAID) was used to identify se gments of youth at greatest a nd least risk of self-injury, frequent self-injury, and know ing a friend who had harmed themselves on purpose (i.e., peer self-injury).

PAGE 14

1 Chapter One: Introduction Self-injury, also known as self-mutil ation, self-harm, and cutting, among other terms, has been referred to as the “faste st-growing adolescent behavioral problem” (Purington & Whitlock, 2004, p. 2). Already es tablished as a risk behavior within clinical and educational sett ings, self-injury is rapidly becoming defined as a problem behavior by society at large. Within school settings, self-injury has been described as a “silent school crisis,” refl ecting insufficient knowledge, co nfusion, lack of effective interventions, and the tendency for adults and youth to shy away from dealing directly with the issue (Carlson, DeG eer, Deur, & Fenton, 2005; Ga lley, 2003). Whether selfinjury is on the rise or is being reported more frequently because of recent media attention is unknown (Favazza, 1998; Purington & Whitlock, 2004). However, schools, hospitals, mental health in stitutes, and clinical repor ts suggest self-injury among adolescents is on the rise (Conterio & La der, 1998; Galley, 2003; Hawton, Harriss, Hall, Simkin, Bale, & Bond, 2003 ; Olfson, Gameroff, Marcus, Gr eenberg, & Shaffer, 2005; Pipher, 1994; Purington & Whitlock, 2004). The emergence and increasing prevalence of this behavior during adoles cence suggest that self-injur y—in clinical or nonclinical settings—is, in part, a developmental phenom enon: aspects of the behavior (e.g., offers immediate reduction in stress), the indivi dual (e.g., difficulties regulating emotion and coping with stress), and the environment (e.g., so cial reinforcement) during this period of development have resulted in its spread (C onterio & Lader, 1998; Rosen & Walsh, 1989;

PAGE 15

2 Ross & Heath, 2002; Whitlock, Powers, & Eckenr ode, 2006). In addition to feelings of release, self-injury offers a dolescents benefits at a time when they are most receptive to influence, most impulsive, and most at risk fo r the negative effects of stress. Self-injury offers vulnerable adolescents, in particular, a way to deal with overwhelming affect and a sense of identity, enables self -expression, and fits with char acteristics of adolescents, including experimentation, imitation, and rebellion (Gladwell, 2000/2002). Although not undertaken for attention, self-injury also enab les youth to shock adults, certainly a perk for some. Although much is known about other adoles cent risk behaviors such as alcohol and tobacco use, little is known about self-injury among the general adolescent population (Purington & Whitlock, 2004). There are three types of direct self-injury: major (e.g., amputation), stereotypic (e.g., rhythmic head banging), and superficial/moderate (e.g., skin cutting) (Favazza, 1998). Favazza (1998) further broke superficial/moderate self-injury, which is the most common type of self-injury, into three types, episodic, repetitive, and compulsi ve. [For a comprehensive review of classifications of self-injury see Claes and Vandereycken (2007).] All three types share similar underlying reasons (e.g., tension relie f); however, they are differentiated by frequency and level of perceived importance to the individual (Strong, 1998). Self-injury has been studied in clinical settings for decades; however, few empirical studies have been conducted to identify the factors that co ntribute to the practi ce of self-injury among adolescents in a general population (Carls on et al., 2005; Purington & Whitlock, 2004). Increased attention will bring w ith it a demand for efforts to c ontrol self-injury, especially within school settings (Jesso r & Jessor, 1977). Before eff ective preventive interventions

PAGE 16

3 can be developed, however, more needs to be learned about the scope of self-injury among adolescents in community settings and fa ctors related to self-injury, especially those amenable to change and useful in identifying vulnerable youth. Research Problem For the most part, self-injury has been a pproached from a psychiatric or clinical framework (Johnstone, 1997). Most research has located self-injury within individuals and, thus, has offered clinical explanations and individua l-level solutions (Johnstone, 1997). Self-injury is a mental health issue, but it is not known whether all youth who self-injure have a diagnosable mental illness, whether self-injury is a sign of distress among vulnerable youth in clinical and nonclinical settings, and/or whether self-injury is a “new” expression of adolescent risk behavior that is being “labeled as risqu by adults in a particular historical and sociocultural setting” and becoming “normative” (Rew, 2005, p. 167). Preliminary evidence suggests th at increasing prevalence rates of selfinjury represent a cultural effect, with more recent c ohorts demonstrating higher prevalence rates than did earlier cohorts (e.g., Whitlock, Eckenrode, & Silverman, 2006). One clinician has associated the rise in self-i njury (and other expressi ons of distress) with the rise in mental and emotional disorder s among children of priv ilege (Levine, 2006). Parenting behaviors associated with privilege (e.g., overinvolv ement, intrusion, criticism, permissiveness) combined with growing up in a culture of affluence has resulted in many privileged children reaching adolescence with a sense of emptiness, an impaired sense of self, which translates, for some, to mental, em otional, and behavioral disorders (Levine, 2006).

PAGE 17

4 There have been many attempts to expl ain self-injury (Conterio & Lader, 1998; Ross & Heath, 2003). Most explanations sugge st self-injury is a maladaptive coping mechanism that provides relief from di stress (i.e., emotional regulation) and communicates what cannot be or is not verbalized; some yout h who lack healthier ways of coping with, or adapting to, stress or have difficulty expressing negative or overwhelming emotions (e.g., hostility, anxiety) use self-injury, a maladaptive coping behavior, as a form of emotional release and survival (Conterio & Lader, 1998; Ross & Heath, 2003; Yates, 2004). Within commun ity samples of adolescents, Yip (2005) suggested self-injury may be used, among othe r things, by adolescents to release tension, gain attention, and/or express their anger at institutions (e .g., schools, families) that seek to control them. This suggestion is similar to Wocjik’s (1995) desc ription of self-injury as rebellion, which has root s in the punk movement. Some researchers suggest selfinjury among adolescents is contagious, sim ilar to what has been known about suicide for years (Crouch & Wright, 2004; Fennig, Gabr ielle, & Fennig, 1995; Gladwell, 2000/2002; Rosen & Walsh, 1989; Taiminen, Kallio-S oukainen, Nokso-Koivisto, Kalionen, & Helenius, 1998; Walsh & Rosen, 1985). On ce tried, self-injury may ‘stick’ with vulnerable youth (Gladwell, 2000/2002; White Trepal-Wollenzier, & Nolan, 2002). The act of self-injury causes the body to release e ndorphins, which result in feelings of relief or release. This chemical reaction and associated release reinforces the behavior (i.e., automatic reinforcement). This process of use—reinforcement—compulsion over time is similar to that seen with other behaviors such as disordered eating and substance abuse. Others caution against viewing self-injury as an addiction because most studies have

PAGE 18

5 demonstrated self-injury is emotionally-based and stopping necessitates perceptions of personal control (Conterio & Lader, 1998). Prior to the 1990s, it was assumed that most individuals w ho tried self-injury discovered the behavior on their own (Adl er & Adler, 2005; H odgson, 2004; Purington & Whitlock, 2004). Among more recent cohorts, however, it is assumed that adolescents have been exposed to self-injury via some social venue (e.g., media, school) (Adler & Adler, 2005; Hodgson, 2004). However, there is a lack of empirical investigations into social influences on self-inj ury, including family and school experiences and exposure to self-injury models in the media and among peer s (i.e., peer contagion). Existing evidence suggests social contagion, or, as Marsden ( 2005) explained, “imitativ e behavior based on the power of suggestion and word of mouth influence,” has played a key role in the dramatic increase of self-injury amo ng youth (Crouch & Wright, 2004; Derouin & Bravender, 2004; Fennig et al., 1995; Lieb erman, 2004; Rosen & Walsh, 1989; Taiminen et al., 1998; Yates, 2004; Young, Sweeting, & West, 2006). Increasi ng media attention, especially since the late 1980s, more than li kely played a central role in tipping the behavior from aberration to social epidem ic. Much of the media that has included references to self-injury targets younger audiences (e.g., 7th Heaven Family Guy Girl, Interrupted ). Although media attention has the potentia l to reach out to youth in need of support with informal social support and res ources for recovery, Carlson et al. (2005, p. 22) and others (e.g., Yates, 2004) have argued th at increased attention without research or scientific information has resulted in a “cl imate of confusion”—self-injury has been normalized and vulnerable youth have been exposed to maladaptive coping behaviors (i.e., social contagion), yet adults and instituti ons are confused as to how best to respond.

PAGE 19

6 Rates of self-injury have increased expone ntially among adolescen ts; self-injury has ‘tipped’ (see Gladwell, 2000/2002 for a discussion of social epidemics). Jensen (2003) suggested that future re search in the area of psychopathology and comorbidity, among other aspects, should focu s on identifying subgroups, interactions associated with comorbidity, environments in which psychopathology is expressed, and the varying pathways to psychopathology. Specific to comorbidity, the theory of problem behavior in adolescence suggests al cohol use, tobacco us e, and other risk behaviors are comorbid among some yout h, possibly due to similar underlying explanatory factors (Jessor & Jessor, 1977; Rew, 2005). Prior research has demonstrated relationships among health-risk behaviors (“co -occurrence,” clusters of risk behaviors); however, to date, this author has been unable to locate empi rical studies conducted within community settings of early adolescents to examine relationships between self-injury and other risk behaviors. Jensen’s (2003) call for the identification of subgroups of individuals is consistent with the use of segmentation in public health and preven tion research. Segmentation refers to the division of an apparently heterogeneous populati on (i.e., dataset) into smaller “homogeneous segments” (John & Miaoulis, 1992, p. 131). The logic behind segmentation within public hea lth is to identify homogenous groups of individuals that will respond to “specific and efficient marke ting strategies designed to elicit particular responses” (John & Miaoulis, 1992, p. 131). Se gmentation is a hallmark of effective public health interventions: combined with audience research (e.g., qualitative research), it enables the identification of target audiences and effective strategies for reaching each with health prevention programming. Within the realm of self-injury, segmentation could

PAGE 20

7 be used to identify groups at risk of adopting self-injury as a maladaptive coping strategy and inform school-based prevention efforts. Conceptual Framework Overall, this study focused on moderate /superficial self-inj ury as a distinct behavioral phenomenon with assumed multiple causes and functions. A broad definition of self-harm, which includes multiple behavi ors noted among early adolescents, guided this study. For the purposes of this study, sel f-injury was defined as the performance of a harmful behavior such as cutting, scratchi ng, burning, not allowing wounds to heal, or pinching, by a person who feels upset as a way to feel better (less ups et). A distinction was not made between episodic and repetitive self-injury given the lack of available measures of psychological symptoms (i.e., indi cators of diagnosable mental illness) and impulsivity. To gain a comprehensive understanding of self-injury among early adolescents, literature from multiple fields was consu lted, including psychology, sociology, education, medicine, and public health. Further, multip le explanatory theories and concepts were considered such as problem behavior theo ry (Jessor & Jessor, 1977), social contagion (see Gladwell, 2000/2002; Marsden, 2005), behavi oral precipitants of self-injury (see Boyce, Oakley-Browne, & Hatcher, 2001; Crouch & Wright, 2004; Strong, 1998; Walsh & Rosen, 1988), developmental theory (i.e., developmental psychology), and behavioral frameworks such as automatic and social reinforcement (see Nock & Prinstein, 2004, 2005). Each of these is discussed in more deta il in Chapter 2. The conceptual framework and theories presented in Chapter 2 guided th e variable selection process. Because this study involved a secondary anal ysis of data obtained from the Youth Risk Behavior

PAGE 21

8 Survey (YRBS), the ability to measure key theories was limited. Ultimately, indicators of the following theories or concepts were identified: problem behavior theory (e.g., “Have you ever tried cigarette smoking, even one or two puffs?”), social contagion (“Have any of your friends hurt themselves on purpose (cutting, scratching, burning, not allowing wounds to heal, pinchi ng)?”), precipitants of self -injury (“During your lifetime, have you ever been cyberbullied?”), and de velopmental theory (Parent Communication). Research Purpose and Questions The purpose of this study was to provid e a description of self-injury within a general adolescent population. This research identified subgroups of self-injurers, identified behaviors associated with self -injury, explored relationships between the environment (e.g., peer, media) and self-injury, and suggested risk and protective factors associated with self-injury. Three broad que stions guided this disse rtation research: (a) What is the status of self -injury within a public middle school setting in terms of prevalence, frequency, exposure, and corre lates, including demographic (e.g., gender), attitudinal (e.g., attitudes toward school), a nd behavioral variables (e.g., having ever been bullied)? (b) How does self-injury relate to ot her risk behaviors, such as tobacco use, alcohol use, suicide, and deviance among yout h? and (c) What factors are useful in identifying meaningful subgroups (segments) of youth who are more likely to self-injure? Research Approach This study is a secondary anal ysis of cross-sectional, self-report data gathered from sixth-and eighth-grade students in eight mi ddle schools in a larg e, southeast county in Florida using the middle school Youth Ri sk Behavior Survey (YRBS). The middle school version of the YRBS is an anonymous survey used by the county school board to

PAGE 22

9 monitor risk health and risk behavior s among middle school youth and for prevention programming and evaluation purposes. The midd le school survey is used to monitor six categories of priority health-risk behaviors among youth and young adults: (a) unintentional and intenti onal injuries, (b) tobacco use, (c) alcohol and other drug use, (d) sexual behaviors that contribute to unintended pregnancies and sexually transmitted diseases, (e) unhealthy dietary behaviors, and (f) physical in activity (Kann et al., 1998). The 2005 middle school YRBS also included que stions about demographics, delinquent behaviors, communication/rela tionship with parents/guardi ans, exposure to prevention interventions, self-reported grades, and truthfulness of re sponses. Three items were developed to measure three aspects of self -injury: lifetime prevalence, past-30 day prevalence, and awareness of p eer self-injury behavior. Given the early state of the literature, this dissertation research focused on mining data for patterns and structure. The concep t of principled stat istical discovery, an iterative analysis approach that involves expl oring datasets, identifyi ng potential patterns or structure, and using further statistical tests and/or inform ation to confirm or disconfirm potential findings, guided the analysis (M ark, 2006). Descriptive and inferential statistics, including multilevel logistic regressi on analysis, were used to answer each of the three broad research questions. Particul ar attention was paid to exploring gender, sociocultural, grade and school-level varia tion with respect to the three dependent variables: having ever trie d self-injury, frequency of self-injury, and knowledge of friends who self-injure. There are numerous multivariate statistical approaches for looking for structure in social and behavioral data, including, for example, multiple regression, cluster analysis,

PAGE 23

10 discriminant analysis, logistic regression, log-linear modeling, and Chi-Square Automatic Interaction Detection (CHAID) an exploratory, criter ion-based response modeling technique (Dillon & Kumar, 1994). Although CHAID (Kass, 1980) has not received substantial attention within the realm of e ducational research and measurement or other fields (Hoare, 2004), it has been used by prev ention researchers to identify unique target audience segments (i.e., mutu ally exclusive and exhaustive subgroups) and has much to offer investigators interested in searching fo r patterns and structur e in large datasets (Hoare, 2004; Magidson, 1994). CHAID is a pred ictive cluster analysis approach in that a set of independent variables (i.e., predicto rs) are used to group participants based on their response to a categorical or polyto mous dependent variable. CHAID produces mutually exclusive and exhaustive segments that result from an itera tive, chi-square test of independence based analysis of the interactions among pr edictor variables, such as demographics, psychographics (e.g., attitudinal variables), and behavioral variables (Magidson, 1994). CHAID was selected for th e dissertation study described herein based on its use in the fields of marketing resear ch and public health, its appropriateness or match to the guiding research questions, and the ease in which potentially meaningful patterns in a dataset are identified in a dataset wi th a large number of variables. Once segmentation results were obtaine d, validity evidence was gathered through the use of theory and applied knowledge in interpreting the segmentation results (i.e., determining the number and nature of segmen ts/classes) and replicating CHAID analysis within a holdout sample (Aldenderfer & Blashfield, 1984; Magidson, 1994).

PAGE 24

11 Significance The overall prevalence of mental health disorders among youth is estimated to be 20% (Spear, 2000). Approximately 16% of boys and 19% of girls meet the criteria for one or more of the following mental illnesses: posttraumatic stress disorder (PTSD), major depression, and substance abuse or depe ndence (Kilpatrick et al., 2003). Further, approximately one third to one half of adolescents may report depressed mood or affective disturbance at any point in time (S pear, 2000). These estimates suggest that, at any one point in time, a substantial proporti on of youth lacking in support or adaptive coping skills may be at risk for trying self-injury. A sma ller subset of youth, for whom the behavior becomes repetitive, may develop a chronic behavioral condition that places them at increased risk for suicide and other long-term, negative outcomes (Hawton, Harriss, & Zahl, 2006; Hawton, Zahl, & Weatherall, 2003; McElroy & Sheppard, 1999; Patton et al., 1997; Shaw, 2002). Whereas much is known about certain risk behaviors such as tobacco and alcohol use, less is known about self-injury, a risk behavior that has taken hold among adolescents in today’s world (Purington & Whitlock, 2004). Currently, schools, mental health institutions, and clinicians suggest it is on the rise and many are at a loss for dealing with it—much less pr eventing it (Carlson et al., 2005; Galley, 2003; Purington & Whitlock, 2004). In a recent study of hospita lizations, Olfson et al. (2005) found a significant increase in the pr oportion of hospitalizations that involved cutting, hanging, and suffocating. More interestingly, Olfs on et al. (2005) found that total estimated inpatient costs for cutting, the most prevalen t form of self-injury, almost tripled in the

PAGE 25

12 past decade from 6.7 million in 1990 to 18.5 million in 2000, along with the proportion of hospitalizations for self-injury fr om 4.3% in 1990 to 12.2% in 2000. Effective primary, secondary, and tertiary prevention programming that addresses self-injury among adolescents could reduce th ese costs and others not yet estimated. Much needs to be learned, however, about the individual, cultural, social, and environmental risk and protective factors asso ciated with self-inj ury among youth in nonclinical settings (Purington & Whitlock, 2004) before effective prevention programs or strategies can be developed. Several aspects of this st udy distinguish it from prior research conducted on selfinjury among adolescents within general popu lations. First, this study used a clear definition of self-injury that was not conflated with atte mpted suicide. Second, selfinjury was studied using a larger, more di verse accessible population. Third, theories such as social contagion and problem be havior theory guided the development of research questions, analysis, and interpre tation of results, thereby, moving beyond a primary focus on psychological theories and variables in understa nding self-injury. Fourth, this study captured the prevalence of self-injury during a time period, early adolescence, when the behavior has been found to emerge (Adler & Adler, 2005; Favazza, 1998). Finally, this study did not a ssume youth who self-inj ure (“cutters”) are a homogeneous group, but rather attempted to identify subgroups within the population who are at risk for self-injury. In addition to contributing to the literature, this study was designed to inform the development of more effective prevention programming and practice. Study results tested the validity of using multivariate marketing approaches to identify segments of youth at risk for self-injury, and the l iterature review combined with

PAGE 26

13 study results were used to develop recommenda tions for the county where the data were gathered. Organization of Remaining Chapters The remainder of this document is divide d into four chapters. Chapter 2 provides a comprehensive review of the literature. Ch apter 3 presents the methods used in this dissertation research. Chapter 4 includes qua ntitative and qualitat ive results. Finally, Chapter 5 provides an overview and discus sion of key study findings, implications for prevention, and suggestions for future research. Definitions of Terms General Adolescent Population : This phrase refers to adolescents who are not in some form of clinical, residential, or juve nile institutional setting. For the purposes of this dissertation, adolescents who attend one of the eight middle sc hools are considered members of the general adol escent population. Individua l members of the general adolescent population may have a clinical diagnosis and/or receive services within the school setting. Prevalence : Prevalence refers to the total number or proportion of cases of a disease, condition, or behavior in a specific population at a specific point in time. Primary prevention : Primary prevention refers to any type of intervention designed to prevent a behavior or negative ou tcome before it occurs. Primary prevention efforts are geared to general populations. Protective factors : Community, school, family, a nd peer/individual level factors that protect against health or behavi oral problems during adolescence.

PAGE 27

14 Risk factors : Community, school, family, and peer /individual level factors that place youth at risk for developing a health or behavioral problem during adolescence. Secondary prevention : Secondary prevention refers to prevention that occurs among those at risk for performing a be havior or developi ng a disease. Tertiary prevention : Tertiary prevention refers to efforts targeted at those who have already adopted a behavior or have deve loped a disease with th e intent of ending the behavior and preventing rela pse, where appropriate, and re ducing the negative impact of the behavior or disease on i ndividual health and wellbeing.

PAGE 28

15 Chapter Two: Literature Review Introduction This chapter begins with an overview of early adolescence and theories used to explain adolescent risk behavior. This overv iew is designed to provide a developmental context for this study and jus tify the consideratio n of relationships between self-injury and other risk behaviors. Se lf-injury then is defined and an overview of the complex etiology of self-injury is provided. The prev alence and trends of self-injury during adolescence, including soci ocultural and gender varia tion, also are reviewed. Relationships between self-injury and adol escent development are discussed, with emphasis on the ways in which self-injury f its with the characteristics and goals of adolescence. The role of popular culture and social contagion in spreading self-injury is discussed. A literature-based discussion of intervention approaches and guiding principles is provided. Segmentation as an approach to identifying homogenous groups of individuals that will respond to public he alth interventions is discussed within the context of self-injury, and st atistical approaches to segmentation are presented. This chapter concludes with a synthe sis and application of the lit erature to the present study protocol. Early Adolescence Middle school-aged youth are in the de velopmental period referred to as early adolescence. Early adolescence, the peri od between 10 and 14 years of age, is

PAGE 29

16 characterized by a multitude of somewhat simultaneous biological, social, and psychological changes (Brooks-Gunn, 1988; El liott & Feldman, 1990; Simmons & Blyth, 1987; Smetana, 1988). Early adolescents assume a new role; they ar e no longer children, but they are not yet adults (Simmons & Blyt h, 1987). Decreased time spent with parents, increased emotional distance from parents, increased conflict over “mundane issues” (e.g., chores), and the desire to hold certain issues private and the related increase in “strategic disclosure” (i.e., car efully selecting what to disc uss with parents) characterize early adolescence (Dowdy & Kliewer, 1998; Paikoff & Brooks-Gunn, 1991). Developmental tasks that begin during this pe riod include establishing identity or selfimage, forming and negotiating peer relati onships, individuati on (i.e., establishing autonomy or individuality while remaining conn ected to parents), planning for the future, dealing with emerging sexuality, learning to interact with same and opposite sex peers, and dealing with conformity issues (C ooper & Cooper, 1992; Elliott & Feldman, 1990; Simmons & Blyth, 1987; Wakschlag, Pittm an, Chase-Lansdale, & Brooks-Gunn, 1996; West, Rose, Spreng, Sheldon-Keller, & Adam, 1998). Adolescents grow psychologically and socially during this period, with thos e who establish caring relationships, find acceptance and belonging, and experience ag e-appropriate intimacy experiencing healthier psychological and social deve lopment than do adolescents who do not (Baumeister & Leary, 1995; Reis & Shaver, 1988; Sullivan, 1953). Early adolescents must cope with numerous issues related to their developmental status, including physical and hormonal changes, sexual or romantic desi res or feelings, changes in parent-child relationships, increased expecta tions associated with their m ove into adolescence, school

PAGE 30

17 changes, and changes in social networks (P apini & Micka, 1991; Simmons, Burgeson, & Reef, 1988). Recent work in developmental neurology and related advances in neuroimaging suggest adolescent behavior is heavily infl uenced by brain development that occurs during this period (see Spear, 2000 for an excellent review of this literature). As is the early adolescent, the adolescent brain is in a state of transition, or as Spear (2000, p. 428) described, a “chronic state of threatened homeo stasis.” Adolescents use the skills they have gained thus far in life to naviga te a time of intense emotion, changes, and expectations. At this time, they have to le arn new skills that will serve them in adulthood (Spear, 2000). As part of this transitiona l period from child to adulthood, there are, on average, increases in social behavior or a ffiliation, risk taking, and/or novelty seeking, with boredom being a common complaint am ong early adolescents (Spear, 2000). A certain level of risk taking, although not always desirable fr om an adult’s (i.e., parent) perspective, is common and may aid youth in making the transition from youth to adulthood (Spear, 2000). For example, risk taki ng, for some, is associated with increased self-esteem and other positive outcomes such as increased knowledge of self and environment; however, it may serve as a m eans for affect regulation (i.e., selfmedication) or maladaptive coping (Spear 2000). Ultimately, the issue involves determining when a behavior become s something to prevent (Spear, 2000). The forebrain regions undergo substa ntial alterations during adolescence (National Institute of Mental Health, 2001; Spear, 2000). Thes e regions are sensitive to stress, placing youth, especially those with underdeveloped copi ng skills or who lack the resources to deal with stress, at risk for affective distur bances (e.g., depressed mood) and

PAGE 31

18 impaired decision making (Spear, 2000). When faced with stress or overwhelming emotion, early adolescents are le ss able than are older adolesce nts to react with reason or problem solving, tending more to react w ith fear and other ‘primitive’ responses (National Institute of Mental Health, 2001). The more life changes that happen during adolescence and the greater the perceived level of stress, the more likely some youth may feel overwhelmed (i.e., unable to cope) and e xperience distress or inner turmoil. Some youth, especially those lacking more adaptive ways to cope with perceived stress, may turn to risk behavior/s (e.g., drinking alcohol) and fail to perform healthy behaviors in response to distress (Spear, 2000). The biological, social, and psychologi cal changes that occur during early adolescence are related to the individuati on process, the primary developmental task associated with early adolescence. Cognitive changes, such as being able to think abstractly and consider multiple perspectiv es, enable adolescents to reason more effectively, and view their pare nts and their relationships in a new light. These changes are hypothesized to contribute to the transformation in th e parent-child relationship (Papini & Micka, 1991; Smetana, 1988, 1991). Biological changes that become readily apparent during early adolescen ce have demonstrated effects on parent-child interactions, particularly when mothers and/or fathers ar e uncomfortable with these changes (Hauser, 1991; Paikoff & Brooks-Gunn, 1991; Papini & Micka, 1991). Effects of the individuation process seem to be most disr uptive during early adoles cence, as evidenced by increased conflict, especially within moth er-daughter dyads, incr eased self-reports of parenting stress, and reports of diminished marital dissatisfac tion (Carlson, Cooper, & Spradling, 1991; Dowdy & Kliewer, 1998; Paikoff & Brooks-Gunn, 1991; Smetana,

PAGE 32

19 1988; Steinberg, 1990; West et al., 1998). Adolescence requir es a shift in parenting: parents must grow with their children (some more rapidly than others), understand and accept the goals of adolescence, and change thei r parenting behaviors to fit the needs of their transitioning child (e.g., rely more on e xplanation, curiosity, a nd problem solving) (Powers, Hauser, & Kilner, 1989). Theoretical Approaches to Adolescent Risk Behavior Individual adolescent problem (risk) behavior can be considered in isolation or in association with other know n problem behaviors (Rew, 2005). Much research has focused on identifying risk and protective fact ors associated with individual problem behaviors such as tobacco use. Information fr om this research has been used to devise interventions targeted at preventing the initia tion of individual risk behaviors. However, some researchers have adopted a more in clusive approach, one that views problem behaviors as related to one another (i.e., comorbid) (Jense n, 2003; Jessor & Jessor, 1977; Rew, 2005; Spear, 2000). Problem-behavior theo ry suggests that problem behaviors such as alcohol use, tobacco use, and others pe rformed during adolescen ce are expressions of similar underlying explanatory factors (Jesso r & Jessor, 1977). Key assumptions of problem-behavior theory include: the rela tionship between academic achievement and individual orientation to conventiona lity; the tension among independence and conventionality, regulation, and adult control; the purposive and instrumental nature of problem behavior; and the need to consider aspects of the individual adolescent, the multiple contexts in which the adolescent opera tes, and the larger society in which the adolescent performs the behavior (s ee Rew, 2005, pp. 169-170 for a review).

PAGE 33

20 Jessor (1991) reviewed em pirical evidence suggesting risk behaviors covary within individuals, which has been dubbed problem (risk) behavior syndrome. Jessor (1991) argued that youth who demonstrat e such a syndrome may be in need of interventions that focus at the lifestyle level rather than at the leve l of individual problem or risk behaviors. In terms of anteced ents, youth who demonstrate risk behavior syndrome tend to be less conventional and f unction within unconventional environments (Donovan & Jessor, 1985). Tests of problem-b ehavior theory have identified numerous other risk and protective factor domains asso ciated with problem behavior syndrome in adolescence, including psychosocial ad justment, school connectedness, family connectedness, and depression (see Rew, 2005). Self-injury Definitions of Self-injury There are many terms used to refer to self-injury, including the following: deliberate self-harm, cutting, self-abuse, self -injurious behavior (SIB), self-mutilation, auto-aggression, and parasuicide (see Cl aes & Vandereycken, 2007; Klonsky, Oltmanns, & Turkheimer, 2003; Strong, 1998). For the purpo ses of the present research, the terms self-harm and self-injury will be used synonymously. Self-mutilation, although some qualify with “superficia l,” carries with it a negative connotation or seriousness not generally demonstrated by youth w ho self-injure, and may be best used only when acts of major injury such as amputation are carried out (see Herpertz, 1995). Self-harm can be classified into two broad cat egories—direct or indirect (Laye-Gindhu & Schonert-Reichl, 2005; Suyemoto, 1998; Yates, 2004). Direct self-harm, which includes cutting, biting, severing, burning, and hitting, is of primar y interest in this study (Yates, 2004).

PAGE 34

21 Examples of indirect self-harm include ove reating and substance abuse (Yates, 2004). Menninger (1935) argued all indivi duals perform some type of non-fatal self-destruction; self-injurious behavior of bot h forms is not uncommon. There are three types of direct self -injury, including: major (e.g., amputation), stereotypic (e.g., rhythmic head banging), a nd superficial/moderate (e.g., skin cutting) (Favazza, 1998). Favazza (1998) further divide d superficial/moderate self-injury, which is the most common type of self-injury, in to three types, episodic, repetitive, and compulsive. Episodic self-injury tends to be associated with mental and personality disorders such as mood disorders, borderline personality disorder, eating disorders, and posttraumatic stress disorder associated with early adve rse experiences (e.g., sexual abuse). Repetitive self-injury, evidence suggest s, is an impulse control disorder or, as some argue, a stand alone behavioral phenomenon (e.g., Klonsky et al., 2003; Muehlenkamp, 2005). Repetitive self-injury is of concern to school administrators and teachers, because it is associated with a chronic condition that functions, in part, to provide a sense of identi ty (Carlson et al., 2005; Li eberman, 2004; Strong, 1998). Compulsive self-injury refers to behaviors that are more subconscious, such as skin picking and hair pulling. All three types sh are similar underlying motivations (e.g., tension relief); however, they are differen tiated by frequency and level of perceived importance to the individual (Strong, 1998). Within the present study, a distinction was not made between episodic or repetitive self -injury given the lack of available measures of psychological symptoms (i.e., indicator s of diagnosable mental illness) and impulsivity.

PAGE 35

22 Within the literature, several definiti ons of self-injury have been offered. Suyemoto (1998) defined self-injury as: …a direct, socially unacceptable, repe titive behavior that causes minor to moderate physical injury; when self-mutilating, the individual is in a psychologically disturbed state but is not attempting suicide or responding to a need for self-stimulation or a stereotypi c behavior characteristic of mental retardation or autism. (p. 532) Woldorf (2005) defined self-i njury as, “Deliberate damage to one’s body that is not culturally sanctioned, is not motivated by suicid al intent, and is meant to relieve intense negative emotions” (pp. 196-197). Muehlenka mp (2005) defined superficial/moderate self-injury as: …repetitive, low-lethality actions th at alter or damage body tissue (e.g., cutting, burning) without suicidal intent. Su perficial/moderate SIBs [self-injurious behaviors] have a unique set of symptoms are viewed as a ty pe of morbid selfhelp, and are exhibited by individuals with and without various mental disorders. (p. 324) Most definitions emphasize that se lf-injury is delibera te, distinct from suicide, and is not culturally sanctioned. Others, such as Mu ehlenkamp’s (2005) definition, specify forms (e.g., cutting), functions (e.g., aff ect regulation), and relationships with mental disorders. Suyemoto (1998) provides a simple defin ition that may apply to many individuals who self-injure within community se ttings: self-injury, she argued, is a temporary maladaptive coping mechanism. Support for this argument is found in the average developmental trajectories associated with depression, self-esteem and anger, all of

PAGE 36

23 which are associated with self-injury (see Br own, 2001 for a review of emotions and selfinjury). Depression, low self-esteem, and a nger peak during early adolescence when the gender gap between males and females is th e largest; however, on average, depression and anger decrease, self-esteem increase s, and the gender gap narrows during the transition to early adulthood, which correspond s to, on average, increased independence and greater emotional regulati on abilities (Galambos, Barker & Krahn, 2006). Overall, although there is some disagreement over wh at self-injury is (and is not); most researchers suggest self-injur y is a form of “morbid self -help” used during times of overwhelming distress or in connection with mental illness or early trauma (Conterio & Lader, 1998; Favazza, 1988; Yates, 2004) a nd, in some cases, is a separate impulse control disorder (e.g., Favazza, DeRosear, & Conterio, 1989). However defined, selfinjury is poorly understood, impacts youth, fami lies, schools, and society, and, evidence suggests, has taken hold within yout h culture (Nock & Prinstein, 2005). Historically, self-injury of ten has been mistaken for attempts at suicide (Favazza, 1998; Muehlenkamp & Gutierrez, 2004). Froe schle and Moyer (2004) argued this view is one of the several myths associated with the behavior (e.g., self -injury is used to manipulate others, self-injury is used to gain attention, and individua ls who self-injure are dangerous to others). Self-injury and suic ide are distinct, yet re lated phenomena: selfinjury is the strongest risk factor for suicide (Hawton et al., 2003). Self-i njury differs from suicide, according to Muehlenkamp (2005), in terms of intent, lethality, chronicity, and preferred methods (e.g., cutting vs. poi soning). Individuals who self-injure distinguish between self-injury and suicide; so me have described self-injury as a way to be in control, and suicide as being ou t of control (Solomon & Farrand, 1996). The

PAGE 37

24 distinction between self-injury and suicide also may be associ ated with attitudes toward life: adolescents who self-inj ure report less repulsion towa rd life than do those who attempt suicide (Muehlenkamp & Gutierrez, 2004). Shaw (2002) expanded on key differences between self-injury and suicid e: “In self-injury, I see confusion, pain, violation, protest and desperation, but also perseverance, a yearning for connection, a struggle to hold on to what is real and a mome nt primed for intervention” (p. 210). In an empirical investigation of differences betw een individuals who had attempted suicide with and without a history of self-injur y, Stanley, Gameroff, Michalsen, and Mann, (2001) discovered that those with a history of self-injury may underestimate the potential lethality of their suicide a ttempts. Among individuals who had attempted suicide, those with a history of self-injury reported higher levels of depression, hopelessness, aggression, anxiety, impulsivity, and suicide id eation than did those who did not have a history of self-injury (Stanley et al., 2001). It is important to note distinctions between self-injury as studied in this dissertation and self-harm as defined by the Child and Adolescent Self-harm in Europe (CASE). In this study, self-injury was define d as the performance of a harmful behavior such as cutting, scratching, burning, not allowing wounds to heal, or pinching, by a person who feels upset as a way to feel bett er (less upset). The CASE study’s definition of self-harm is as follows: “An act with a non-fatal outcome in whic h an individual deliberately did one or more of the following: Initiated behaviour (for example, self cutting, jumping from a height), which they intended to cause self harm

PAGE 38

25 Ingested a substance in excess of th e prescribed or generally recognised therapeutic dose Ingested a recreational or illicit drug that was an act that the person regarded as self harm Ingested a non-ingestible substance or object.” (H awton, Rodham, Evans, & Weatherall, 2002) The latter definition is more inclusive th an the latter—self-inj ury with and without suicidal intent are included, as is the ingestion of substances (ingestible and noningestible). Etiology and Functi ons of Self-injury Multiple pathways lead to self-injury (Tiefenbacher, Novak, Lutz, & Meyer, 2005).1 The etiology of self-injury may diffe r according to the population of interest— clinical or nonclinical. For example, with in clinical settings, sexual abuse has been identified as the single best predictor of self-injury (Strong, 1998). Whether this holds in nonclinical populations is unclear. Within community samples of adolescents, Yip (2005) suggested self-injury may be used, am ong other things, by a dolescents to release tension, gain attention, and/or e xpress their anger at institutions that seek to control them (e.g., parents, schools). This suggestion is sim ilar to Wocjik’s (1995) description of selfinjury as rebellion, which has roots in the punk movement (e.g., Sex Pistols). Tiefenbacher et al. (2005) sugge st two developmental pathways to self-injury: one that begins with genetic or biol ogical risk and the other that begins with adverse early 1 Self-injury associated with suicide attempts, need for self-stimulation, or conditions such as mental retardation or autism is excluded from this discussion of etiology in order to be consistent with Suyemoto’s (1998) guiding definition.

PAGE 39

26 experiences. Whether these two pathways ar e mutually exclusive is arguable given the complex interrelationships between genetic or biological vulnerability, certain mental illnesses (e.g., depression, bipolar disorder), a nd increased susceptibility to the negative impact of early adverse expe riences. In some cases, self -injury may be a symptom of mood or personality disorders that when treated abates (Woldorf, 2005), whereas in others it may be a risk beha vior experimented with by a curious, exposed adolescent either during times of stress or in response to some developm ental need or drive (Derouin & Bravender, 2004). Shaw (2002) summarized the complex etiology and trajectory of self-injury among women: Self-injury is not only protest or resistance. It is a product of culture as well as physiology, unconscious processes, trau matic experiences, life events and environmental triggers. Paradoxically, se lf-injury may at once be a symbol of protest, a marker of violations, a catha rsis and a behavior through which women unwittingly engineer their own incarcerati on as they become entrapped in an isolating cycle of self-abuse. (p. 209) The concept of “biological fragility” (vulne rability) is important to consider when discussing self-injury (Conterio & Lader, 1998). Although approximately one-half (or more) of individuals who self -injure report a history of a buse or maltreatment, many have no such history (Conterio & Lader, 1998; St rong, 1998). Thus, some practitioners have recognized that self-injury, even within clin ical settings, has multiple causes, one of which may be the tendency for some indivi duals to be emotionally “hypersensitive” (Conterio & Lader, 1998). Th is hypersensitivity is descri bed by some as an innate temperamental influence on developmen t (Conterio & Lader, 1998). Although

PAGE 40

27 environment plays an important role in deve lopment, perhaps even greater than that played by individual-level factor s, especially during adolescen ce, biological fragility may play a key role in clinical and nonclinica l settings—one that would explain why selfinjury ‘sticks’ with some but not with others (Conter io & Lader, 1998; Gladwell, 2000/2002). Based on Suyemoto’s (1998) review of the functions of self-injury, Klonsky (2004) identified the following seven models of self-injury: (a ) the interpersonalinfluence model, (b) the self -punishment model, (c) the antisuicide and sexual models (i.e., drive models – self-injur y reduces these drives), (d) th e affect regulation model, (e) the dissociation model (i.e., self-injury stops dissociation), and (f) the interpersonal boundaries model (i.e., self-injury serves to create or delineate boundaries between self and other). In reality, more than one of th ese models may explain the initiation and/or maintenance of self-injury within any one individual. Klonsky (2004) did not address directly the environmental model, which pos its self-injury is modeled and reinforced within the child’s immediate e nvironment. Children learn th at behavior is rewarded (e.g., tension relief, attention, sympathy), imitate the behavior, and experience reinforcement (Suyemoto, 1998). Klonsky (2004) provided one of the first tests of the functions of deliberate self-harm. Using a semi-structu red interview assessing the functions and consequences of self-injury and feelings asso ciated with self-injur y episodes of selfinjury, Klonsky (2004) interviewed 39 college students who had repetit ively self-injured. Results supported the affect regulation model as the prim ary functional motivation for self-injury. Secondary functions of self-injury such as self-punishment, interpersonal influence, and sensation-seeking also were identified (Klonsky, 2004).

PAGE 41

28 In many cases, the act of self-injury repres ents a cry for help (i.e., self-injury as communication) and a quick means of emotional regulation (i.e., self-inj ury as self-help). Self-injury serves as a language of sorts (i.e., words, history, and experiences written on the body) that allows individuals who self -injure to communicate psychological distress (Abrams, 2003; Austin & Kortum, 2004; Cont erio & Lader, 1998; Derouin & Bravender, 2004; Harrison, 1997). Whereas some individuals externalize distress (i.e., ‘act out’) through some form of defiance such as fighti ng, substance use or sexual behavior, others internalize distress through be haviors such as self-injury (Abrams, 2003). Individuals who self-injure report that it offers quick relief fro m overwhelming affect, racing or chaotic thoughts, depersonaliz ation, anxiety, and emotional distress (Adler & Adler, 2005; Favazza, 1998; Favazza & Conterio, 1989; Solomon & Farrand, 1996; Woldorf, 2005). Self-injury also has been associated with relief from gu ilt, rejection, boredom, hallucinations, and sexual preoccupation, whic h is a symptom of bi polar disorder among adolescents (Favazza, 1998). Self-injury conve rts emotional distress into physical pain that is within the control of the self-injurer (Liebling, Chipchase, & Velangi, 1997; Solomon & Farrand, 1996). In addition to pr oviding a means of self-soothing, self-injury leaves behind marks, scars, or wounds that tell a story of pain that either cannot be verbalized or has been ignored or trivialized by others (Shaw, 2002; Solomon & Farrand, 1996; Woldorf, 2005). All of these models (except the ‘selfinjury as rebellion’ model) share a foundation in the clinical literature. Klons ky’s (2004) study supported the validity of the affect regulation model within a community se tting. However, the sample was small and limited to repetitive self-injurers. Whether clinical models that link self-injury to

PAGE 42

29 diagnosable mental illness and/or trauma w ill remain valid within nonclinical populations remains uncertain. The emergence and increa sing prevalence of this behavior during adolescence, however, suggests self-injury—in clinical or non clinical settings—is, in part, a developmental phenomenon: aspects of the behavior (e.g., offers immediate reduction in stress), the indi vidual (e.g., difficulties regulat ing emotion and coping with stress), and the environment (e.g., social reinforcement) during this period of development have resulted in its spread (C onterio & Lader, 1998; Rosen & Walsh, 1989; Ross & Heath, 2002; Whitlock, Powers, & Eckenrode, 2006). Young et al. (2006) used a longitudinal cohort design to study the factor s that predict self-injury among Scottish youth, serving as one of the first—if not the first—longitudinal examination of self-injury within a general population. Unfortunately, par ticipants were not recruited into the study until they were 11 years of age, thereby precl uding the ability to examine factors that occurred earlier in the developmental trajectory. Results suggested that self-reported identification with the Goth subculture was the strongest predicto r of self-injury and suicide attempts, even after controlling for other factors examined (Young et al., 2006). Additional significant predictors of self-injury included ge nder (i.e., being female), parental divorce or separati on, smoking, and other substance use (excluding alcohol), and a history of depression (Young et al., 2006). In a retrospective, cross-sectional study involving undergraduate and graduate student s in the general population, Whitlock et al. (2006) found that, when controlling for dem ographic characteristics, self-reported emotional or sexual abuse, having ever cons idered or attempted suicide, elevated psychological distress, and characteristics of eating disorders were associated with repetitive self-injury. In addition to reporting greater di stress and poorer psychological

PAGE 43

30 functioning than did their nonself-harming peers, youth who self-injured reported, on average, greater repulsion with life, greater a ttraction to death, and less attraction to life (Muehlenkamp & Gutierrez, 2004). For more information on the etiology of self-injury consult Conterio and Lader (1998), Favazza (1998), Lloyd-Richardson, Perrine, Dierker, and Kelley (2007), Suyemoto (1998), Wals h and Rosen (1988), and Yip (2005). Prevalence and Trends of Se lf-injury during Adolescence For the most part, estimates of the prev alence of self-injury during adolescence and early adulthood have been calculated wi thin clinical settings or using small, convenience samples. Among clinical popula tions, approximately 20% of patients or clients self-injure, with higher rates among sp ecific groups (e.g., 32% of individuals with eating disorders) (Dieter, Nicholls, & Pearlman, 2000; Solano, Fernandez-Aranda, Aitken, Lpez, & Vallejo, 2005). Even though the behavior is said to emerge during early adolescence (13 to 14 years of age), fe w studies have focused on self-injury during early adolescence within community settings (Muehlenkamp, 2005). Estimates of the general prevalence (including adults) varies from a low of 750 per 100,000 persons per year (0.75%) (Yates, 2004) to a high of 1.7% (Patton et al., 1997). The prevalence of self-injury among adults may be similar to the estimated pr evalence rates (~1%) of other disorders such as eating disorders and bipolar disorder (American Psychiatric Association Work Group on Eating Disorders, 2000; Narrow, 1998; Regier et al., 1993). However, Briere and Gil (1998) suggested 4% of the general population in the United States may self-injure. Three studies conducted within community settings documented similar rates of having engaged in self-injury amo ng adolescents: 15% (Laye-Gindhu & SchonertReichl, 2005), 16% (Muehlenkamp & Gutie rrez, 2004), and 14% (Ross & Heath, 2002)

PAGE 44

31 (see Table 1). Lloyd-Richardson et al. (2007) found that 46.5% of adolescents reported some form of non-suicidal self -injury (NSSI), 60% of whom (28% of the entire sample) reported moderate/severe forms of NSSI (e.g., cutting/carving skin). Rates of hospitalizations for self-injury have increased, as has the be havior, with a rate of 4.3% among youth hospitalized in 1990 to 12.2% of youth hospitalized in 2000 (Olfson et al., 2005). Cutting (wrist or arm) is the most common form of self-injury (Favazza & Conterio, 1989; Hawton et al ., 2003; Ross & Heath, 2002). Table 1 Sample of Self-Injury Measures Used with A dolescents and Associated Prevalence Rates Study* Measure Prevalence Laye-Gindhu & SchonertReichl (2005) Have you ever done anything on purpose to injure, hurt, or harm yourself or your body (but you weren’t trying to kill yourself)? Followed by open-ended questions about specific behaviors 15% Lloyd-Richardson et al. (2007) FASM – A checklist of non-suicidal self-injury asking whether participants had practiced each of 11 self-harm behaviors. 46.5% Muehlenkamp & Gutierrez (2004) Self-harm Behavior Scale – open-ended, free response scale 5 items on methods of self-harm (i.e., cutting, scratching, burning, selfhitting, punch/kicking, banging, and other) 15.9% Ross & Heath (2002) Screening: hurt themselves on purpose (Likert scale) Semi-structured, followup interview: elaborate on hurting themselves on purpose Urban School Screening: 21.2% Follow-up: 13% Suburban School Screening: 19.6% Follow-up: 14.8% *All studies were conducted with high school students.

PAGE 45

32 The duration of self-injur y varies anywhere from a single experimentation to chronic, repetitive self-injury lasting a decad e or more (Favazza, 1998; Suyemoto, 1998). According to Suyemoto (1998), for most a dolescents, self-injury may be a temporary coping mechanism. Most adolescent females who self-injure eventually stop, with most stopping the behavior at around 18 to 19 years of age (Suyemoto, 1998). One-half of those who had self-injured had done so on a minimum of 50 different occasions (Favazza & Conterio, 1989). Muehlenkamp (2005) sugge sted self-injury becomes repetitive at around five or more times. Favazza (1998) s uggested switching between occasional and repetitive self-injury occurs at different times for different individuals. In addition to increased lifetime prevalence rates of self-inj ury, some have reported an increase in the frequency of repetitive self-injury (Hawton, Fagg, Simkin, Bale, & Bond, 1997). Sociocultural and Gender Variation There is a lack of information on socioc ultural and gender vari ation in self-injury prevalence and frequency within community sa mples. Traditionally, self-injury has been reported to be a White, female, middle-to -upper middle class issue (Abrams, 2003; Conterio & Lader, 1998; Ross & Heath, 2002). However, this may represent a sampling artifact: White, female inpatients have b een over-represented in clinical studies (Suyemoto, 1998). However, self-injur y may represent a symptom of the disproportionate rates of depression, anxiet y, and substance use disorders among children of privilege in the United St ates (see Levine, 2006 for a re view). Parental pressure combined with growing up in a culture of affluence has resulted in many privileged children reaching adolescence with a sense of emptiness and a lack of core self, which translates, for some, to mental, emotional, and behavioral disorder s (Levine, 2006). On

PAGE 46

33 the other hand, sociocultural vari ations in vulnerability to suicide, depression, and eating disorders suggest ethnic groups, particularly low-income, Hispanic females, may be at increased risk for self-injury (Abrams, 2003). However, studies (e.g., Muehlenkamp & Gutierrez, 2004) have been limited by insufficien t numbers of particip ants within ethnic groups to study variation. As with gender differences in other expressions of emo tional distress (e.g., depression), there may be gender differences in self-injurious be haviors and underlying motivations (Laye-Gindhu & Schonert-Reichl, 2005). The performance of self-injury may vary across genders. For example, wher eas girls may self-injure when alone, boys also may self-injure when in the company of others (Laye-Gindhu & Schonert-Reichl, 2005). There is a lack of information on self-injury among males due to their underrepresentation in clinical se ttings (Gratz, 2003; Laye-Gindh u & Schonert-Reichl, 2005). Ross and Heath (2002) hypothesized that differe nces in the prevalence of self-injury between males (36%) and females (64%) in their sample represen ted preferences for different coping behaviors that are not new (i.e., internalizat ion versus externalization). However, research conducted among male inpa tients (Winters, 2005) suggested that rates of self-injury among males is on the rise, indicating either an increase in distress and related depression among males and/or the infl uence of media exposure to self-injury on males’ choices of coping behaviors. Inte restingly, Muehlenkamp and Gutierrez (2004) found no statistically significant gender diffe rences in self-injur y rates among high school students in a community setting. Goodman ( 2005) suggested repetitive self-injury may be more common among females; however, th is has not been established empirically within community settings.

PAGE 47

34 Shaw (2002) suggested self-injury is gendered, representing women’s internalization of cultural objectification a nd violence; through se lf-injury, females recreate and control the violence that is infl icted on them every day in the media, at school, and in their own homes (Shaw, 2002). Within a feminist framework (e.g., Shaw, 2002), girls use self-injury as a way to reflect back to societ y the violence that has been perpetuated on them (e.g., objectification, vi olence in the media and home). The body becomes their means of expression, with some youth carving words or symbols into their flesh (Derouin & Bravender, 2004; Suyemoto, 1998). Shaw (2002) argued self-injury may be “uniquely distressing because it reflects back to the culture what has been done to girls and women” (p. 208). Self-injury and Adolescent Development To understand why self-injury emerges and peaks during adolescence, one must understand the prevalence of emotional di sturbances during a dolescence, biological characteristics of early adolescents, the deve lopmental characteristics and tasks of early adolescence, and the role that self-inj ury plays during adolescence (e.g., benefits, precipitants). Studies of psychopathology in community samples of adolescents suggest that the prevalence of severe emotional disturbance ranges from 10% to 20%, which represents the percenta ge found in the adult population (Kilp atrick et al., 2003; Powers et al., 1989; Suyemoto, 1998). However, Spear (2 000) pointed out that approximately onethird to one-half of adolescents may report depressed mood or affective disturbance at any point in time. A substantial proportion of youth may be at risk for self-injury and other risk behaviors because of early experiences that do not equip them with the skills and resources necessary for navigating adolescenc e, such as affect re gulation in the face

PAGE 48

35 of overwhelming experiences, self-soothing behaviors, and dealing with sexuality (Suyemoto, 1998). Self-injury is bodily communication of trauma or emotional distress and a “sign that something has gone wrong in the development of self-regulatory functioning and the separation-indivi duation process” (Hemme, 2001, p. 647). Impulsivity (i.e., urgency, lack of prem editation, lack of perseverance, and sensation seeking) and aggression peak during adolescence in association with developmental and neurological changes (d’Acremont & Van der Linden, 2005; Muehlenkamp, 2005; Spear, 2000). Neurological changes that occur during adolescence increase adolescents’ sensitivity to stress and result in poorer decision making when compared to adults (Spear, 2000). In additi on to impulsivity, internalizing problems also increase during adolescence, with differ ences in girls and boys becoming pronounced beginning with the transition from chil dhood to early adolescence, and girls demonstrating higher levels th ereafter (Bongers, Koot, van de r Ende, & Verhulst, 2003). Increased distress among girls once they reach adolescence has b een noted (Gilligan, 1991; Pipher, 1994). Differences in internal ization (‘anger in’) vers us externalization (‘anger out’) are associated with gender diffe rences in preferred coping styles, which may help to explain greater rates of self-injur y among females than males (Bongers et al., 2003). Whereas females tend to ‘act in’ and de monstrate self-destructive behaviors such as self-injury, males tend to act out, be have aggressively, an d ‘accidentally’ hurt themselves (e.g., punching a hole through th e wall) (Clarke & Whittaker, 1998). Self-injury emerges during adolescence because it fits in well with the conflicts and developmental issues associated with this phase of life (C rouch & Wright, 2004; Suyemoto, 1998). These include the tension between needing and not wanting help, the

PAGE 49

36 struggle for autonomy (i.e., individuation), self -definition (i.e., identity formation), the tension between disclosure and privacy, fear of rejection versus the need to be understood, and affect regulation during a ti me of marked physiological and social change. When shared within a group setti ng, whether a clinical setting (e.g., mental health ward) or community setting (e.g., Go th subculture), self-injury may offer group cohesion, acceptance, and understanding (Crouch & Wright, 2004; Machoian, 2001; Muehlenkamp, 2005; Young et al., 2006). Most acts of self-injury are precipitated by a sense of loss, interpersonal conflict or perceived rejection, or is olation (Boyce et al., 2001; Crouch & Wright, 2004; Walsh & Rosen, 1988; Strong, 1998). Relationship a nd communication difficulties between parent and youth may place some youth at risk for se lf-injury (Derouin & Bravender, 2004). Discord between parent and youth peaks duri ng early adolescence, with greater tension noted in mother-daughter relationships (C arlson et al., 1991; Dowdy & Kliewer, 1998; Paikoff & Brooks-Gunn, 1991; Smetana, 1988; Steinberg, 1990; West et al., 1998). Interestingly, in a community sample of adolescents, Lloyd-Ric hardson et al. (2007) found that getting “a reaction” from another person was one of the most common reasons cited for deliberate self-harm. Further, a dolescents move away from parents and toward their peers as a part of the individuation pro cess, thereby setting the scene for some youth to experiment with self-injury when expos ed within their peer networks (Derouin & Bravender, 2004). In addition, youth exposur e to media increases substantially during the teenage years, which may lead some yout h to attempt the behavior on impulse when exposed to self-injuring models on the Inte rnet or in the media (Teens Health, 2005; Whitlock et al., 2006).

PAGE 50

37 Self-injury can be linked to adolescents’ search for identity and truth, for a sense of self (see Erickson’s and G illigan’s theories of adolescent development; Shaw, 2002). Self-injury offers one solution to the struggle for identity; some youth who try self-injury self-identify as “cutters,” “bur ners,” or “emo” (i.e., emotional). Adolescents may even distinguish between ‘genuine ’ self-injurers and those ‘f aking’ the behavior (for belonging) using criteria such as level of physical damage and secrecy (Crouch & Wright, 2004). Whereas self-inj ury offers some benefits a nd meets developmental needs, being labeled as a ‘cutter’ or ‘burner’ or being linked to groups known for high rates of self-injury (i.e., Goths; see Young et al., 2006) may further traumatize vulnerable youth and place them at risk for developing a ch ronic behavioral condition (Adler & Adler, 2005; Johnstone, 1997; Machoian, 2001). The youth may be labeled as manipulative, attention seeking, and severely emoti onally disturbed (Machoian, 2001). The investigator’s personal experience with self-injury among adol escents has uncovered disturbing youth backlash against other yout h who self-injure (“cu tters”) on social networking sites such as www.vampirefreeks.com a popular social networking site among Goths. One virtual ‘cult’ open to members of www.vampirefreeks.com “fuck_emo,” made available banners that read, “Next time cut deeper” against a background of superficial cuts indicative of self-injury. Youth easil y could save and add these banners to their own site, thereby leading to the spread of images of self-injury and backlash against youth who se lf-injure. In addition to poems and other narratives romanticizing the benefits of self-injury (i.e ., “problems flow away as the blood flows”), there is evidence that some adolescents have begun to create j okes about self-injury, which may serve to normalize the behavior:

PAGE 51

38 Question : Why are ‘emo’ lawns the best? Answer : They cut themselves. Cutting is an effective yet maladaptive way for adolescents to release their frustration, gain relief from tens ion, gain attention (i.e., someone to listen to them), and express their anger towards people and inst itutions charged with controlling them— schools, parents, and society (Yip, 2005). Self-injury serv es as a way to communicate distress and may result in improved relationships with parents, in a subset of cases (Hilt, Borelli, Nock, & Prinstein, 2004). Evidence su ggests adolescents who self-harm differ from those who do not in help-seeking, communication and choice of coping (Evans, Hawton, & Rodham, 2005). Comp ared to those who did not self-harm, youth who selfharmed were more likely to need help but not seek it, were less “able” to talk with social network members (e.g., teachers, family), ha d fewer groups they could turn to for support, were more likely to choose avoida nt coping over problem focused coping, and were more likely to turn to their frie nds for support (Evans et al., 2005, p. 585-586). Although a desire for control may precipita te many cases of self-injury among youth, ironically, self-injury often resu lts in a loss of control (Liebl ing et al., 1997). Adolescents may discover the behavior becomes compul sive (i.e., difficult to control without intervention) over time, and, if discovered, youth who self-injure may be considered a danger to self and others by schools, families, and clinicians. This may result in their freedoms being limited by concerned and of ten uninformed/misinformed adults and institutions (Carlson et al., 2005; C onterio & Lader, 1998; Shaw, 2002).

PAGE 52

39 Popular Culture and Self-injury The increased prevalence of self-inj ury among youth, especially over the past several years, suggests a cultura l trend. Thus, self-injury can not be considered separate from the cultural and historical period in which it occurs (Clarke & Whittaker, 1998; Johnstone, 1997; Kleinman 1988; Oliver, Hall, & Murphy, 2005). Feminist researchers such as Harrison (1997) and Shaw (2002) suggest ed self-injury is a natural yet admittedly maladaptive reaction to living in “a harming so ciety – a society that seeks to control and maintain us” (p. 438). Levine (2006) associated self-injury with the culture of affluence that leads to disconnection, emptiness, and depression among adolesce nts and adults. In addition to individual, familial, and community level influences on expressions of distress (i.e., internalizati on versus externalization), cu lture impacts an individual’s preferred method of expression. Culture cr eates the options and reinforces their expressions (Abrams, 2003; Gladwell, 2000/2002). Using the body as a “bulletin board for the frustrations and feelings that have gone ignored” is not a new phenomenon (Conterio & Lader, 1998, p. 11). Body modifi cation involving breaking of the skin has occurred since the beginning of recorded hi story (Conterio & Lader, 1998). Self-injury as defined herein is not the same pheno menon as piercing or ta ttooing. Although these behaviors have in common piercing of th e skin, the behaviors are differentially motivated—piercing and tattooing represent a desire to care for the body and may actually protect against self-injury (C laes, Vandereycken, & Vertommen, 2005). There is a lack of empirical investiga tions into aspects of popular culture that have contributed to increased rates of self -injury. Derouin and Br avender (2004) suggest the high rate of separation and divorce (appr oximately 50% of marriages end in divorces;

PAGE 53

40 Bramlett & Mosher, 2002 ) may place youth at increased risk for self-injury. Combined with high rates of separation and divorce, yout h in recent cohorts have had to cope with increasing levels of stress and violence in thei r lives (e.g., media, community), placing vulnerable youth at increased risk for self-i njury through internationalization of violence and social learning (exposure) (Derouin & Bravender, 2004). Conterio and Lader (1998) suggested several factors may be related to increasing rates of self-injury, including disconnection at the familial and community levels (e.g., extended families live apart, youth spend more time alone when not in schoo l), reductions in talking with confidants and increases in acting on emotions, incr eased reliance on technol ogy, the ‘quick fix’ nature of our culture, emphasis on addiction, less time with family and more time with peers, a focus on appearance, and gender bi as. Gender bias or living within a ‘girlharming’ culture may help to explain highe r rates of self-injury among females than males; by adolescence, girls are angry, afrai d, and frustrated (Cont erio & Lader, 1998; Pipher, 1994). Although self-injury has been studied fo r several decades, media attention has increased substantially since the late 1980s to early 1990s (Adler & Adler, 2005; Derouin & Bravender, 2004). An anthology of self -injury in the media can be found at http://anthology.self-injury.net/ A recent study of self-injur y on the Internet discovered more than 400 self-injury message boards dedi cated to self-injury, most of which were developed within the past five years (Whitlock et al., 2006). Much of the media that have included references to self-injury target s younger audiences (e.g., 7th Heaven, Family Guy, Girl, Interrupted). Princess Diana was one of the earliest (1996) famous individuals to talk of her personal struggl e with self-injury (Derouin & Bravender, 2004). Since that

PAGE 54

41 time other celebrities, many popular with yout h, have discussed their experiences with self-injury, including but not limited to: Johnny Depp, Angelina Jolie, Fiona Apple, Marilyn Manson, and Christina Ricci. The Intern et is riddled with web sites devoted to self-injury, and attention has increased in the news, advice columns, personal narratives, the research literatur e, and novels (e.g., Cut ) (Shaw, 2002). Today’s adolescent cohort in the United States (i.e., GenTech, GenM) is wired (~80% use the Internet; 50% acce ss the Internet daily); they are technologically savvy and use the Internet to express themselves and connect socially (Becker, 2000; Gross, 2004; Lenhart, Madden, & Hitlin, 2005; Roberts, Foehr, & Rideout, 2005). The average 8 to 18 year old is exposed to 8.5 hours of media a day, with an average of 6.5 hours of direct media use per day (Roberts et al., 2005). Although the average total media use among adolescents has not changed signi ficantly from 1999–2004, time spent using computers has more than doubled during this time, from an average of 27 minutes per day to just over one hour per day (Roberts et al., 2005). Relative to media exposure during a typical day (i.e., 8.5 hours), youth reported spending just over two hours per day “hanging out with parents” and just over two hours per day “hangi ng out with friends” (Roberts et al., 2005). Exposure to media violen ce, which is present in high levels in a substantial portion of media to which youth are exposed, has been linked to increased verbal and physical aggression (O’Keefe, 2002) Whether self-injury, in particular, is associated with media exposure is unknown. Whereas studies have suggested that Internet use may decrease social isolation among youth and help them connect with likeminded others and assume different identi fies (Maczewski, 2002; Suzuki & Calzo, 2004), at least one investigator has suggested the increasing prevalence of self-injury

PAGE 55

42 ‘communities’ on Internet message boards and web sites devoted to self-injury, in full or in part, may serve to fuel the behavior among adolescents (Whitlock et al., 2006). Although media attention has th e potential to reach out to individuals in need of support with informal social support and res ources for recovery, Carlson et al. (2005, p. 22) and others (e.g., Yates, 2004) have argued th at increased attention without research or scientific information has resulted in a “clim ate of confusion”—self-injury is normalized and vulnerable individuals are exposed to ma ladaptive coping behavior (i.e., social contagion) yet adults and inst itutions are confused as to how best to respond. Whereas most adults exposed to self-injury during adu lthood may react with hor ror or an inability to understand when exposed to self-injury, ad olescents, who tend to be drawn to dramatic and romantic notions of death and dying, are more susceptible to behavioral contagion and may find self-injury attr active (Gould, 2001; Muehlenkamp & Gutierrez, 2004). For example, Whitlock et al. (2006) found that In ternet message boards are most frequently populated with messages of informal support a nd discussions of self -injury triggers. However, the Internet also provides “access to a virtual subculture of like-minded others,” exposure to explicit content (ideas, suggestions), and connect ions to sources of pro-self-injury sites (e.g., site s that serve as self-injury t echnique information), and may serve to normalize and encourage the behavior (Hodgson, 2004; Whitlock et al., 2006). Social Contagion & Self-injury Existing evidence suggests self-injury has in creased dramatically due, in part, to the dynamic of social contag ion (Crouch & Wright, 2004; De rouin & Bravender, 2004; Fennig et al., 1995; Hodgson, 2004; Lieber man, 2004; Rosen & Walsh, 1989; Taiminen

PAGE 56

43 et al., 1998; Yates, 2004; Young et al., 2006).2 According to Marsden (2005), social contagion refers to “imitative behavior ba sed on the power of suggestion and word of mouth influence.” Social contagion is “subtle,” working through imitation and “permission to act from someone else who is engaging in a deviant act” (Gladwell, 2000/2002, p. 223). Emotions, behaviors, and ideas all can spread vi a social contagion (Marsden, 2005). One branch of social co ntagion research has focused on identifying aspects of the person and the behavior that affect contagion (e.g., Marsden, 1998). Gladwell (2000/2002) built on the social contag ion literature base in his national bestseller, The Tipping Point. According to Gladwell (2000/ 2002), three characteristics interact to explain the spr ead of emotions, behaviors, and ideas through a culture: contagiousness, the idea that little changes or causes can trigger big effects, and geometric rather than gradual change. The idea of geometric or dramatic shifts in cultural trends is referred to as the “tipping point” (Gladw ell, 2000/2002, p. 9). Efforts to explain why some epidemics “tip” (i.e., take off) and others falter must address three factors, including: (a) charac teristics of individuals who transmit the emotion, behavior, or idea; (b) aspects of the emotion, behavior, or idea that make it attractive or “sticky”; and (c) the environment in which the poten tial contagion is transmitted (Gladwell, 2000/2002). The spread of an emotion, behavior, or id ea through a culture se rves as a form of communication, a form of advertisement of sorts. For example, within the realm of selfdestructive behaviors, social contagion posits that messenge rs who perform the behavior serve to advertise one potential response to dealing with life’s challenges (Gladwell, 2 For a case study of social contagion and adolescent ri sk behaviors see Gladwell (2000/2002, pp. 216-252).

PAGE 57

44 2000/2002). Self-aggressive behavi ors, in particular, serve as a shared language among a particular group of individuals at a particular point in time (Gladwell, 2000/2002). Contagiousness is a function of the “messenge r”; thus, efforts to stop the spread of a behavior must consider aspects of the messenge r’s personality to whic h others are drawn. Thus, in addressing self-inj ury, one would need to iden tify aspects of individuals transmitting the self-injury message that make them attractive sources of information. These may include traits that are known to be attractive to yo uth: rebelliousness, impulsivity, risk-taking, precociousness, and indifference to others (Gladwell, 2000/2002, p. 232). Whether a behavior sticks is a function of the message and the person exposed to the behavior (Gladwell, 2000/2002, p. 232). Self -injury may be particularly sticky for adolescents because it offers a way to deal with overwhelming affect and a sense of identity, horrifies parents and adults, enables self-expression, and fits with characteristics of adolescents, including experimenta tion, imitation, and rebellion (Gladwell, 2000/2002). In other words, self-injury is a simple yet powerful way to meet numerous psychological needs at once (Strong, 1998). Cutt ing may be especially effective in aiding the individuation process because wounds are visible and disturbing to adults who may not be familiar with the behavior and may react with horror and disbelief. Within Gladwell’s (2000/2002, p. 268) framework, self-i njury may represent an “epidemic of isolation” in that it makes sens e only to those within the group performing the behavior. Whether a behavior becomes repetitive for a particular individual is dependent upon the individual’s initial reac tion. This is the reason why highly addictive substances such as heroin or nicotine are “only addi ctive in some people, some of the time”

PAGE 58

45 (Gladwell, 2000/2002, p. 235). Differences in the number of individuals who report trying self-injury and the sma ller subgroup who continue on to repetitive or compulsive self-injury reflect this differential stickiness. This initial reaction to the behavior then becomes a key time point for intervention—some youth will cut once and move on, whereas others cut once and find it works. Individual level characteristics, such as genetics, biological frailty, attitudes a nd beliefs, and early adverse experiences, determine, in part, whether and to whom self-injury sticks. Though evidence suggests that social contag ion or social learning theory plays a role in initiation of se lf-injury, whether self-injury associ ated with social contagion (e.g., peer influence) differs in meaningful ways from self-injury studi ed within clinical settings (e.g., self-injury associated w ith abuse and/or psyc hopathology) is unknown (Yates, 2004).3 Rosen and Walsh (1989) discover ed evidence of social contagion among adolescents (i.e., adolescents imitated th e self-injury behavior of group leaders). Fennig et al. (1995) suggested self-injury in the school envi ronment may differ from that found within clinical settings. This is sim ilar to Austin and Kortum’s (2004) discussion of the “traits” of adolesce nts who self-injure, includi ng, for example perfectionism, intelligence, moodiness, body image issues, inab ility to tolerate intense feelings, and difficulties expressing feelings or needs. In their study, most youth who self-injured were high functioning socially and academically but exhibited internalizing traits (e.g., anxiety)—not severe emotional disturbance. A more recent study supported this finding among college students at Ivy League institu tions: 17% of undergraduate and graduate 3 Exposure to self-injuring models is not necessar y for experimentation with self-injury, as some individuals who self-injure report accidentally discover ing the power of self-injur y to alleviate distress (Hodgson, 2004; Strong, 1998).

PAGE 59

46 students self-reported lifetime self-injurious behavior, with 36% reporting that no one knew of their behaviors (Whitlock et al., 2006). Although the secret or privat e nature of self-injury ha s been emphasized, evidence of social contagion indicates self-injury during a dolescence may not be as private as the literature would suggest (see Adler & Adle r, 2005 for a discussion of the social transformation of self-injury). Adler a nd Adler (2005, pp. 348-349) used a sociological framework to explain differences between th e secretive self-injurer (“loner deviant”) typically described in the lit erature and youth who self-injur e in private but share their experiences with members of their social network (“individual de viant”). Adolescent developmental theory suggests adolescents may share evidence of self-injury with some people and not others using, perhaps, the same criteria used when selectively disclosing parts of their lives to parents, peers, and other members of th eir social network. Further, the infiltration of self-injury into popular culture over the past two decades suggests the social unacceptability of self -injury may be giving way to some level of tolerance (Adler & Adler, 2005). This is not to say that adolescents who self-inj ure do not attempt to manage their deviant identities (i.e., stigma management) by hiding their injuries (e.g., wearing long sleeves), creating stories to ex plain their injuries (e .g., a cat scratch), or accounting (i.e., justifying) for their self-i njurious behaviors (A dler & Adler, 2005; Hodgson, 2004). Some adolescents who self-i njure (“individual deviants”) may be surrounded by “fellow deviants” who share their views of self-injury (i.e., the benefits, motivations) (e.g., Goths; Young et al., 2006), which may make it difficult for them to cease the behavior (Adler & Adler, 2005, p. 372). Being surrounded by their “fellow

PAGE 60

47 deviants” confirms the “deviant identity” and makes it difficult for some adolescents to stop self-injuring and adopt healthier coping behaviors (Adler & Adler, 2005, p. 372). Traditionally, attention has been rejected as a primary motivator for self-injury within clinical settings, alt hough attention is certainly a si de-effect of the behavior. However, evidence suggests that whereas automatic reinforcement (e.g., the sense of relief) may drive repetition of the behavior for some, social reinforcement (e.g., attention, sympathy) may, in part, explain the shift be tween experimentation and repetition (Nock & Prinstein, 2004; Oliver et al., 2005). This tendency toward social reinforcement may be one factor that differentiate s self-injury as discussed in the clinical literature (i.e., clinical psychology) from self-injury as disc ussed in non-clinical settings (i.e., middle school setting), which begs the question of isolation and privacy—a key assumption made in the literature. Are youth in non-clin ical settings aware of self-injury among their peers? Are there some youth who try se lf-injury during middl e school or beyond for attention (“fakes”; Ta iminen et al., 1998) and some who self-injure ‘legitimately’ (Crouch & Wright, 2004)? What are youths’ reactions to other youth who self-injure (e.g., social reinforcement, isolation)? S hould schools remain quiet (“reluctant”) about the issue and isolate those who self-inj ure to prevent cont agion (e.g., Derouin & Bravender, 2004; Lieberman, 2004) if a sizable proportion of youth are already discussing the behavior and awar e of its presence among their pe ers (Fennig et al., 1995)? Youth spend more time with their peers th an ever before; they are connected 24/7 via cell phone, Internet, telephon e, and face-to-face contact at school and other locations (Roberts et al., 2005). Peer contagion refers to peer influence on the spread of behavior. Peer contagion works through competition and false consensus bias (i.e., thinking more

PAGE 61

48 peers are performing a behavior than actua lly are) (Dishion & Dodge, 2005). Although scant research has examined empirically th e relationship between peer contagion and self-injury, there is a body of literature that o ffers insight into how self-injury may spread among adolescents (e.g., Dishion & Dodge 2005; Hartup, 2005; Prinstein & Wang, 2005). For example, the effects of peer co ntagion may be greatest among youth who are not at the extremes of deviancy ; youth who are in the middle or sitting on the fence, so to speak, may be at increased risk for ‘catching’ risk behavior s from their peers (Dishion & Dodge, 2005; Hartup, 2005). Further, mixed groups of youth demonstrate higher levels of peer contagion than do ‘pure’ groups (i.e ., deviant youth). Thus public school settings where there is a mixture of deviance levels with most youth bei ng not at the extreme levels of deviance, represen t potential breeding grounds for th e spread of health risk behaviors such as self-i njury (Hartup, 2005). What makes some youth vulnerable or sus ceptible to peer co ntagion is not well understood; what is known with certainty is there are numer ous individual level factors that may be related to vulnerability, which ma y or may not be specifi c to the behavior of interest (Hartup, 2005). The literature does highlight the importance of considering relationships within social networks and social norms (i.e., shared beliefs, attitudes) when studying peer contagion (Hartup, 2005). De velopmentally, behaviors present before adolescence (e.g., tendency to be overwhelmed when faced with intense emotion) may be amplified within peer groups (Hartup, 2005). Peer contagion must be considered in association with the way in which relationships are formed; individuals select their peers based, in part, on the ways in which they ha ve been socialized (Hartup, 2005). Basic social psychology suggests like individuals tend to gravit ate toward one another and

PAGE 62

49 develop relationships. During adolescence, this tendency is demonstrated in the formation of groups, such as Goths, Preps, a nd Skaters. Within the realm of aggression and deviance, aggressive or deviant youth who spend time with one another tend to be more aggressive or deviant th an they would if left to th eir own devices (Hartup, 2005). The mechanisms underlying this tendency are not well understood; suggestions have included modeling, coaching, and deviancy talk (Dishion, Spracklen, Andrews, & Patterson, 1996). Joiner (2003) suggested that selection or assortive relating may be responsible for bringing indivi duals vulnerable to suicide in to contact with one another (i.e., similar people cluster together before self-injury occurs). A recent longitudinal cohort study suggested identification with the Goth subculture was th e best pred ictor of having self-injured or attempted suicide (Young et al., 2006). The authors suggested selection and modeling effects we re at play in the initiati on and spread of self-injury among youth; vulnerable youth are more attract ed to the Goth subculture and, once ‘accepted’ into the culture, were at increased risk for adopting self-injury when exposed (Young et al., 2006). Once adopted, affiliation with a deviant identity—Goth or cutter— may make it difficult for youth to adopt a hea lthier identity (Adler & Adler, 2005). In addition to competition (i.e., one-upmanship; Crouch & Wright, 2004), false consensus bias, or the tendency for some adolescents to overestimate the prevalence and/or frequency of health risk behavior am ong their peers, plays a role in behavioral contagion (Prinstein & Wang, 2005). Affilia tion with similar others may partially explain this phenomenon (Prinstein & Wang, 2005). For example, Goths who ‘hang out’ together may be surrounded by a number of you th in their peer gr oup who self-injure, which may lead them to overestimate the number of youth who self-injure, thereby

PAGE 63

50 normalizing the behavior within this group. Adolescents also may conform to the perceived ‘leaders’ of their peer group, imitating the behavi ors of those they respect (Prinstein & Wang, 2005). Behavi oral conformity offers adolescents benefits, such as the avoidance of social ‘sanctions’ and incr eased self-esteem (Prinstein & Wang, 2005). When the behavior is not consistent with the individuals’ values, they will either terminate the behavior or align their values, beliefs, and attitudes to be consistent with performance of the behavior (Prinstein & Wang, 2005). This may help to explain why some individuals experiment with self-injur y, whereas others shift from experimentation to behavioral adoption. Among a dults and institutions, the choi ce to remain silent versus intervene may encourage the latt er (Prinstein & Wang, 2005). Behavioral Correlates of Self-injury Whereas comorbidity between self-injur y and psychological disorders has been established (e.g., eating disord ers; Favazza & Conterio, 1989; Solano et al., 2005; Strong, 1998), there is reason to believe self-injury may be related to other risk behaviors. For example, given the relationship between low serotonin levels and cigarette smoking, one would expect to see a relationship between self-injury and cigarette smoking (Malone, Waternaux, Haas, Cooper, Li, & Mann, 2003). Also, alcohol use may increase disinhibition and risk taking, setting the stage for se lf-injury (McCloskey & Berman, 2003). It is important to note, however, that within clinical samples, at least, alcohol or other substance use is not a necessary c ondition for self-injury to occur (Nock & Prinstein, 2005). Although suicide and self -injury are distinct phenomena, a substantial proportion of those who self-harm commit suic ide; thus, a relati onship among suicidal ideation, planning, and attempts and self -injury would be expected (McElroy &

PAGE 64

51 Sheppard, 1999). Antisocial behaviors (e.g., vi olence) also have been associated with self-injury (Patton et al., 1997) Self-injury is an impulsive behavior; thus, relationships with other impulsive behavior s including alcohol, substance use, suicide, shoplifting, skipping school, and so on would be expect ed (Lieberman, 2004). However, one study failed to support relationships between se lf-injury and other impulsive behaviors including alcohol abuse, st ealing, and suicide attempts (Solano et al., 2005). Psychological distress has been associated with health risk behavior s such as unprotected sex, sex with multiple partners, dating vi olence, smoking, weapon carrying, attempted suicide, and poor health (Rew, 2005). Assuming self-injury is a symptom of psychological distress, it should be associated wi th other health risk behaviors that have demonstrated relationships with psychological distress. Prevention and Intervention Given the impulsive nature of self-i njury, Goodman (2005) questioned whether self-injury can be preven ted before it occurs, and how to prevent youth who experiment with self-injury from becoming repeaters. In tervening in the self-injury process may be especially difficult because most cases of self-injury go undetected and without intervention (Whitlock et al., 2006). Whereas self-injury does not ‘stick’ with most who try it, efforts to teach alternative copi ng behaviors (i.e., primary and secondary prevention) and intervening before self-inj ury becomes compulsive or repetitive (i.e., tertiary prevention) should be made given th e relationship with suic ide and other negative outcomes (Hawton et al., 2006; Hawton et al., 2003; McElroy & Sheppard, 1999; Patton et al., 1997). If in most cas es self-injury emerges during ea rly adolescence, efforts to prevent self-injury should begi n as early in the developmenta l trajectory as possible (e.g.,

PAGE 65

52 through supporting parents in teaching emotiona l regulation skills, identifying vulnerable youth), making recommendations (e.g., Muehlenkamp & Gutierrez, 2004) to focus primary prevention efforts on high school-a ged youth misguided. Primary prevention must occur before the behavior has had a ch ance to stick; by high school, risk and protective factors associated with self-injur y have been established, and many youth have already experimented with self-injury, with a smaller proportion having already switched into repetitive self-injury. There is currently no public healthor population-based approach to the primary prevention of self-injury, which is not surpri sing given the current st ate of the literature (Hawton et al., 1997). Studies of peer contag ion associated with other risk behaviors suggest when adults remain silent, youth a dopt more favorable attitudes toward deviant and health risk behaviors and tend to overe stimate the number of youth performing them (Prinstein & Wang, 2005). Failing to implemen t primary prevention programs targeted at the promotion of adaptive coping behaviors and reliance on after-the-fact interventions that rely on isolation and tr eatment of youth who are alrea dy self-injuring may facilitate the spread of the behavior. As with resear ch conducted within th e realm of media and suicide risk (see Gould, 2001), researchers shoul d attempt to identify ways of addressing self-injury within non-clinical settings that do not romanticize the behavior or make it attractive to vulnerable youth. Further, the current literatu re base, along with empirical research such as that reported herein, could be used to guide the development of primary and secondary prevention programming. For example, the review conducted for this dissertation suggested the following prelimin ary prevention recommendations:

PAGE 66

53 Address self-injury openly (Suyemoto, 1998) but with caution due to potential contagion (Lieberman, 2004). Limit cont agion through taking a ‘low key’ approach that focuses on identifying a nd treating those practicing the behavior and preventing social contagion among th eir peers (Derouin & Bravender, 2004). Given the potential for triggering th e behavior among vul nerable youth, avoid holding assemblies about sel f-injury (Lieberman, 2004). Incorporate a self-injury component in suicide prevention strategies, including screening for self-injury along with su icide risk (Hawton et al., 2003; LayeGindhu & Schonert-Reichl, 2005). Support interv entions that offer alternatives for dealing with the emotional demands of the environments in which middle school youth are situated (Ross & Heath, 2002) Educate adolescents on how to help frie nds who have tried self-injury or are having emotional problems because adolesce nts who self-injure are most likely to rely on their friends for help (Evans et al., 2005). Reposition self-injury as an unacceptable, pathological behavior—not romantic, desirable, or positive (Suyemoto, 1998), a behavior that goes against the goal of adolescence (e.g., self-injury is an imita tive behavior) (Taiminen et al., 1998; Walsh & Rosen, 1985), and a behavioral ch oice (Saxe, Chawla, & Van Der Kolk, 2002). Teach youth skills, such as problem solving, emotional regulation, affect tolerance, and ways to meet safety and comfort needs (Crouch & Wright, 2004; Gratz, 2003; Laye-Gindhu & Schonert-Rei chl, 2005; Suyemoto, 1998). Offer

PAGE 67

54 positive alternatives to self-injury, including opportunities for group cohesion (Crouch & Wright, 2004; Suyemoto 1998; Taiminen et al., 1998). Support parents of adolescent youth th rough sharing knowledge of adolescent development, the cultural tr end of self-injury, and th e transitional nature of adolescence and praising c ontinued efforts to support their adolescents (Derouin & Bravender, 2004). Train those who come into contact w ith youth who self-injure, including counseling professionals (Zila & Kiselica, 2001). Adults should be trained in appropriate demeanors (i.e., nurturing) to take on when dealing with youth who self-injure because evidence suggests that adults perceived as uncaring, overprotective or intrusive, or uninformed undermine intervention effectiveness (Huband & Tantam, 2004). Offer support, including intervention a nd treatment, for those who self-injure (Suyemoto, 1998), with potentially differe nt approaches required for boys and girls (Laye-Gindhu & Schonert-Reichl, 2005). Factors that contribute to relationships between soci oeconomic deprivation and suicide (i.e., mediating fact ors) also may impact self-i njury, including family factors (e.g., genetics, family instability, lack of fam ily support, mental illness, unemployment); peer groups; violence and bullying; educa tion and the school environment; nutrition; smoking and substance abuse; and housing (e.g., overcrowding, crime) (Ayton, Rasool, & Cottrell, 2003; Gunnell, Peters, Kammer ling, & Brooks, 1995). Thus, policies that address socioeconomic deprivation and relate d mediators may be helpful in reducing the prevalence of self-injury.

PAGE 68

55 The complex interplay among the numerous factors that have contributed to the spread of self-injury among adolescents w ill make selecting a prevention approach difficult. Should we target the messengers (e.g., isolate and treat youth who are selfinjuring)? Try to make the behavior less stic ky (e.g., reposition it as an imitative behavior that goes against the adolescent desire to be unique)? Modify the environments in which youth interact (e.g., lower stress levels, elimin ate social reinforcers)? Lessons learned from efforts to prevent other risk behaviors s hould inform efforts to address self-injury. For example, having adults tell youth to ‘just say no’ to self-injury would most certainly make the behavior more attractive. Sec ond, equating experimentat ion with addiction should be avoided (Gladwell, 2000/2002). Ra ther than trying to “tackle the whole problem at once” (i.e., the war against drugs approach), efforts should attempt to “make sure experimentation doesn’t have se rious consequences” (pp. 250–251). Although a substantial proportion of youth may experime nt with self-injur y once exposed, it will only stick to a smaller proportion of vul nerable youth. Focusing on the early identification of vulnerable youth and teachi ng/modeling adaptive coping skills may be a more effort-, time-, and cost-effective appro ach than a universal approach (Gladwell, 2000/2002). Most interventions discusse d in the literature are clini cal in nature (see Brown, 2001 for a review), which is not surprising gi ven the number of studies conducted within clinical settings. There is currently a lack of empirical evidence to support effective treatments for deliberate self-harm, including repeat suicide attempts and self-injury (Hawton et al., 1998). Specific therapeutic approaches reco mmended in the literature include: problem solving therapy, dialectic behavior therapy (Lin ehan, 1993), cognitive

PAGE 69

56 therapy, behavioral therapy, and anger manage ment therapy (Boyce et al., 2001; Jones & Daniels, 1996; Milligan & Waller, 2001; Yate s, 2004; Zila & Kiseli ca, 2001). A mixture of approaches based on the needs of each indi vidual youth who self-injures may represent the best approach (Zila & Kiselica, 2001). Yip (2005) advocated for a multidimensional intervention with emphasis on the social en vironment, including supportive parents and peers, teaching youth to handle frustration a nd anger and regulate emotions in positive ways, and nurturing youth with the goal of developing their self-image and promoting their competence. At the core of any inte rvention designed to a ddress self-injury once established, is an effort to ‘cure a cure’ (Yat es, 2004). A review of the literature suggests efforts to ‘cure a cure’ should: Identify individual vulnerabilities (e. g., attitudinal, emo tional, relational), consider developmental and current expe riences, and offer training and support in the adoption of skills needed to ameliorate vulnerabilities (e.g., affect regulation, interpersonal) (Yates, 2004). Foster the development of a relationsh ip with active listening, talking, understanding, caring, compassion, patience, modeling of alternative ways of coping and assertiveness, and enc ouragement of self-expression and individualism (Austin & Kortum, 2004; Derouin & Bravender, 2004; Huband & Tantam, 2004; Liebling et al., 1997; Zila & Kiselica, 2001). Recognize self-injury as a maladaptive su rvival strategy and offer alternatives (Boyce et al., 2001; Harrison, 1997) Focus on support and teaching alternatives/skills—not th e cessation of self-injury (Derouin & Bravender, 2004; Saxe et al., 2002; Solomon & Farrand, 1996; Suyemoto, 1998). Avoid reliance on

PAGE 70

57 relaxation techniques because they ma y make self-injury worse (Huband & Tantam, 2004). Decrease environmental stress through fo stering bonds with parents and friends and reducing triggers of self-injury, es pecially social problems (Boyce et al., 2001; Derouin & Bravender, 2004), and identify and address behavioral reinforcers (Suyemoto, 1998). Address diet issues, such as caffeine cons umption, that can affect anxiety; employ efforts to prevent substance abuse, which can decrease inhibiti ons and alter mood; and screen for depression and anxiety, wh ich may be ameliorated with the use of appropriate psychotropic medication (Boy ce et al., 2001; Derouin & Bravender, 2004). Each school should have a protocol or internal plan for addressing self-injury (Onacki, 2005). School staff including teach ers, counselors, nurses, and security personnel need training in recognizing the signs of self-injury, listening and empathizing with students, and adopting a nurturing postu re (Froeschle & Moyer, 2004; Lieberman, 2004; Onacki, 2005). Further, staff should be trained to release students from class when negative emotions emerge (Froeschle & M oyer, 2004). Lieberma n (2004) recommended incorporating training into th e school’s crisis team respons ibilities. Once students who self-injure are identified, teachers are require d to refer students for further assessment, and schools are required to repor t self-injury to parents becau se students are considered a danger to themselves (Froeschle & Moye r, 2004; Lieberman, 2004; Onacki, 2005). Further, Froeschle and Moyer (2004) emphasized the need to report suspected abuse.

PAGE 71

58 In addition to external counseling an d therapeutic support, schools offer an essential environment in which students who self-injure can r eceive resiliency or skills training. Depending on the underlying motiva tion for self-injury, youth who self-injure could benefit from a number of individua l or group counseling foci, including selfesteem, grief, loss, divorce, assertiveness tr aining, substance abus e (including alcohol), and/or anger management (Froeschle & M oyer, 2004). Specific skills that may ameliorate the dependence on self-injury as a coping mechanism include: problem solving, interpersonal skills, distress toleran ce, and emotion regulation (Suyemoto, 1998). Johnstone (1997) discussed a need for deve loping partnerships with youth who selfinjure, with emphasis placed on understandi ng feelings versus physical action and behavioral choices, the mean ing youth place on self-inj ury, cultural influences on individual behavior, and gi ving youth a voice in interventi ons. Froeschle and Moyer (2004) emphasized the need to create a s upportive environment for youth that offers alternatives means of empowerment, enc ourages youth to voice their feelings, and models appropriate ways of handling negative affect. Parents and communities play integral role s in youths’ lives, and, thus, must be considered when addressing self-injury. Supporting parents of youth who self-injure should be a part of each school’s external pl an (Onacki, 2005). At a minimum, parents should be notified of their yout h’s self-injurious behavior and provided with resources (Froeschle & Moyer, 2004; Lieberman, 2004). Pa rents can play an important role in their children’s recovery thro ugh participation in counseling an d/or family therapy and needed support in how to deal with the behavi or and communicating with their children (Froeschle & Moyer, 2004; Suyemoto, 1998). Schools need to collaborate with parents

PAGE 72

59 and clinicians to ensure supportive connections among youth, schools, parents, and communities (Lieberman, 2004). Community involvement through local parent organizations, agencies, and churches could be used to reach parents through identifying and supporting speakers and training for pa rents and community members (Onacki, 2005). Segmentation Available evidence suggests that individuals who self-injure do not represent a homogeneous group. More than likely th ere are smaller homogeneous subgroups, or segments, of individuals who self-injure that share traits in common (e.g., motivation for self-injuring, preference for self-injury behavi or). Segmentation is the process used to divide an apparently heter ogeneous population (i.e., dataset) into smaller “homogeneous segments” (John & Miaoulis, 1992, p. 131). The logic behind segmentation within social marketing in public health is to identify homogenous groups of individuals who will respond to “specific and efficient marketing st rategies designed to elicit particular responses” (John & Miaoulis, 1992, p. 131). According to Yankelovich and Meer (2006), “good segmentations identify the gr oups most worth pursuing – the underserved, the dissatisfied, and those likely to make a first-time purchase” (p. 124). Within the realm of self-injury, segmentation provides a way to identify groups at risk of adopting self-injury as a maladaptive coping strategy and inform school-based prevention efforts. Segmentation is a hallmark of effective public health interventions. Social marketing, a strategy employed by some pub lic health professionals, relies on segmentation to identify target audiences and effective strategies for reaching each with health prevention programming. Principles of social marketing include the following:

PAGE 73

60 segment the target audience into homogene ous groups, analyze characteristics that discriminate segments, such as knowledge, atti tudes, social norms, and behavior; identify communication channels specific to each segm ent, develop strategies based on analysis of characteristics of each segment, and pret est materials and interventions with members of each segment (Slater & Flora, 1991). Segmentation, when undertaken well, can “improve the reach, utilization, and effectivene ss of health interventions” (Slater & Flora, 1991, p. 222). Rather than segmenting groups based on general attitudes, beliefs, personal characteristics, and psychographics (e.g., lifestyle se gmentation schemes), Yankelovich and Meer (2006) argued that segmentation stra tegies should reflect the “relationships of consumers to a product or product [behavior] category” (p. 124). In other words, emphasis should be placed on cons umer behavior and what this behavior reveals about the consumer (Y ankelovich & Meer, 2006). There are two basic approaches to statistical segmentation: a priori and clusterbased (Malhotra, 1989). In a priori segmentation, segmentation variables and categories are determined before data are gathered (M alhotra, 1989). In clus ter-based segmentation approaches, responses to a number of vari ables are used to determine segments (Malhotra, 1989). There are numerous variab les used to segment heterogeneous groups into smaller, homogenous groups, including general observable variables such as demographic variables, produc t (behavior)-specific observa ble variables, such as frequency, general unobservable variables such as values, beliefs, and attitudes, and product (behavior)-specific unobs ervable variables, such as benefits, preferences, intentions, and so on (Vriens, 2001, p. 5).

PAGE 74

61 Chi-square Automatic Interaction Detection (CHAID) The present study focused on mining data for patterns and structure. Although there are numerous statistical approaches for looking for structure in social and behavioral data, such as multiple regressi on, factor analysis, multidimensional scaling, discriminant analysis, logistic regression, and log-linear modeling, and for segmenting a population, such as cluster analys is and latent class analysis this dissertation used the following multivariate approach: Chi-Square Automatic Interaction Detection (CHAID), an exploratory, criterion-based response modeling technique (Dillon & Kumar, 1994). Procedures such as CHAID can be categorized into predictive and descriptive approaches to finding structure in data (Vriens, 2001). CHAID is a predictive cluster analysis approach in that a set of independent va riables (i.e., predictors) are used to group participants based on their res ponses to a categorical or poly tomous dependent variable. CHAID was selected based on its use in the fields of marketing research and public health, its appropriateness or match to the guiding research questions, and its ability to handle a large number of variables and identi fy potentially meaningful patterns in a dataset. Although CHAID (Kass, 1980) has not receiv ed substantial atte ntion within the realm of educational research and measuremen t or other fields (Hoare, 2004), it has been used by social marketers to identify unique audience segments (i.e., mutually exclusive and exhaustive subgroups) to target with public health interventions (Hoare, 2004; Magidson, 1994). CHAID is a hierarchical, criterion-based appro ach to segmentation that defines segments based on combinati ons of predictor variables (Magidson, 1994; Vriens, 2001). CHAID results in mutually ex clusive and exhaustive segments that result

PAGE 75

62 from an iterative, chi-square test of independence based analysis of the interactions among predictor variables, such as dem ographics, psychographi cs, and behavioral variables (Magidson, 1994). A lthough CHAID is used with ca tegorical variables, it was initially modeled on stepwise analysis of variance (Kass, 1980). Traditionally, CHAID has been used to create segments based on pr edictors of a single categorical, criterion variable; however, recent methodological wo rk has resulted in a hybrid algorithm for using CHAID and latent class analysis to se gment using multiple, correlated dependent variables (see Magidson & Vermunt, 2005). As a criterion-based model, CHAID is si milar to regression in that it is designed for prediction purposes (Magidson, 1994). W ithin the CHAID analysis approach, the initial sample is considered one segment (Vriens, 2001). Th is large, initial segment, which consists of all respondents, is por tioned into subgroups (segments) based on interactions among predictor va riables, which will, by defi nition, predict the criterion variable. For example, a segment may form based on the interaction between age and ethnicity where the criterion variable is res ponse to a diabetes sc reening opportunity. One possible finding may show African Americans between the ages of 25 and 35 are most likely to respond (i.e., be screened) to a diabetes screeni ng opportunity. Unlike regression analysis, CHAID assumes that the pr edictor variables will interact and enables the investigator to identify the most si gnificant predictors from a large number of possible predictors, thus simplifying the interpretation of complex interactions (Magidson, 1994). CHAID has three options for categoriz ing predictor types, including free, monotonic, and floating. The choice between predictor types determines how categories

PAGE 76

63 are merged (Magidson, 1994). Ordinal variab les are typically treated as monotonic; in other words, only those categories of a vari able that are adjacent can be merged (Magidson, 1994). Free variables are those variables that have no inherent ordering, such as occupation. Thus, whether free variable categories are combined does not depend upon adjacency (Magidson, 1994). Floating variab les are similar to those classified as monotonic, with the exception of the la st category (e.g., missing, unknown), which is combined with the category that is most alike in terms of distribution (Magidson, 1994). Magidson (1990, 1994) provides an overvi ew of the basic steps in a CHAID analysis of categorical data. Overall, there are three basic components of a CHAID analysis: the categorical or pol ytomous dependent variable, a set of predictor variables, and settings for CHAID paramete rs, including variab le classifications (e.g., floating) and stopping criterion (i.e., smallest segment size) There are three steps to the CHAID algorithm, including merging of categories base d on their similarity in relation to the dependent variable, splitting the overall group on the ‘best’ predicto r (i.e., the lowest statistically significant, Bonferroni adjusted p -value), and returning to the merging step if the stopping criterion has not been met or there are more subgroups to analyze (Magidson, 1994, p. 124). The merging step is th e most complex. Categories are merged within and across independent variables (Vri ens, 2001). Two-way cr oss-tabulations are formed between each independent variable and the dependent variable, categories are merged where appropriate, a nd the Bonferroni adjusted p -value is calculated for the merged cross-tab (Magidson, 1994; Vriens, 2001). The results of a CHAID analysis are presen ted in the form of a tree diagram (see Figure 1) and a gains table is produced that ranks each segment in terms of its likelihood

PAGE 77

64 of response to the behavior of interest (e.g., re sponse). Tree diagrams consist of a root node, parent nodes, child nodes, and terminal nodes (segments), each of which provides the following information: the category that defines the group, percentage response for the particular group, and the sample size for the group (Magidson, 1994, p. 128). Settings for parent and child node size depend, in part, upon av ailable sample size: within smaller sample sizes, minimum sample size se ttings are typically 10 for parent node and 5 for child nodes, and, within larger sample sizes, minimum sample sizes can be set at 20 for parent node and 10 for child nodes (The Measurement Group, 1999-2005). Figure 1 represents a segmentation tree with only one predictor variable, gender. Within this diagram, differing prevalence rates between ma les and females are represented (i.e., 15% among males, 35% among females) and the to tal sample size and the sample size per gender are displayed. Figure 1. Sample Tree Diagram. Total Sample Yes, injured: 25% No self-injury: 75% n=2000 Male Yes, Injured: 15% No self-injury: 85% n=1000 Female Yes, injured: 35% No self-injury: 65% n=1000

PAGE 78

65 CHAID offers several key be nefits. CHAID does not require data to be normally distributed. In addition, inde pendent variable categories th at do not differ statistically significantly are merged, resu lting in a simplified pictur e of relationships between predictors and the dependent variable (assu ming a Type 2 error has not occurred). Further, CHAID is useful as an exploratory da ta analysis approach in that a large number of predictors can be included in the anal ysis, and a preliminary segmentation model can be developed and verified using confirmatory approaches such as logistic regression or can be replicated using CHAID within a holdout sample (Magidson, 1990). CHAID allows for the inclusion of cluster variable s to determine whether group-level variables (e.g., school) are useful in segmenting the popul ation into subgroups (Magidson, 1990). CHAID includes a Bonferroni al pha adjustment to control in flated Type I error rates associated with the use of multiple, simulta neous statistical tests (Magidson, 1990, 1994). Additional benefits such as th e ability to treat missing values for each predictor variable as a “floating category” are discussed in The Measurement Group (1999-2005). A key benefit to CHAID is the ease in which out put is understood and communicated to lay individuals (Vriens, 2001). Important issues to consider when usi ng CHAID are detailed in Vriens (2001). CHAID is a forward stepwise approach; thus segmentation results depend upon the order in which variables enter the model (The Measurement Group, 19992005; Vriens, 2001). Once a predictor has entered the model, it cannot be removed later in the analysis (Vriens, 2001). Also, segments are developed using statistical criter ia, not practical or theoretical criteria. Thus, segmentation results may not be useful, and not every

PAGE 79

66 important relationship is identified because the focus is on identifyi ng relationships with the greatest odds of being replicated in new samples (The Measurement Group, 19992005). Fortunately, CHAID trees can be revi sed manually to reflect theoretical or applied knowledge (Vriens, 2001). Investigat ors can choose to ‘force’ in independent variables at different stages in the tree ba sed on non-statistical crite ria (Vriens, 2001). Although the ability to consider a large number of independent va riables is a benefit, this increases the risk of including an ‘irrelevant’ variable that may di minish the validity of the segmentation solution (Vriens, 2001). Finally, specifying stopping rules and other CHAID settings can be difficult because ther e are no agreed upon, objective guidelines. For example, the investigator must specify the minimum number of observations in a segment. This decision must be made with close consideration to practical constraints— how small can the group be and still be wort h targeting/considering, and how large can the group be and still be interpretable and re sponsive to targeted efforts (Magidson, 1990; Vriens, 2001)? Finally, because CHAID relies on significance testing, if the sample size used for a CHAID analysis is small or the tr ee is ‘grown’ to too ma ny levels (i.e., smaller and smaller subgroups), it is “susceptible to capitalizing on chance” (Magidson, 1990, p. 108). Segmentation Validity Gathering validity evidence to support se gmentation results is a key aspect of segmentation analysis. Three sources of validity evidence emerged from the literature: the use of theory and applied knowledge in developing segmentations, the use of holdout samples, and predictive valid ity studies (Aldenderfer & Blashfield, 1984; Magidson, 1994). First, ideally theory and applied know ledge are used in in terpreting segmentation

PAGE 80

67 results (i.e., determining the number and na ture of segments/classes). Second, holdout samples (i.e., randomly splitting the original sample into two separate samples) can be used to determine the stability/replicability of segmentations across samples and/or provide evidence of predictive validity (Magidson, 1994). Third, Aldenderfer and Blashfield (1984) suggested determining whether cluster or segment membership predicted “theoretically-related criterion va riables” was the strongest form of validity evidence (p. 224). Summary Early adolescence provides a perfect bac kdrop for the emergence of self-injury. Self-injury offers adolescents a way to re gulate overwhelming affect, gain a sense of identity, separate from parents, solidify re lationships with peer groups, and address other conflicts or goals associated with adolescence (e.g., need fo r self-expression). Evidence suggests self-injury has take n hold among youth in recent cohorts—media attention has increased, schools have taken note, and parent s and other adults are bewildered. Selfinjury is a mental health i ssue, but it is not known whether all youth who self-injure have a diagnosable mental illness, whether self-inj ury is a sign of distress among vulnerable youth in clinical and nonclinical settings, and/or whet her the self-injury is a “new” expression of adolescent risk be havior that is being “labeled as risqu by adults in a particular historical and so ciocultural setting” and b ecoming “normative” (Rew, 2005, p. 167). Current research suggests self-injury is, in many cases, a symptom of distress (i.e., maladaptive coping mechanism) that, during adolescence, is influenced by the environment, especially the phenomenon of social contagion. Se lf-injury may be a

PAGE 81

68 temporary, maladaptive coping mechanis m (‘behavioral dysfunction’) that is automatically and socially reinforced for ma ny youth that ends with the transition to adulthood, with a smaller proportio n switching to chronic, repe titive self-injury (Walsh & Rosen, 1988). In this respect, self-injury is arguably similar to other problem/risk behaviors such as tobacco, al cohol, and other substance us e among adolescents that can, in some cases, be defined as expressions of underlying psychological distress and become addictive over time (Rew, 2005). Because suicide is one of the leading causes of death among adolescents, and self-injury is a str ong predictor of suicide, self-injury among youth should be considered a significant public health issue in need of attention. Whereas recommendations have been to scre en older adolescents for self-injury and implement interventions during mid-to-late ad olescence, efforts to prevent self-injury should be made before the behavior has a ch ance to ‘stick’. This study had three purposes: (a) contribu te to what is known about self-injury among early adolescents in the general middle school population (i.e., non-clinical population), (b) identify behavior s that are comorbid with se lf-injury, and (c) identify segments of youth who self-injure. Overa ll, the study focused on moderate/superficial self-injury as a distinct behavioral phe nomenon with multiple causes and functions. A broad definition of self-harm was used, in cluding multiple behaviors noted among early adolescents. For the purposes of this study, se lf-injury was defined as the performance of a harmful behavior such as cutting, scratc hing, burning, not allowi ng wounds to heal, or pinching, by a person who feels upset as a way to feel better (less upset). This study provided general adolescent population estimate s of the prevalence, 30-day frequency rates of injury among self-injurers, and inform ation about the extent to which adolescents

PAGE 82

69 know a friend who self-injures. Relationshi ps between self-injury and other risk behaviors were described. Segmentation an alyses were used to identify factors associated with self-injury among middle school youth and meani ngful segments of youth who self-injure. Reco mmendations (e.g., Gratz, 2003) to examine sociocultural and gender variations in the prevalence, frequency, and corre lates of self-injury were followed (Gratz, 2003). The interaction between environment (e.g., self-reported exposure to peers who self-injure, exposure to bullying and violence in the school setting, social climate) and individual behavior (i .e., having ever tried self-injury and 30-day frequency rate of self-injury) were co nsidered (see Dishion & Dodge, 2005).

PAGE 83

70 Chapter Three: Method This chapter describes the research approach, accessible population, preliminary prevalence estimates of self-injury, instru mentation, measures of self-injury, data collection, study design, and anal ysis procedures. A discu ssion of the protection of human research subjects and di ssemination of study results is included at the close of this chapter. Research Approach This study involved secondary analysis of data gathered using the middle school Youth Risk Behavior Survey (YRBS) from si xthand eighth-grade students in eight middle schools in a large, south eastern county in Florida. Given the early state of the literature, the dissertation rese arch focused on mining data for patterns and structure. The concept of principled statistical discovery, an iterative analysis approach that involves exploring datasets, identifying potential patterns or structure, and using further statistical tests and/or information to confirm or disc onfirm potential findings, guided the analysis (Mark, 2006). A model of this approach as appl ied to the research is provided in Figure 2. Overall, there were thr ee distinct, yet related, phases to the study. The first phase focused on providing a description of self -injury within a general school-population setting. The second phase involved explor ation and confirmation of relationships between demographic, attitudinal, and beha vioral variables and the three self-injury items. The third phase involved the discove ry and validation of segments or unique

PAGE 84

71 subgroups of youth who self-injur e, self-injure frequently, and know a peer who has selfinjured. The reader should note the multileve l nature of the data was considered in confirmatory analyses (e.g., logistic regres sion) but not in expl oratory analyses (e.g., bivariate). Sampling, methods, key decisions, and other cons iderations are summarized in Figure 2 and are discusse d in the next section.

PAGE 85

72 Figure 2. Model of research approach. Phase 1 Phase 2Phase 3 Purpose : To describe selfinjury in the general adolescent population (i.e., students in regular middle schools whose clinical diagnosis and receipt of services is unknown to the investigator) Methods : Calculate descriptive statistics including measures of the prevalence, central tendency, and variation Sample : Full sample (~1900) including 6th and 8th grade students from one of the eight regular middle schools who responded to the self-injury item Key Variables : Lifetime prevalence of self-injury, 30day frequency of self-injury, peer exposure Considerations: Variation (i.e., subgroup analysis), confounding relationships, scale creation (i.e., to increase reliability) Purpose : To explore and confirm relationships between study variables and the three self-injury items. Methods Step 1 : Calculate bivariate statistics including Chi-square test of independence, Indendent samples t-test, Spearman's rank order correlation Sample : Full sample (~1900) including 6th and 8th grade students from one of the 8 regular middle schools who responded to the lifetime prevalence selfinjury item. Considerations : Given the large sample size alpha = .01 and measures of practical signifcance will be calculated; effect size will be criterion used to select predictors for logistic regression. Methods Step 2 : Conduct multilevel logistic regression with self-injury items as outcome variables. Purpose : To determine if there are meaningful subgroups of youth who selfinjure, self-injure frequently, and know freinds who have self-injured Sample : Original sample randomly split into two samples one for 'learning' the model (learning sample) and one for validating the model (hold out sample) Methods Step 1 : Run CHAID using automatic growth function for each outcome variable within learning sample; force in demographic variables with lifetime prevalence variable as outcome Methods Step 2 : Run CHAID using automatic growth function for each outcome variable within hold out sample; force in demographic variables with lifetime prevalence variable as outcome Methods Step 3: Compare and contrast segmentation results obtained within two samples Considerations : A predetermined effect size rather than statistical significance was the criterion used to decide when to stop splitting (ie., growing the tree). Methods Step 3 : Calculate adjusted odds ratios and 95% confidence intervals for each predictor. Decision : Handling missing data

PAGE 86

73 Accessible Population The accessible population for this study in cluded sixthand ei ghth-grade students (~ 10 to 14 years of age) in eight middle sc hools in a large, s outheastern county in Florida. Special education students were not included in the accessible population as they were not included in the survey admi nistration as per the study county’s district policy. Although data were available from six alternative and private schools in the study county, these were excluded given the sma ll, unrepresentative samples obtained from each site (range = 8 to 21 students). Yout h between 10 and 14 years old were selected because many adolescents of this age are in itiating a variety of risk behaviors (e.g., sexual activity, smoking, drinking an d other drug use) as well as self-injury (Carlson et al., 2005). According to the Florida Departme nt of Education’s Statistical Brief (20052006) in the fall 2005, the study c ounty had 41,884 students in its public pre-kindergarten through 12th grades. Of those students, 9,663 were in middle school, with 2,939 (30.41%) in sixth grade and 3,423 (35.42%) in eigh th grade. The Florida Department of Education reports racial/ethnic data at the county level for public school student membership. The majority of students in the study county’s public schools were White, non-Hispanic (N = 31,097; 74.25%), Hispan ic (N = 4,516; 10.78%), or Black, nonHispanic (N = 3,735; 8.92%), with an overa ll minority population of 10,787 (25.75%). Total enrollment, demographic, and grade level enrollment information specific to each participating middle school are provide d in Tables 2 and 3. A total of 1,748 students were included in the study sample (see Table 2). Examination of free/reduced price lunch information suggests study school s represented a range of socioeconomic (SES) classes, with the lowest percentage of free/reduced price l unch at School 6 and the

PAGE 87

74 highest at School 1. The majo rity of students at most st udy schools were White, which is consistent with study county demographics (see Table 2). However, students at School 1 were more ethnically diverse than were t hose at other study schools (see Table 2). Table 2 Description of the Accessible Popu lation by School (N=1743, December 2005) SCHOOL 1 2 3 4 5 6 7 8 2 Total # of Students 222 176 431 122 254 170 158 210 Gender % Female 51 52 51 51 56 58 48 50 5.31, p = .62, df=7 Race/Ethnicity % White % Black or African American % Hispanic or Latino % Other Race or Ethnicity 34 28 33.5 4 74 10 10 6 76 8 9.5 7 81 6 5 8 78 7 10 6 84 3 6 7 81 2 8 9 84 1 9 7 310.89, p < .0001, df = 35 Grade* % 6th grade 48 24 42 39 58 53 44 57 69.04, p < .0001, df = 7 % Free/Reduced Price Lunch 66.0 35.7 39.8 23.0 33.5 4.1 15.7 27.4 211.34, p < .001, df = 7 Note: Five students included in the sample did not report school attended. *The sample was limited to students in 6th and 8th grades. A total of 5,592 sixthand eighth-grade students were enrolled in study schools in 2005-2006 (Table 3). More eighth graders than sixth graders were enrolled. Overall, sampling resulted in an obtained sample of 31% of enrolled sixth graders and 32% of enrolled eighth graders (Table 3). Random sampling was not used. Whereas samples obtained from most study schools were within the 1/3 of the accessible population range (N=1748), samples obtained from Schools 2 and 7 were lower than those obtained from other study schools. At School 2, surveys were obtained from only 13% of enrolled sixth graders compared to 35% of enrolled eighth graders. At School 7, surveys were obtained from only 19% of enrolled sixth graders and 20% of enrolled eighth graders.

PAGE 88

75 Table 3 Comparison of Sample Obtained and Enrollment by School (December 2005) School 6th Grade Sample (2005) 6th Grade Enrollment (2005/2006) % of 6th Grade Population Obtained 8th Grade Sample (2005) 8th Grade Enrollment (2005/2006) % of 8th Grade Population Obtained 1 107 285 37.5 115 359 32 2 42 324 13 134 384 35 3 182 569 32 249 661 38 4 48 146 33 74 250 30 5 148 349 42 106 355 30 6 90 256 35 80 228 35 7 70 365 19 88 450 20 8 120 335 36 90 276 33 Total 807 2629 31 936 2963 32 Unlike clinical samples where the diagnos is and receipt of services are known, individuals included in the accessible populatio n may or may not have had a clinical diagnosis associated in the clin ical literature with self-injur y (i.e., depression). Further, some students may have been receiving psyc hological services at the time of survey administration either from a private clinicia n or from a school psychologist. According to the school board of the study county, a pproximately 2% to 3% of middle schools students received psychological servi ces in the schools during the 2005–2006 school year. The proportion of students receiving psyc hological services from private clinicians was unknown. Instrumentation The middle school version of the YRBS is used by the county school board to monitor risk health and risk behavior s among middle school youth and for prevention programming and evaluation purposes. The YRBS questionnaire was developed by the Centers for Disease Control and Prevention (CDC), with input from the Methods and Evaluation Unit of the University of South Fl orida Prevention Resear ch Center (FPRC).

PAGE 89

76 The YRBS is a school-based classroom survey of risk behaviors self-reported by middle school youth (see Appendix A). Usually condu cted at the high school level (Grades 912), the 104-multiple-choice questionnaire was modified to include questions relevant to middle school students. The mi ddle school survey is used to monitor six categories of priority health and risk behaviors am ong youth and young adults: (a) unintentional and intentional injuries, (b) tobacco use, (c) alcohol and other dr ug use, (d) sexual behaviors that contribute to unintended pregnancies a nd sexually transmitted diseases, (e) unhealthy dietary behaviors, and (f) phys ical inactivity (Kann et al ., 1998). The 2005 middle school YRBS also included questions about demographics, delinquent behaviors, communication/relationship with parents/guardians, exposure to prevention interventions, and self-reported grad es (see Table 4). Table 4 Middle School Youth Risk Behavior Survey Item Categories Item Category Number of Items Demographics 7 Personal safety and violence-related behaviors 8 Bullying 12 Cyberbullying 4 Suicide 3 Self-harm 3 Tobacco use 10 Alcohol use 6 Marijuana use 4 Other drug use 4 Body weight 7 Physical activity 9 AIDS education 1 Sexual intercourse 4 General health behavior 2 Delinquent behavior 4 Exposure to Believe Campaign 4 Parental communication about drugs and alcohol 2 Feelings about future, substance use, and family 4 Attitudes toward school 3 Self-reported academic performance 1 Truthfulness in answering survey questions 2

PAGE 90

77 Measures of Self-Injury In the study county, Safe School Liaisons were responsible for monitoring risk and protective factors among youth and assi sting schools and community agencies in addressing reoccurring and remerging issues. In addition to increases in suicidal ideation, Safe School Liaisons noted increases in the numbers of students practicing self-harm or requiring services for the behavior. To in crease their ability to develop or locate interventions to address self -harm among youth, Safe Schools Liaisons needed to be able to identify youth at risk for self-injury and f actors to consider when addressing self-injury (e.g., co-morbid behaviors, gender or grade differences, school level variation). In response to those identified need s, the investigator assisted the Safe School Liaisons in developing three items specific to self-harm. These items were designed to assess the prevalence and frequency of se lf-injury and level of peer exposure. Item development was informed by a review of th e self-injury literature. Safe School Liaisons, who worked with middle school youth and were trained in guidance and prevention, helped define se lf-injury and played a key role in item generation. Self-injury was de fined for youth to help ensure each participant responded using the same frame of reference. The follo wing lead in was placed directly before the series of self-injury items: The next 3 questions ask about self -harm (cutting, scratching, burning, not allowing wounds to heal, pinching). Sometimes people who feel upset hurt themselves on purpose as a way to feel better (less upset).

PAGE 91

78 Three items were developed to measure three aspects of self-injury: lifetime prevalence, past 30-day prevalence, and awar eness of peer self-injury beha vior. Each of these items is reprinted below: 1. Have you ever hurt yourself on purpose (c utting, scratching, burning, not allowing wounds to heal, pinching)? a. Yes b. No 2. During the past month, how often ha ve you hurt yourself on purpose (cutting, scratching, burning, not allowing wounds to heal, pinching)? a. Never b. 1 time c. 2 or 3 different times d. 4 or 5 different times e. 6 or more different times 3. Have any of your friends hurt themse lves on purpose (cutting, scratching, burning, not allowing wounds to heal, pinching)? a. Yes b. No Data Collection Safe School Liaisons with the assistan ce of middle school teachers administered the YRBS to sixthand eight h-grade students at eight middl e schools and six alternative and private schools in the county in December 2005. Approximately 2,350 surveys were distributed across schools. Each school c onducted an in-service tr aining for teachers describing the data collection protocol. A letter was sent home to students allowing parents to opt out their child from the surv ey administration. Students who were opted out (~10% of eligible students) were not allowed to take the survey on the day of administration. An effort was made to survey one-half of all stude nts enrolled in sixth

PAGE 92

79 and eighth grades in the eight schools. Speci al education students were excluded from participation as per district policy. Teach ers, in their respective subjects, then administered the self-reported questionnaire to students during a regul ar class period (~ 45 minutes). Survey procedures were designe d to protect the student s’ privacy and allow for anonymous, voluntary participation. Sta ndard electronic answ er sheets (“bubble sheets”) were used by students to record thei r responses. Data were then read by an optical scanner. Visual inspection reveal ed that out of approximately 2,350 surveys distributed, a total of 2,003 valid surveys were completed, resulting in an initial response rate of 85.23%. A total of 1,907 students (~81% of the original sa mple) self-reported attendance at one of the eight middle schools. Protection of Human Subjects Parents were informed of the possibility of their child being administered the YRBS and were provided with a means for opti ng their child out of survey participation through distribution of a letter to parent s at the beginning of the 2005–2006 school year. Students who were opted out of participating were not allowed to complete the YRBS on the day of survey administration. The inves tigator obtained permissi on from the director of pupil support services of the school board to utilize the data from the 2005 YRBS administration for dissertation purposes. The study protocol was re viewed and approved by the University of South Florida Institutional Review Board, Social and Behavioral Sciences Division.

PAGE 93

80 Analysis Procedures Step 1: Data Entry and Cleaning Youth who agreed to participate record ed their responses to each item on a scantron sheet. Scantrons that were wri nkled or smudged were numbered with a unique identifier and hand entered in a Microsoft Excel database to ensure data quality. Once all surveys were entered into a Microsoft Excel database, the inves tigator calculated frequencies for each variable to identify response values outside of the established response categories. Values outside of th e expected range were double checked against the original scantrons using the unique identi fier (i.e., ID variable ). Corrections were made where possible. When a correction wa s not possible, the response was recoded as missing. SAS v. 9.1.3 was used to calculate al l statistics, with the exception of CHAID analysis, which was conducted using SPSS Answertree v. 3.1 software, and MPLUS (Muthn & Muthn, 1998-2006) and HL M 6 (Raudenbush, Bryk, Cheong, & Conadon, 2004), which was used to conduct multilevel modeling. Step 2: Creation of Study Datasets Multiple datasets, based on the original, were used in the research reported herein. The following actions were taken to limit the overall dataset. Only students who self-reported atte nding one of the eight middle schools were retained. Responses were valida ted using the second school item that listed private and alternative schools: st udents who self-report ed attendance at both a public middle school and a private or alternative school (i.e., an invalid response pattern) were excluded.

PAGE 94

81 Fifty-eight participants who responded something other than sixth or eighth grade (e.g., “other”) to the grade level item were excluded because the YRBS was primarily administered to sixth and eighth graders. Forty-seven participants who did not respond to the having ever tried selfinjury item (i.e., those with a missing re sponse), the main dependent variable, were excluded. Missingness on this item was statistically significantly, but weakly associated, with gender: males we re more likely to not respond to this item than were females (2.4% vs. 71%; 2(N = 1959, 1) = 8.78, p < .01, Cramer’s V = -0.07). Missingness also was statistically significantly, but weakly associated, with race or ethnic ity: White students were more likely to not respond than Black students (4.6% vs.1.5%; 2(N = 1580, 1) = 8.51, p < .01, Cramer’s V = 0.07) and students of other ethnicities (6.1% vs.1.5%; 2(N = 1545, 1) = 15.80, p < .0001, Cramer’s V = 0.10). Twenty six participants who reported answering trut hfully less than one-half of the time and none of the time were excluded. However, participants who did not respond to this item were not excluded given the number of students who were unable to finish the survey and, therefore, were unab le to respond to the ‘truth item’ (i.e., survey item #103). These actions resulted in a final sample size of 1,748, representing approximately 92% of participants who self-reported attendance at one of the eight middle school (N = 1,907) and 74% of the 2,350 surveys or iginally distributed. The nature of missing data also was c onsidered. Some youth may have skipped items they did not want to answer, especially those specific to risk behaviors, and some

PAGE 95

82 youth may not have been able to complete all 104 items due to time constraints associated with survey administration. The nature of missing data was explored using descriptive and bivariate statistics (e.g., correlations). Un ivariate and bivariate statistics were used to describe differences, if any, between those with no missing data and those with some missing data with in the reduced sample of 1,748. Approximately 70% of students had zero missing responses. A nother 14% had only one or two missing responses. The average number of missing re sponses was 2.5, with a range of 0 to 46. Missingness was negatively associated with ag e: as age increased, the number of missing responses decreased ( r = -.09, p < .01). On average, males had higher numbers of missing responses than did females (2.94 vs. 2.01; t (1738) = 3.16, p = .0016; Cohen’s d = 0.15). On average, sixth graders had highe r numbers of missing responses than did eighth graders (3.28 vs. 1.78; t (1746) = 5.12, p < .0001; Cohen’s d = 0.24). Missingness was not significantly statistical ly associated with the main outcome variable of this study, having ever tried self-injury, t (1746) = -0.84, p= .40. Given the size of the available sample and the fact that most participan ts had zero to two missing responses (84%), listwise deletion was used to eliminate cases with missing data on each variable used in each analysis conducted. Associations between gender and age and missingness were considered when interpreti ng key study findings. Step 3: Variable Selection and Modification Because this study sought to provide a description of self-injury during early adolescence, many of the variables from the 2005 YRBS were used (see Tables 5 and 6). In addition to demographic (e .g., ethnicity) and descriptive items (e.g., perceived health

PAGE 96

83 status), indicators of problem behavior th eory, social contagion, precipitants of selfinjury, and developmental theory were identi fied and are summarized in Tables 7 and 8. Table 5 Interval-Level Variable Descriptive Statistics Variable N RangeMeanMedianSD Skewness Kurtosis Age 1746 10 – 16 12.52 13.00 1.18 0.06 -0.88 Age at first alcohol use 640 8 – 14 10.56 11.00 1.98 0.03 -1.37 Age at first cigarette use 291 8 – 14 10.70 11.00 1.88 -0.17 -1.26 Age at first marijuana use 194 8 – 14 11.56 12.00 1.83 -0.65 -0.66 Age at first sex 266 8 – 14 11.41 12.00 1.96 -0.58 -0.91 Grades 1519 1 – 9 7.41 8.00 1.58 -1.63 3.02 Health 1734 1 – 5 3.94 4.00 0.91 -0.55 -0.20 Number of sexual partners 253 1 – 3 1.87 2.00 0.86 0.25 -1.62 Time on computer or video games 1610 0 – 7 2.26 2.00 1.82 0.91 0.30 TV hours per day 1659 0 – 6 3.03 3.00 1.75 0.16 -0.84 Note: All variables were coded so that a higher score represented a higher amount of the characteristic, behavior, or attitude being measured. Table 6 Prevalence Information for Categorical Study Variables Individual Variables Yes (%) During your lifetime, have you ever been cyberbullied? 22.6 During the past 12 months, did your boyfriend or girlfriend ever hit, slap, or physically hurt you on purpose? 7.7 Have you ever seriously thought about killing yourself? 21.7 Have you ever made a plan about how you would kill yourself? 13.5 Have you ever tried to kill yourself? 7.6 Have you ever tried cigarette smoking, even one or two puffs? 25.1 During the past 30 days, have you smoked cigarettes, even one or two puffs? 10.7 Have you ever had a drink of alcohol, other than a few sips? 36.3 In the past 30 days, have you had any alcohol to drink, other than a few sips? 17.3 In the last year, have you had five or more drinks of alcohol in one day? 12.6 Have you ever used marijuana? 14.0 Have you ever sniffed glue, or breathed the contents of spray cans, or inhaled any paints or sprays to get high? 15.0 Have you ever used prescription drugs or over the counter medicine (cough/cold medicine) to get high? 5.4 Have you ever had sexual intercourse? 17.6 Have any of your friends hurt themselves on purpose (cutting, scratching, burning, not allowing wounds to heal, pinching)? 46.8 Have you ever hurt yourself on purpose (cutting, scratching, burning, not allowing wounds to heal, pinching)? 28.4

PAGE 97

84 Table 7 Individual Variables Selected for Use and Associated Theoretical or Conceptual Framework Theory or Concept Individual Variables During your lifetime, have you ever been cyberbullied? During the past 30 days, how many times were you the victim of cyberbullying? Precipitants of Self-injury During the past 12 months, did your boyfriend or girlfriend ever hit, slap, or physically hurt you on purpose? Have you ever sniffed glue, or breathed the contents of spray cans, or inhaled any paints or sprays to get high? Problem Behavior Theory Have you ever had sexual intercourse? Have any of your friends hurt themselves on purpose (cutting, scratching, burning, not allowing wounds to heal, pinching)? On an average school day, how many hours do you watch TV? Social Contagion On an average school day, how many hours do you spend playing video games or using a computer for fun? (Include activ ities such as Nintendo, Game Boy, Play Station, and computer games.) Table 8 Scales Developed for Use and Associated Theoretical or Conceptual Framework Theory/Concept Scale Number of Items Cronbach’s Range of Item to Total Correlations Attitude Toward School 3 .55 .33 .43 Belief in Possibilities 3 .76 .45 .67 Developmental Theory Parent Communication 2 .83 .71 .71 Precipitants Bully – Victim 5 .74 .39 .59 Abnormal Eating 3 .59 .39 .51 Deviant Behavior Scale 2 .51 .34 .34 Suicide Scale 3 .75 .58 .63 Problem Behavior Theory Substance Use Scale 10 .88 .50 .70 Cronbach’s alpha (Cronbach’s ), a measure of internal consistency reliability, was calculated for item sets that were de signed to measure the same behavior or underlying construct (i.e., to be used as a scale), including attitudes toward school, belief in possibilities, parent communication, and bu llying (see Tables 8 and 9). Many of the

PAGE 98

85 scales had a small number of items and, theref ore, reliabilities were generally lower than the minimal levels commonly accepted for research (i.e., .70). Exploratory factor analysis (EFA) with tetrachoric (i.e., dichotomous items) and polychoric (i.e., polytomous items) correlations was conducted us ing Mplus v. 3.0 to ai d in the reduction of the number of variables used in the multiv ariate component of the study (see Appendix B). Variables that were not necessarily desi gned to create a scale were included, such as substance use (e.g., tobacco, alcoho l, marijuana, inhalants, pr escription drugs), theft, and skipping. Promax rotations were used because it was assumed factors would be correlated. Results from the promax solution revealed substantial correlations between factors, so Promax rotated pattern coeffi cients were interpreted (see Appendix B). Pattern coefficients combined with theory we re used to create scales (see Table 8). Cronbach’s alpha was calculated for each sc ale. Where appropriate, variables were modified (e.g., dichotomized) for use in the segmentation analysis (i.e., a set of dummy variables were created for each nominal vari able). Tables 8 through 10 present scale definitions and psychometric information. Al l variables were coded so that a higher score represented a higher amount of the ch aracteristic, behavior or attitude being measured. Table 9 Scale Definitions and Internal Consistency Reliability 1. Have you ever gone without eating for 24 hours or more (also called fasting) to lose weight or to keep from gaining weight? 2. Have you ever taken any diet pills, powde rs, or liquids without a doctor’s advise [sic] to lose weight or to keep from gaining weights? (Do not include meal replacement products such as Slim Fast.) Abnormal Eatinga (Cronbach’s = .59) 3. Have you ever vomited or taken laxatives to lose weight or to keep from gaining weight. Attitude Toward 1. People at my school notice when I am good at something.

PAGE 99

862. I participate in activities (clubs, sports, WEB, etc.) at this school. Schoolb (Cronbach’s = .55) 3. There is at least one teacher or adult at this school I can talk with if I have a problem. 1. I believe I can choose to not smoke cigarettes or drink alcohol, even if I’m going through tough times. 2. I believe my future holds many possibilities. Belief in Possibilitiesb (Cronbach’s = .76) 3. I believe I have better things to do than smoke cigarettes or drink alcohol. 1. During the past 30 days, how many times did another student tease or call you names? 2. During the past 30 days, how many times did another student threaten to hit or hurt you? 3. During the past 30 days, how many times did another student spread rumors about you? 4. During the past 30 days, how many times did other students not let you join in what they were doing? Bully – Victimc (Cronbach’s = .74) 5. During the past 30 days, how many times did another student push, shove, slap, hit, or kick you on purpose? 1. Have you ever tried cigarette smoking, even one or two puffs? 2. During the past 30 days, have you smoked cigarettes, even one or two puffs? 3. During the past 30 days, on how many days did you smoke cigarettes? 4. Have you ever had a drink of alcohol, other than a few sips? 5. In the past 30 days, have you had any alcohol to drink, other than a few sips? 6. In the last year, have you had five or more drinks of alcohol in one day? 7. During the past 30 days, how many times have you had 5 or more drinks in one day? 8. Have you ever used marijuana? 9. During the past 30 days, how often have you used marijuana? Substance Used (Cronbach’s = .88) 10. Have you ever used prescription drugs or over the counter medicine (cough/cold medicine) to get high? 1. My parents have talked to me about their feelings toward me smoking cigarettes. Parent Communicatione (Cronbach’s = .83) 2. My parents have talked to me about their feelings toward me drinking alcohol. 1. Since school started this year how many times have you skipped school? Deviant Behaviorsf (Cronbach’s = .51) 2. During the past 12 months, how often have you shoplifted (stolen something from a store)? 1. Have you ever seriously thought about killing yourself? 2. Have you ever made a plan about how you would kill yourself? Suicidea (Cronbach’s = .75) 3. Have you ever tried to kill yourself? a Response scale for Items ranges from 0 (No ) to 1 (Yes). b Response scale for Items ranges from 1 (Strongly Disagree) to 5 (Strongly Agree). c Response scale for Items ranges from 0 (0 times) to 4 (10 or more times). d Response scale for Items 1 – 2, 4 – 6, 8, and 10 goes from 0 (No ) to 1 (Yes). Response scale for Items 3, 7, and 9 ranges from 0 days to 30 days. e Response scale for Items ranges from 0 (No ) to 2 (Yes). f Response scale for Item 1 ranges from 0 (Never) to 4 (More than 3 times). Response scale for Item 2 ranges from 0 (0 times) to 4 (6 or more times). Original variables were used to create most scales with the exception of the Substance Use and Deviant Behaviors scales (see Table 10). Becau se response scales

PAGE 100

87 differed between items used to create each scale, item responses needed to be standardized before scales were created. Four out of eight scales demonstrated nonnormal distributions (i.e., high skewness and kurtosis values) including: Abnormal Eating, Belief in Possibilities, Substance Us e, and Deviant Behaviors. The Abnormal Eating Scale was transformed to normalize the distribution using the natural log function in SAS. The transformation reduced the skewness and kurtosis from 2.60 and 6.65 to 1.88 and 2.05, respectively. The Belief Scale was transformed to normalize the distribution using the cos(in e) function in SAS. The transformation reduced the skewness and kurtosis from -2.13 and 5.39 to -0.91 and -0.56, respectively. The Substance Use Scale was transformed to norma lize the distribution using the natural log function in SAS. The transformation redu ced the skewness and kurtosis from 2.70 and 8.31 to 0.69 and kurtosis -0.84, respectivel y. The Deviant Behavior Scale was transformed to normalize the distribution usi ng the natural log function in SAS. The transformation reduced the skewness and kurtosis from 2.46 and 6.58 to 0.90 and -0.83, respectively. Statistical tes ting was conducted using the orig inal and transformed scales and results were compared to examine the sensitivity of the results to nonnormality. Unless otherwise noted, results are reported based on tests conducted with original scales. Table 10 Scale Descriptive Statistics Scale N Range M Median SD Skewness Kurtosis Abnormal Eating (Original) 1646 0-3 0.26 0.00 0.62 2.60 6.65 Abnormal Eating (Transformed)a 1646 -0.69-1.25 -0.45 -0.69 0.52 1.88 2.05 Attitudes Toward School 1535 1-5 3.74 4.00 0.94 -0.69 0.12 Belief in Possibilities (Original) 1538 1-5 4.53 4.67 0.70 -2.13 5.39 Belief in Possibilities (Transformed)b 1538 -0.99-0.54 -0.06 0.28 0.43 -0.91 -0.56 Bully – Victim 1746 0-4 0.73 0.40 0.78 1.51 2.17 Substance Use (Original)c 1708 -0.43–3.86 0.00 -0.39 0.69 2.70 8.31

PAGE 101

88Substance Use (Transformed)a 1708 -2.63–1.47 -1.32 -2.21 1.05 0.76 -0.68 Parent Communication 1542 0-2 1.40 2.00 0.81 -0.85 -0.96 Deviant Behavior (Original)c 1595 -0.44–3.74 -0.00 -0.44 0.82 2.46 6.58 Deviant Behavior (Transformed)a 1595 -2.81–0.44 -1.81 -2.81 1.41 0.87 -0.93 Suicide 1732 0–3 0.43 0.00 0.85 1.96 2.75 Note: All variables were coded so that a higher score repr esented a higher amount of the characteristic, behavior, or attitude being measured. aThis scale was transformed to normalize the distribution usi ng the natural log function in SAS. Statistical testing was conducted using the original and transformed scales. bThe belief scale was transformed to normalize the distributio n using the cos(ine) function in SAS. Statistical testing was conducted using the original and transformed scales. cVariables were standardized ( M = 0, SD = 1), and a composite variable was created by taking the average of the standardized variables. Step 4: Description of Self-injury in General Middle School Population Within the full sample, univariate statistics including frequencies, measures of central tendency, and measures of variati on were calculated for each study variable, where appropriate. The normality of c ontinuous variables was assessed and the implications of nonnormality were considered when conducting bivariate analyses. The following research questions were addressed through the calculation of frequencies and proportions: What is the prevalence of self-i njury among middle school youth? What is the frequency of self-injury am ong middle school youth who self-injure? What proportion of middle school yout h know a friend who self-injures? Confidence intervals were pr ovided. Because of potential differences between groups, univariate statistics for these three items also were calculated by gender, racial or ethnic classification, age, grade, and school, which partially answered the following questions: Are there gender, racial or ethnic, age, grade, and sc hool differences across rates of self-injury, frequency of self-injury, a nd knowledge of friends who self-injure? Interrelationships among these variables (e.g., gender and ethnicity) were examined to address potential confoundi ng relationships.

PAGE 102

89 Step 5: Exploration of Relationships Between Self-Injury and Other Behaviors Bivariate relationships between possible co rrelates and self-injur y were calculated using appropriate statistical techniques such as Pearson correlations, Spearman correlations, independent samples t-tests, and chi-square tests of independence (see Appendices B and C). The following questions were answered, in part, using bivariate analyses: What demographic, attitudina l, and behavioral variables are related to self-injury (see Table 2)? Are there gender, racial or ethnic, age, grade, and school differences across rates of self-injury, frequency of self-injury, a nd knowledge of friends who self-injure? Where does self-injury fit in with other risk behaviors such as alcohol use, tobacco use, suicide, and deviance? Measures of statistical and practical significance were calculated. The overall alpha level, given the large sample size, was set at .01. Measures of practical significance (e.g., Cramer’s V for chi-square tests of independence) were calculated where appropriate (e.g., to de scribe differences in m eans or proportions among youth who have tried self-injury and those who have not and those who se lf-injure frequently vs. infrequently). Cohen’s “rule-of-thumb” for interpreting effect sizes was used (see Table 11).

PAGE 103

90 Table 11 Cohen’s Effect Size Interpretation Rules-of-thumb Cohen’s d Correlation Coefficient Odds Ratio Cramer’s V Small .20 .10 1.50 df = 1; 10 < V < .30 df = 2; 07 < V < .21 df = 3;.06 < V < .17 Medium .50 .25 2.50 df = 1; 30 < V < .50 df = 2; 21 < V < .35 df = 3; 17 < V < .29 Large .80 .40 4.30 df = 1; V > .50 df = 2; V > .35 df = 3; V > .29 Note: The guideline for chi-square tests of independence with 3 degrees of freedom was used for tests with greater than three degrees of freedom. To confirm relationships id entified at the bivariate level, multilevel logistic regression analysis was conducte d using the predictor variable s identified in Tables 7 and 8 and demographic variables (e.g., gender, race, grade). Bivariate relationships between predictors were considered to rule out possible multicoll inearity (see Appendix C). Multilevel modeling was used because student s (Level-1) were nested within schools (Level-2). Only Level-1 predictors were used. Models were run with three outcome variables: having ever self -injured (dichotomous), the frequency of self-injury (polytomous), and peer self-injury (dichotom ous). Multinomial logistic regression was conducted with a modified version of the fr equency of self-injury outcome variable. Frequency of self-injury (past 30 days) was m odified to included three categories: (0) never self-injured, (1) self-inj ured once, and (2) self-injur ed two or more times. Two models were run, allowing for the following comparisons to be made: once versus never, more than once versus never, and once versus more than once. The models were estimated using penalized quasi-likelihood estimation (PQL) and were conducted using HLM version 6. The Bernoulli distribution at Level-1 was used for both dichotomous

PAGE 104

91 outcome variables, and the multinomial distribution was used for the polytomous outcome variable. Adjusted odds ratios we re calculated, along with 95% confidence intervals for each (see Wright, 1998). The assumptions of logistic regression were considered, such as model sp ecificity, mutually exclusive and collectively exhaustive categories, and a minimum of 50 cases per pr edictor variable (Wright, 1998). Logistic regression results were summari zed in tables specific to each outcome variable. Step 6: Identification of Meaningful Segments of Youth Who Self-injure CHAID analyses using SPSS Answertr ee v. 3.1 audience segmentation software were used to answer the fo llowing research question: Are there meaningful segments of yout h who self-injure? If so, what characteristics are useful in defining each segment? More specifically, CHAID was used to divi de the sample into subgroups (segments) based on interactions between pr edictor variables identified in Step 4, which predict each criterion variable. Having ever tried self-i njury was the first [dichotomous] dependent variable analyzed. Predictor variables were identi fied as nominal, ordinal, or continuous (Magidson, 1994). Given the sample size, settin gs for parent and child node size were as follows: n = 20 for parent node and n = 10 for child nodes (The Measurement Group, 1999-2005). The overall alpha level was set at .01; however, Bonferroni adjustments were used to control for alpha inflation resu lting from simultaneous statistical testing. The size of subgroups and the av ailability of stat istically significant predictors were considered when assessing tree depth. An eff ect size in addition to statistical significance was used as the criterion for determining when to stop splitting (i.e., growing the tree). Cramer’s V (i.e., effect size appropriate fo r chi-squared tests of independence) was

PAGE 105

92 calculated for each node. [Cramer’s V is e quivalent to the Phi coefficient when calculated for two-by-two tables.] Nodes th at did not meet the minimum value for a small effect size were not considered pract ically meaningful and, thus, were excluded from the segmentation tree. Segmentation an alyses were conducted using the automatic growth function. Segmentation analyses were conducted using origin al and transformed predictor variables (e.g., belief). Segmentation tress with original pr edictor variables are presented and differences between trees (i.e., original vs. transformed) are noted. The resulting tree diagram and gains table were reviewed to determine predictor variables useful in segmenting middle school-aged yout h according to self-injury behavior and segments of youth most likely to self-injure. Classification accuracy was determined by examining a crosstabulation of the actual ca tegories of the cases and their predicted categories using the model (i.e., the segmenta tion tree). The risk estimate, or the proportion of misclassified cases, is reported, as is the classification accuracy, or the proportion of correctly classified cases. A description of each segment was developed, including the size and characteristics. There is a lack of agreement in the lite rature as to the best approach for model building/testing when using C HAID. Given the fact that the inclusion of extraneous variables can change segmentation results a nd the number of variab les included in this analysis, two approaches were used and the re sults of each were compared, including: use of all predictor vari ables (i.e., exploratory approach) and use of predictor variables selected using logistic regres sion results (i.e., confirmatory approach). Predictors that were found to be statistically significant usi ng logistic regression at the alpha = .10 level were included in the confirma tory approach (see Forthofer & Bryant, 2000). Comparison

PAGE 106

93 of results suggested interpre tation of the inclusive model (i.e., that which included all predictors) resulted in a more well-devel oped tree. Thus, only trees grown using all predictor variables are presented. A final se gmentation was develope d using results that were statistically and practically significant across methods approaches, and theory. The frequency of self-injury during the past month and knowing a friend who self-injures also were used as dependent variables in segmentation analyses. Using having ever self-injured as a dependent variable, the freque ncy of self-injury during the past month was ‘forced in’ as a predictor variable. Descriptive information and the results of the segmentation (e .g., where the categorie s split) were used to transform the original variable into a new dependent vari able based on where the frequency variable split. Results suggested differences between those who had never tried self-injury, those who had self-injured once, and those w ho had self-injured more than once ( p < .01). Thus, a new variable was created with three response options. The new frequency dependent variable/s was used as a criteri on variable in a second segmentation analysis that sought to identify variables that statis tically significantly inte racted to distinguish between each group. Test-sample cross-validation was used to validate the CHAID analysis results for each criterion variable. The dataset was randomly split in to two samples: a training sample used for initial CHAID analysis, and a test (hold-out) sample for cross-validation analysis. The predictive accuracy of each classification tree developed within the learning sample was tested within the holdout sample (i.e., misclassification rates were compared).

PAGE 107

94 Segmentations were judged using the following criteria (see Malhotra, 1989 for discussion): mutual exclusivity (i.e., segments are distinct) and exhaustivity (i.e., each target member is included in a segment), m easurability (i.e., size and other characteristics of segments can be measured), substantiality (i.e., segments are of sufficient size to warrant pursuit), and actiona bility (i.e., segments can be reached and served). Step 7: Present Findings Results are summarized in narrative form at, and tables and graphs are used to summarize and illustrate key findings. Resu lts are presented according to each of the three guiding research objectives. Segmenta tion trees are included. Finally, an overall summary of answers to each re search question is provided. Issues to Consider Self-injury is affected by numerous, indi vidual and contextual level factors. For example, the literature suggests variation in self-injury rates acro ss gender, grades, and schools. Variability across eight middle schoo ls was considered. Descriptive statistics were calculated for each study variable by gende r, grade, and school. Due to the small number of schools, examination of between -school variability was restricted to descriptive and bivariate stat istics such as chi-square tests of independence. This study involved a large number of va riables, which can increase the odds of including irrelevant variables that may dist ort segmentation results (Vriens, 2001). To reduce the number of variables used, summary scales consisting of multiple items were created and internal consistency reliability estimates (i.e., Cronbach’s alpha) were calculated for each. Predictor variables us ed in CHAID can be variables of mixed measurement levels, including cat egorical or continuous vari ables (Vriens, 2001). This

PAGE 108

95 poses issues, however, for categorical variable s with more than two levels. Categorical levels with more than two levels were transf ormed into dummy variables (Vriens, 2001).

PAGE 109

96 Chapter Four: Results Introduction This chapter begins with a review of the research purpose and questions. The next section, Description of Self-injury in General Middle School Population describes the prevalence and frequency of self-injury among middle school st udents in this study and the phenomenon of peer self-injury. The re mainder of the chapter is organized into three major sections repeated for each of th e three dependent variab les: having ever tried self-injury, the frequency of self-injury in the past 30 days, and knowing a friend who had tried self-injury (i.e., peer se lf-injury). The major sections are Relationships between the Outcome Variable and Other Variables Multilevel Logistic Regression Analyses and CHAID Analyses The chapter concludes with a summa ry of answers to the three broad questions that guided this dissertation research. Research Purpose and Questions The purpose of this study was to provid e a description of self-injury within a general adolescent population. This research was designed to identify subgroups of selfinjurers, identify behaviors associated with self-injury, explore relationships between environmental factors (e.g., peer, media) and self-injury, and suggest risk and protective factors associated with self-injury. Thr ee broad questions guided this dissertation research: (a) What is the stat us of self-injury within a public middle school setting in terms of prevalence, frequency, exposure, and correlates, including demographic (e.g., gender), attitudinal (e.g., attitudes toward sc hool), and behavioral variables (e.g., having ever been bullied)? (b) How does self-injury relate to other risk behaviors, such as

PAGE 110

97 tobacco use, alcohol use, suicide, and de viance among youth? and (c ) What factors are useful in identifying meaningful subgroups (s egments) of youth who are more likely to self-injure? Description of Self-injury in General Middle School Population Prevalence of Self-injury Self-injury was defined on the YRBS as a wa y to “feel better or less upset.” After reading the definition, students were asked wh ether they had “ever hurt themselves on purpose (i.e., cutting, scratching, burning, not allowing wounds to heal, pinching).” The prevalence of self-injury among 1,734 middl e school youth in this study was 28.4% ( n = 492), with a margin of error of 2.1% at 95% confidence. Frequency of Self-injury During the past month, most youth (74.6% 95% CI = 73.6-75.6), in general, had never harmed themselves on purpose. Approximately 15% had harmed themselves one time. Smaller proportions of youth had harm ed themselves more frequently, including two or three different times (5%), four or five different times (2%), and six or more different times (3%). There was a significan t and large relationship between having ever tried self-injury and past m onth frequency of self-injury, 2(N = 1746, 4) = 755.74, p < .0001, Cramer’s V = .66. Among youth who self -reported having ever tried self-injury (N = 495), 35% had harmed themselves one ti me during the past month, 18% had harmed themselves two or three different times, 5.5% had harmed themselves four or five different times, and 11% had harmed themse lves six or more different times.

PAGE 111

98 Peer Self-injury Almost one-half (46.8%, 95% CI = 4 5.6% 48.0%) of youth surveyed reported knowing of a friend who had harmed himself/hers elf on purpose to feel better. There was a significant, yet small relationship between knowing a friend who had tried self-injury and having ever tried self-injury. Whereas 39% of those who had not tried self-injury reported knowing of a friend who had tried sel f-injury, 66% of those who had tried selfinjury reported knowing of a frie nd who had tried self-injury, 2(N = 1,732, 1) = 105.01, p < .0001, Cramer’s V = .25. Bivariate Relationships Between Studen t Demographic Variables and Self-injury Outcomes Possible gender, racial or et hnic, age, grade, and school differences across rates of self-injury, frequency of se lf-injury, and knowledge of fr iends who self-injure were examined. Although the relationship between having ever tried self-injury and gender was statistically significant ( p < .01), the effect size was negligible (i.e., .07). Approximately 32% of females and 25% of males had ever tried self-injury, 2(N = 1,740, 1) = 9.75, p < .01, Cramer’s V = .07. There was no statistically significant or meaningful association between having ever tried self-injury a nd race or ethnicity, 2(N = 1,726, 5) = 7.08, p = .21, Cramer’s V = .06; grade level, 2(N = 1,748, 1) = .10, p = .75, Cramer’s V = .01; age, t (1744) = -.01, p = .99; or school attended, 2(N = 1,743, 7) = 12.53, p = .08, Cramer’s V = .08. The frequency of self-injury ranged from a low of 22.2% at School 7 to a high of 33.3% at School 1. Interrelationships among gender, race or ethnicity, age, grade, and school were examined to address potential confounding re lationships. Results suggested race or

PAGE 112

99 ethnicity was statistically significantly associated with school attended, reflecting variations in ethnic di versity across schools, 2(N = 1,721, 35) = 310.89, p < .0001, Cramer’s V = .19. The strength of this relationship, however, did not suggest confounding. Age and grade also were st atistically significantly associated, 2(N = 1,746, 6) = 1635.26, p < .0001, Cramer’s V = .97. The strength of this relationship, on the other hand, does suggest confounding. Thus, only grad e was used in logistic regression and CHAID analyses. Finally, grade level and sc hool attended were sign ificantly associated (see Table 1), with the proportion of surveys re turned by sixth or eighth graders varying across schools, 2(N = 1,743, 7) = 69.04, p < .0001, Cramer’s V = .20. The strength of this relationship, however, did not suggest confounding. Relationships Between Self-injury and Other Variables Results suggested small effects of having ever self-injured on student health and academic performance. On average, student s who had ever tried self-injury reported poorer health than those who had not tried self-injury ( M = 3.74 vs. 4.02; t (843) = 5.72, p < .0001; Cohen’s d = -0.31). On average, stud ent who had ever trie d self-injury reported lower grades than those who ha d not ever tried self-injury ( M = 7.01 vs. 7.57; t (674) = 5.82, p < .0001; Cohen’s d = -0.35). Having ever tried self-injury was statistically significantly associated with not going to school during the 30 days prior to survey administration because of feeling uns afe, but the effect was small ( r = .08, p < .01). Having ever tried self-injury was relate d to lower average scores on three key factors associated with adolescent developmen t, namely attitudes toward school, belief in possibilities, and parent communication (see Table 12). On average, students who reported they had tried self-injury reported less positive attitudes toward school, lower

PAGE 113

100 belief in their possibilities, and lowe r levels of parent communication ( p < .0001). Overall, small effects were noted with att itudes toward school a nd parent communication and a medium effect with belief in possibi lities. Attitudes toward school, belief in possibilities, and parent communi cation did not vary by gender ( p > .01). Table 12 Self-Injury and Developmental Theory Variables Self-injury Yes No Scale M SD M SD tb Cohen’s d Attitudes Toward School 3.50 0.99 3.84 0.90 6.25 -0.36 Belief in Possibilitiesa 4.20 0.89 4.67 0.55 10.18 -0.64 Parent Communication 1.24 0.85 1.46 0.79 4.86 -0.27 aThe results reported are for the original scale. Results from analysis conducted using the transformed scale were parallel (i.e., statistically si gnificant mean difference). bAll relationships reported were statistically significant ( p < .0001). Having been a victim of bullying, havi ng been a victim of cyberbullying, the frequency of having been a victim of cyberbul lying, and having been physically hurt by a boyfriend or girlfriend were defi ned as possible behavioral preci pitants of self-injury. All four behavioral precipitants demonstrated statistically significant relationships with having ever self-injured, a ll of which were in the small effect size range ( p < .0001; see Tables 13 and 14). Having been a victim of bullying and the frequency of having been a victim of cyberbullying in the past 30 days demonstrated the stronge st relationships with having ever tried self-injury (see Tables 13 and 14). Students who had not tried selfinjury reported a mean bullying score of 0.63, whereas those who had tried self-injury reported an average of 1.00 ( p < .0001). Males reported, on av erage, greater frequency of bullying than did females ( M = 0.80 vs. 0.67, Cohen d = 0.17). A greater proportion of

PAGE 114

101 females (26%) than males (19%) had ever b een cyberbullied; however, this relationship was negligible (Fis her’s Exact, N = 1,732, p < .01, Cramer’s V = .07). In terms of the frequency of cyberbullying, whereas 10% of students who had not ev er tried self-injury had been cyberbullied one or more times duri ng the month prior to survey administration, 20% of students who had ever tried self-injury had been cy berbullied. Males and females did not differ significantly in the frequency of having been a victim of cyberbullying ( p > .01). A greater proportion of males (10%) compared to females (5%) had been physically hurt by a girl/boyfriend in the past 12 months (Fisher’s Exact, N = 1,707, p < .0001, Cramer’s V = .10). Interestingly, however, a greater proportion of females who had been physically hurt by a boyfriend/girlfriend (56.5%) had ever self-injured compared to males who had been physically hurt by a girlfriend/boyfrie nd (45%). Table 13 Self-Injury and Preci pitants of Self-Inju ry (Chi-square tests of independence) Ever Self-Injured* Precipitants of Self-injury Yes (%) No (%) N Cramer’s V During your lifetime, have you ever been cyberbullied? 35 18 1740 .19 During the past 12 months, did your boyfriend or girlfriend ever hit, slap, or physically hurt you on purpose? 13 6 1715 .13 *All relationships reported were statistically significant ( p < .0001). Table 14 Self-injury and Precipitants of Se lf-injury (Independent t-tests) Precipitants of Self-injury* Yes No Scale M SD M SD t Cohen’s d Bully-Victim 1.00 0.87 0.63 0.72 -8.52 0.36 During the past 30 days, how many times were you the victim of cyberbullying? 0.30 0.71 0.12 0.44 -5.03 0.30 *All relationships reported were statistically significant ( p < .0001).

PAGE 115

102 Two out of three indicators of social contagion, knowing a friend who had harmed themselves on purpose and time spent on the computer or video games, demonstrated significant relationships with having ever tried self-injury, both of which were in the small effect size range (see Tables 15 and 16). Compared to those who had not ever selfinjured, a greater proportion of youth who had tried self-injury repor ted being aware of friends who had hurt themselves on purpose (Fisher’s Exact, N = 1,732, p < .0001; see Table 15). Females (54%) were significan tly more likely to know a friend who had harmed themselves than were males (38%; Fisher’s Exact, N = 1,724, p < .0001, Cramer’s V = .16). On average, youth who had ever tried self-injury spent a greater number of hours playing video games or usi ng a computer for fun on an average school day than those who had not ever tried self-injury ( p < .0001; see Table 16). Males spent significantly more time, on average, playing video games or using a computer for fun on an average school day than did females ( M = 2.60 vs. 1.94, p < .0001, Cohen’s d = 0.36). Table 15 Self-Injury and Social Contagion (C hi-square tests of independence) Ever Self-Injured Social Contagion Yes (%) No (%) N p -value Cramer’s V Have any of your friends hurt themselves on purpose (cutting, scratching, burning, not allowing wounds to heal, pinching)? 66 39 1732 <.0001 .25 Table 16 Self-Injury and Social Contagi on (Independent t-tests) Self-injury Social Contagion Yes No M SD M SD t p -value Cohen’s d Time on computer or video games 2.59 2.00 2.13 1.73 -4.29 <.0001 0.32 TV hours per day 3.09 1.79 3.01 1.73 -0.80 .42 0.05

PAGE 116

103 Having ever tried self-injury was statisti cally significantly associated with the multiple risk behaviors studied, including suicide, substance use, deviancy, sexual intercourse, and abnorma l eating behaviors ( p < .0001; see Tables 17-19). Self-injury demonstrated the st rongest relationship with suicide4 (Cohen’s d = 0.93; see Table 19). Youth who had self-inj ured, on average, scored statistically significantly higher on th e suicide scale than did those who had not self-injured ( p < .0001). There was no significant difference be tween males and females on suicide scale scores ( p > .01). Relationships between the indi vidual items included in the suicide scale (i.e., suicidal ideation, pla nning, and attempts) and self -injury were explored. Statistically significant and substantial relations hips (i.e., medium effect size) were noted between self-injury and suicidal id eation (Fisher’s Exact Test, N = 1,739, p < .0001, Cramer’s V = .44), having a suicide plan (Fisher’s Exact Test, N = 1,739, p< .0001, Cramer’s V = .39), and having attempted suicide (Fisher’s Exact Test, N = 1,739, p< .0001, Cramer’s V = .32). Whereas half of th e students who had ever tried self-injury reported thinking about suicide, only 10% of those who had not ever self-injured reported thinking about suicide. Most of those (66%) who had thoug ht about suicide also had tried self-injury. Whereas 5% of youth who ha d not tried self-injury had made a suicide plan, 35% of youth who had trie d self-injury had made a suic ide plan. The majority of those who had made a suicide plan had also tr ied self-injury (73%). Six percent of the sample ( n =103) had tried self-injury and attempte d suicide. Only 2% of youth who had 4 This relationship would be expected given the failure to distinguish between the two behaviors within the definition of self-injury (i.e., item lead-in).

PAGE 117

104 never tried self-injury had ever attempted suic ide. Most of those who had tried suicide had also tried self-injury (78%). Results suggested age, gender, race or et hnicity, grade, or school attended did not differentiate between those who had ever trie d self-injury and attempted suicide and those who had not tried both ( p > .01). Having tried both self-injury and suicide was statistically significantly associated with frequency of self-injury ( r = .32, p < .0001). Trying both behaviors was associated with in creased frequency of self-injury. Having tried both self-injury and suicide also was a ssociated with knowing a friend who harmed themselves on purpose; however, this relationship was weaker, 2(N = 1,726, 1) = 34.98, p< .0001, Cramer’s V = .14. Whereas 45% of st udents who had not tried both behaviors knew a friend who had harmed himself/he rself on purpose, 75% of students who had tried both behaviors knew friends who had harmed himself/herself on purpose. Having ever tried self-injury was statisti cally significantly associated with higher scores on the substance use scale (i.e., indicating greater use) ( p < .0001; see Table 19). Substance use scores did not differ by gender ( p > .01). In addition, youth who had tried self-injury were more likely to have sniffed glue, breathed the contents of spray cans, or inhaled any paints or sprays to get high ( p < .0001). The effect sizes for substance use were in the medium range (see Tables 18 and 19). Although relate d to substance use, having ever tried self-i njury was not statistica lly significantly associated with average age of first usage of alcohol, cigarettes, or marijuana ( p > .01; see Table 19).

PAGE 118

105 Table 17 Self-Injury and Substance Use Ever Self-injured* Yes (%) No (%) N Phi Cigarettes 42 18 1738 .25 Alcohol 52 30 1720 .20 Marijuana 22 11 1709 .14 Inhalants 32 8 1702 .30 Prescription 13 3 1706 .20 *Fisher’s Exact tests revealed statistical dependence between all substances and having ever tried selfinjury ( p < .0001). Having ever tried self-injury also was significantly associated with deviant behaviors, with relationships in the small effect-size range ( p < .0001; see Table 19). Deviant behaviors did not vary by gender ( p > .01). Having ever tried self-injury demonstrated a significant yet small relationshi p with sexual behavior (see Tables 18 and 19). A greater proportion of students who had ever tried self-injury had also had sexual intercourse ( p < .0001). However, having ever tried self-injury was not associated with age at first sexual intercourse or the number of sexual pa rtners among those who had had sexual intercourse ( p > .01; see Table 19). Table 18 Self-Injury and Problem Behaviors (C hi-square tests of Independence) Ever Self-Injured* Problem Behaviors Yes (%) No (%) N Cramer’s V Have you ever sniffed glue, or breathed the contents of spray cans, or inhaled any paints or sprays to get high? 32 8 1702 .30 Have you ever had sexual intercourse? 26 14 1605 .14 *All relationships reported were statistically significant ( p < .0001). Finally, having ever tr ied self-injury was statis tically significantly and substantially associated with the abnormal eating behaviors scale (Cohen’s d = 0.56, see Table 19). Students who had ha d ever tried self-injury were statistically significantly

PAGE 119

106 more likely, on average, to report abnormal ea ting behaviors such as fasting, using diet pills, powders, or liquids, or using laxatives to lose or control their weight than did those who had not tried self-injury ( p < .0001; see Table 19). Fe males, on average, reported higher levels of abnormal eati ng behaviors than did males ( M = 0.33 vs. 0.19, p < .0001, Cohen’s d = 0.23). Table 19 Self-Injury and Problem Behavior Co mparisons (Independent t-tests) Self-injury Yes No Variable/Scale M SD M SD t p -value Cohen’s d Abnormal Eating Scale* 0.54 0.84 0.16 0.47 -9.34 <.0001 0.56 Age at first alcohol use 10.22 1.87 10.43 2.05 1.26 .21 -0.11 Age at first cigarette use 10.66 1.76 10.74 2.00 0.38 .71 -0.04 Age at first marijuana use 11.64 1.68 11.49 1.97 -0.60 .55 0.08 Age at first sex 11.50 2.01 11.34 1.93 -0.67 .50 0.08 Deviant Behavior Scale* 0.20 0.96 -0.13 0.69 -6.64 <.0001 0.39 Number of sexual partners 1.85 0.87 1.89 0.86 0.40 .69 -0.05 Substance Use Scale* 0.20 0.80 -0.16 0.49 -9.08 <.0001 0.54 Suicide Scale 1.07 1.12 0.18 0.53 -16.72 <.0001 0.93 *Results reported are for the original scale. Results from analysis conducted using the transformed scale were parallel (i.e., statistically significant mean difference). Multilevel Logistic Regression Analyses To confirm relationships id entified at the bivariate level, multilevel logistic regression analysis was conducte d using the predictor variable s identified in Tables 7 and 8 and demographic variables (e.g., gender, race, grade) and having ever tried self-injury (outcome variable), which was coded as 1 (y es) and 0 (no) (see Table 19). Multilevel modeling was used because students (Level-1 ) were nested within schools (Level-2), thus, creating a lack of independence in the data. Only Level-1 st udent variables were used as predictors. The models were estimated usi ng penalized quasi-likelihood estimation (PQL) and conducted using HLM ve rsion 6. Odds ratios were reported.

PAGE 120

107 Six variables statistica lly significantly related ( p = .01) to having ever tried selfinjury while controlling for all other variables in the model: peer self-injury, having ever tried inhalants, grade level, belief in possibilities, abnorma l eating behaviors, and suicide (see Table 20). With the exception of suicid e (medium effect), all relationships were within the small effect size range. In terms of demographics, grade level was the only characteristic that emerged as statistically significant. Students in sixth grade were at decreased risk of having ever tr ied self-injury when compared to students in eighth grade (Odds Ratio [OR] = 0.80, p < .01). Abnormal eating behavi ors had the strongest effect on having ever tried self-injury, with an odds ratio of 3.76. Suicide demonstrated the second strongest relationship with having ever tried self-injury: as suicidal tendencies increased, the odds of having ever tr ied self-injury increased (OR = 2.82, p < .01). Two additional factors placed youth at risk for ha ving ever tried self-injury—peer self-injury and having ever tried inhalants. Youth w ho knew a friend who had harmed themselves on purpose were 1.84 times as likely to have harmed themselves on purpose as did those who did not know a friend who had self-harmed (OR = 1.84, p < .01). Youth who had tried inhalants were twice as likely to have tried self-inj ury as were youth who had not tried inhalants (OR = 2.06, p < .01). Youth who had a stronger belief in their possibilities were less likely to have tried self-injury (OR = 0.64, p < .01). Table 20 Multilevel Logistic Regression Analysis of Fa ctors that Predict Ha ving Ever Tried SelfInjury (N=1748) Predictor Coefficient p -value SE Odds Ratio 95% CI Femalea 0.34 .03 0.16 1.41 1.03, 1.94 Hit by boy/girlfriendb 0.56 .04 0.28 1.76 1.02, 3.03

PAGE 121

108Cyberbulliedb 0.28 .11 0.18 1.32 0.94, 1.87 Peer self-injuryb 0.61 .00 0.16 1.84 1.34, 2.54 Inhalant useb 0.72 .00 0.22 2.06 1.35, 3.16 TV viewing time 0.01 .78 0.05 1.01 0.92, 1.12 Sex (ever had) b -0.13 .58 0.23 0.88 0.56, 1.39 Video/computer use 0.08 .12 0.05 1.08 0.98, 1.19 Grades -0.07 .20 0.05 0.94 0.85, 1.04 Grade levelc -0.23 .01 0.08 0.80 0.68,.094 Attitudes toward school -0.01 .91 0.09 0.99 0.83, 1.18 Belief in possibilities -0.44 .00 0.15 0.64e 0.48, 0.87 Parent communication 0.18 .17 0.13 1.20 0.93, 1.56 Bully (victim) frequency 0.10 .01 0.04 1.10 1.02, 1.20 Abnormal eating behaviors 1.32 .00 0.40 3.76 1.79, 7.91 Substance use 0.05 .76 0.16 1.05 0.76, 1.45 Suicide 1.04 .00 0.10 2.82 2.32, 3.43 Deviant behavior -0.24 .04 0.11 0.79 0.63, 0.98 Blackd -0.26 .41 0.31 0.78 0.43, 1.42 Hispanicd -0.10 .70 0.25 0.91 0.56, 1.47 Other ethnicityd -0.22 .49 0.33 0.80 0.42, 1.52 aMale is the reference category. bNo is the reference category. cSixth grade is the reference category. dWhite is the reference category. eThe inverse of the odds ratio (1/.64 or 1.56) was used to judge the magnitude (i.e., Cohen’s Rule of Thumb). Given the strength of the relationship between self-injury and suicide, the multilevel logistic regression analysis was rer un with suicide removed from the model to determine whether suicide masked relationshi ps among other predictors in the model and self-injury. Three additional variab les became statistically significant ( p = .01): gender (OR = 1.54, 95% CI = 1.15, 2.08), having been hi t or pushed by a girlfriend or boyfriend (OR = 1.95, 95% CI = 1.19, 3.21), and the frequenc y of having been a victim of bullying (OR = 1.16, 95% CI = 1.08, 1.25). Once suicide was removed from the model, females were oneand-a-half times mo re likely to have ever self-injured than males ( p < .01). Finally, having been a victim of violence plac ed youth at increased risk for having ever

PAGE 122

109 tried self-injury compared to those who had not experienced violence at the hands of a boyfriend or girlfriend. Howeve r, the frequency of having b een a victim of bullying did not meet the minimal criterion for a small effect size (i.e., OR = 1.50). CHAID Analyses CHAID was used to explore interactions between predictors of having ever tried self-injury (i.e., the same predictors used in the multilevel analyses see Table 20) with the intent of identifying mutua lly exclusive, meaningful subgroups or segments at risk for having ever tried self-injury (see Figures 3 and 4). The training sample, which was created through randomly splitti ng the sample into two sepa rate samples using the 50% sample size option in CHAID, was used to de velop the model, and the test sample was used to examine classification accuracy. CHAID searches through the potential predictors to identify the predictor with the most significant relationship with the dependent or criterion variable —in this case, having ever trie d self-injury. This process is repeated until a stopping criterion is reache d. The overall alpha level was set at .01; however, Bonferroni adjustments were used to control for alpha infl ation resulting from simultaneous statistical testing. The size of the subgroups and th e availability of significant predictors were considered when assessing tree depth. An effect size, in addition to statistical significance, was used as the criterion for determining when to stop splitting (i.e., growing the tree ). Classification accuracy was determined by examining a crosstabulation of the actual categories of the cases and th eir predicted ca tegories using the model (i.e., the segmentation tree). The risk estimate, or the proportion of misclassified cases, is reported, as is the classification accuracy, or the proportion of correctly classified cases.

PAGE 123

110 The analysis began with a total traini ng sample of 901 cases (29% SI and 71% non-SI). CHAID analyses identified multiple interactions between predictors, with suicide, belief in possibilities, inhalant us e, gender, and substance use emerging as the best predictors of having ev er tried self-injury (see Figure 3). All relationships were within the small effect size range with the exception of suicide, which was within the medium range (see Table 21). The best predic tor of having ever self-injured, according to CHAID, was suicide ( p < .0001, Cramer’s V= .49; see Figure 3). Suicide was further divided into three distinct groups: (a) t hose who had not thought about, planned, or attempted suicide ( 0); (b) those who had a low level of suicidal tendencies (>0 to 1); and (c) those who had moderate to high levels of suicidal te ndencies (>1; see Figure 3). As seen in Figure 3, the segment at great est risk comprised female youth who have moderate to high levels of suic idal tendencies and used substan ces in the past. More than 97% of these students reported having injure d themselves on purpose. In contrast, the segment with the smallest proportion of youth who have se lf-injured had not thought about, planned, or attempted suicide, had hi gh belief in their possi bilities (>4), and had not used inhalants (12%, n =538). Inhalant use attenuate d the relationship between high belief in possibilities and having ever trie d self-injury, with 12% of those who had no suicidal tendencies, high beliefs and no inhalant use having tr ied self-injury compared to 33% of those with no suicidal tendencies, hi gh beliefs, and inhalant use having tried selfinjury (see Figure 3). There was a positive relationship between suicide and self-injury; youth who had self-injured had higher levels of suicidal tendencies than youth who had not self-injured (see Figure 3) Having a low level of suicidal tendencies (0, 1) interacted significantly with belief in possibilities: spec ifically, low belief pl aced youth at increased

PAGE 124

111 risk for having ever tried self-injury (92%, n = 13) and strong belief protected youth against having ever tried self-injury (43%, n = 99).5 Having a moderate to high level of suicidal tendencies (>1) interacted signifi cantly with gender: being female and reporting higher levels of substance use placing youth at risk for having ever tried self-injury (98%, n = 46). The overall model result ed in a classification accu racy of approximately 80% within the training sample (i.e., risk estimate = .20) and 79% within the test sample (i.e., risk estimate = .21).6 5 This interaction did not occur with the CHAID anal ysis conducted using the transformed variables. 6 The author was unable to locate guidelines for determining acceptable va lues for the risk estimate. The higher the classification, and conver sely, the lower the risk estimate, the better the model is in terms of performance.

PAGE 125

112 Category%n No70.81638 Yes29.19263 Total(100.00)901 Node 0 Category%n No21.9527 Yes78.0596 Total(13.65)123 Node 3 Category%n No12.169 Yes87.8465 Total(8.21)74 Node 9 Category%n No2.171 Yes97.8345 Total(5.11)46 Node 13 Category%n No28.578 Yes71.4320 Total(3.11)28 Node 12 Category%n No36.7318 Yes63.2731 Total(5.44)49 Node 8 Category%n No50.8957 Yes49.1155 Total(12.43)112 Node 2 Category%n 056.5756 143.4343 Total(10.99)99 Node 7 Category%n No7.691 Yes92.3112 Total(1.44)13 Node 6 Category%n No83.18554 Yes16.82112 Total(73.92)666 Node 1 Category%n No86.30504 Yes13.7080 Total(64.82)584 Node 5 Category%n No673931 Yes326115 Total(511)46 Node 11 Category%n No87.92473 Yes12.0865 Total(59.71)538 Node 10 Category%n No60.9850 Yes39.0232 Total(9.10)82 Node 4 Self-Injury (Training Sample) SUCIDE Adj. P-value=0.0000, Chi-square=212.8826, df=2 >1, GENDER Adj. P-value=0.0038, Chi-square=10.3896, df=1 Female, SUBSTANCE USE Adj. P-value=0.0045, Chi-square=11.3532, df=1 >-0.19750685378273475 <=-0.19750685378273475 Male >0 to <=1 BELIEF Adj. P-value=00083, Chi-square=10.9826, df=1 >3.3333333333333335, <=3.3333333333333335 <=0 BELEF Adj. P-value=0.0000, Chi-square=32.9684, df=1 >4, INHALE Adj. P-value=0.0003, Chi-square=15.1036, df=1 Yes, No <=4 Figure 3. Segmentation of having ever tried self-injury with suicide included in the model.

PAGE 126

113 Table 21 Effect Size Values for Segmentation of Having Ever Tried Self-Injury – Suicide Included Relationship Node Chi-square Cramer’s V Self-injury with suicide 0 212.8826 .49 Belief with suicide 1 32.9684 .22 Belief with suicide 2 10.9826 .31 Gender with suicide 3 10.3896 .29 Inhale with belief 5 15.1036 .16 Substance use with gender 9 11.3532 .39 Given the strength of the relationship be tween self-injury and suicide, the CHAID analysis was conducted with suicide remove d from the model to determine whether suicide masked relationships among other pred ictors in the model and self-injury (see Figure 4). CHAID analyses identified multiple interactions between predictors, with belief in possibilities, peer se lf-injury, inhalant use, and bullying emerging as the best predictors of having ever tried self-injury (s ee Figure 4). Interes tingly, once suicide was excluded from the model, gender and subs tance use were no longer statistically significant (see Figure 4). All relationships we re within the small effect size range with the exception of peer self-inj ury and belief, which was with in the large range (see Table 22). After eliminating suicide, the best pred ictor of having ever self-injured, based on CHAID results, was belief in possibilities ( p < .0001, Cramer’s V= .33; see Figure 4). Belief in possibilities demonstr ated a negative relationship with having ever tried selfinjury; as level of belief decr eased, the proportion of youth w ho had ever tried self-injury increased (see Figure 4). Belief in possibili ties was further divided into three groups roughly corresponding to those with low ( 3.33), medium (> 3.33 to 4.5) and high belief (>4.5; see Figure 4). As seen in Figure 4, the segment at greatest risk comprised youth with low belief in their possibilities ( 3.33) who knew a friend who had harmed

PAGE 127

114 themselves on purpose (88%, n = 58). In contrast, the segment with the smallest proportion of youth who have self-i njured had high belief in thei r possibilities (> 4.5), not used inhalants, and low bullying ( .80) (12%, n = 285). Low belief in possibilities statistically significantly and substantially (i.e., large effect size) inter acted with peer selfinjury: youth with low belief who knew a frie nd who had harmed themselves on purpose were at increased risk for self-injury ( p < .0001, Cramer’s V = .58). Having ever used inhalants significantly interacted with m oderate belief (> 3.33 to < = 4.5): whereas moderate belief appeared to protect against ha ving ever tried self-inj ury, having ever tried inhalant use attenuated this effect (see Figure 4). Sixty three percent ( n = 173) of those with moderate belief had never tried self -injury, and, similarly, 71% of those with moderate beliefs who had not tried inhalants had never tried self-i njury. However, 69% of those with moderate beliefs who had tried inhalants also had tried self-injury. High belief in possibilities (> 4.5) significantly interact ed with inhalant use ( p < .0001, Cramer’s V = .18). Among those with high be lief, a greater proportion of youth who had never tried inhalants also had never tried self-injury (see Figure 4). Inhalant use significantly interacted with the frequenc y of having been a victim of bullying ( p < .01, Cramer’s V = .18). Among youth with high belief and no history of inhalant use, those with a lesser frequency of having been a victim of bullying ( .80) were more likely to have never tried self-injury th an those with a greater freque ncy of having been a victim of bullying. The overall model resulted in a cl assification accuracy of approximately 77% within the training sample (i.e., risk estimate = .23) and 75% within the test sample (i.e., risk estimate = .25).

PAGE 128

115 Category%n No70.81638 Yes29.19263 Total(100.00)901 Node 0 Category%n No69.8174 Yes30.1932 Total(11.76)106 Node 4 Category%n No79.67435 Yes20.33111 Total(60.60)546 Node 3 Category%n No53.6622 Yes46.3419 Total(4.55)1 Node 10 Category%n No81.78413 Yes18.2292 Total(56.05)505 Node 9 Category%n No74.09163 Yes25.9157 Total(24.42)220 Node 12 Category%n No87.72250 Yes12.2835 Total(31.63)285 Node 11 Category%n No63.01109 Yes36.9964 Total(19.20)173 Node 2 Category%n No31.4311 Yes68.5724 Total(3.88)35 Node 8 Category%n No71.0198 Yes28.9940 Total(15.32)138 Node 7 Category%n No26.3220 Yes73.6856 Total(8.44)76 Node 1 Category%n No72.2213 Yes27.785 Total(1.00)18 Node 6 Category%n No12.077 Yes87.9351 Total(6.44)58 Node 5 SELF-INJURY (Training Sample) BELIEF Adj. P-value=0.0000, Chi-square=98.6805, df=3 >4.5 INHALE Adj. P-value=0.0000, Chi-square=18.5182, df=1 Yes No BULLY Adj. P-value=0.0007, Chi-square=15.4778, df=1 >0.80000000000000004 <=0.80000000000000004 >3.33 to <=4.5 INHALE Adj. P-value=0.0000, Chi-square=18.7702, df=1 Yes, No <=3.3333333333333335 PEER SELF-INJURY Adj. P-value=0.0000, Chi-square=25.6339, df=1 No Yes, Figure 4. Segmentation of having ever tried self-inju ry with suicide excluded from the model.

PAGE 129

116 Table 22 Effect Size Values for Segmentation of Having Ever Tried Self-Injury – Suicide Excluded Relationship Node Chi-square Cramer’s V Self-injury with belief 0 98.6805 .33 Belief with peer self-injury 1 25.6339 .58 Inhale with belief 2 18.7702 .33 Inhale with belief 3 18.5182 .18 Bully with inhale 9 15.4778 .18 Comparison of CHAID analys es conducted with the original versus transformed variables suggested the model excluding suicide was sensitive to nonnormality (see Figures 4 and 5). Whereas in the model c ontaining the original variables, belief in possibilities was the best predic tor of having ever self-injur ed (suicide excluded), when transformed variables were used, having ev er used inhalants emerged as the best predictor of having ev er self-injured ( p < .0001, Cramer’s V = .31; see Figure 5). Overall, the two models—original and transf ormed—were more similar than different, sharing the following best predic tors of having ever self-injur ed: inhalant use, belief in possibilities, and peer self-injury. In the tr ansformed model, having never used inhalants statistically significantly interacted with belief in possibilities (transformed); relative to those with lower belief, youth with higher belie f in their possibilities were more likely to have never tried self-injury ( p < .0001, Cramer’s V = .22). Knowing a friend who had harmed themselves on purpose statistically significantly interacted with belief in possibilities (transformed); peer self-injur y placed youth who had never tried inhalants but had low belief in their possib ilities at further risk for ha ving ever tried self-injury ( p < .01, Cramer’s V = .34). The frequency of which youth had been a victim of bullying significantly interacted with belief in possibilities (t ransformed); however, this relationship did not meet minimal criteria for a small effect size (see Table 23).

PAGE 130

117 Therefore, the decision was made to not grow the branch (i.e., Node 4, see Figure 5). Youth who had tried inhalants and knew a fr iend who had harmed themselves on purpose comprised the greatest proportion of youth who had injured themselves on purpose ( p < .01, Cramer’s V = .31). The overall model re sulted in a classification accuracy of approximately 78% within the training sample (i.e., risk estimate = .22) and 75% within the test sample (i.e., risk estimate = .25).

PAGE 131

118 Category%n No70.81638 Yes29.19263 Total(100.00)901 Node 0 Category%n No35.2044 Yes64.8081 Total(13.87)125 Node 2 Category%n No62.9617 Yes37.0410 Total(2.00)27 Node 6 Category%n No27.5527 Yes72.4571 Total(10.88)98 Node 5 Category%n No76.55594 Yes23.45182 Total(86.13)776 Node 1 Category%n No80.15541 Yes19.85134 Total(74.92)675 Node 4 Category%n No70.80160 Yes29.2066 Total(25.08)226 Node 10 Category%n No41.6710 Yes58.3314 Total(2.66)24 Node 14 Category%n No74.26150 Yes25.7452 Total(22.42)202 Node 13 Category%n No84.86381 Yes15.1468 Total(49.83)449 Node 9 Category%n No91.51194 Yes8.4918 Total(23.53)212 Node 12 Category%n No78.90187 Yes21.1050 Total(26.30)237 Node 11 Category%n No52.4853 Yes47.5248 Total(11.21)101 Node 3 Category%n No72.0931 Yes27.9112 Total(4.77)43 Node 8 Category%n No37.9322 Yes62.0736 Total(6.44)58 Node 7 SELF-INJURY (Training Sample) INHALE Adj. P-value=0.0000, Chi-square=89.0421, df=1 Yes PEER SELF-INJURY Adj. P-value=0.0006, Chi-square=11.6376, df=1 No Yes No, BELIEF TRANSFORMED Adj. P-value=0.0000, Chi-square=37.4747, df=1 >-0.41614683654714241, BULLY VICTIM Adj. P-value=0.0001, Chi-square=18.6746, df=1 >0.59999999999999998 SEX Adj. P-value=0.0027, Chi-square=11.0204, df=1 No;Yes <=0.59999999999999998 ATTITUDES TOWARD SCHOOL Adj. P-value=0.0034, Chi-square=13.8384, df=1 >3.6666666666666665 <=3.6666666666666665, <=-0.41614683654714241 PEER SELF-INJURY Adj. P-value=0.0020, Chi-square=11.5555, df=1 No Yes, Figure 5. Segmentation of having ever tried self-inju ry with suicide excluded from the model (transformed variables).

PAGE 132

119 Table 23 Effect Size Values for Segmentation of Havi ng Ever Tried Self-Injury – Suicide Excluded (transformed variables) Relationship Node Chi-square Cramer’s V Self-injury with inhale 0 89.0421 .31 Belief(transformed) with inhale 1 37.4747 .22 Peer self-injury with inhale 2 89.0421 .31 Peer self-injury with belief(transformed) 3 11.5555 .34 Bullying (victim) with belief(transformed) 4 18.6746 .03 Attitudes toward school with bully(victim) 9 13.8384 .05 Sex with bully(victim) 10 11.0204 .22 Relationships Between the Frequency of Self-injury and Other Variables This section addresses the outcome of frequency of self-injury (i.e., never, once, more than once). The analyses that are presen ted parallel those for having ever tried selfinjury. Frequency of self-inj ury was not statistica lly or meaningfully associated with gender, 2(N = 1,738, 4) = 7.12, p = .13, Cramer’s V = .06; race or ethnicity, 2(N = 1,725, 20) = 27.34, p = .13, Cramer’s V = .06; grade, 2(N = 1,746, 4) = 7.26, p = .12, Cramer’s V = .06; school attended, 2(N = 1,741, 28) = 35.90, p = .15, Cramer’s V = .07; or age, r = .00025, p = .99. Students who self-injured mo re frequently during the past 30 days tended to report poorer health than did those who se lf-injured less frequently ( r = .12, p < .0001). Frequency of self-injury was si gnificantly associated with not going to school during the 30 days prior to the survey administration because of feeling unsafe ( r = .15, p < .0001). As the frequency of self-injur y increased, the frequency of not going to school because of feeling unsafe increased. As the frequency of self-injury increased, self-reported academic performance tended to decrease ( r = -.17, p < .0001).

PAGE 133

120 The frequency of self-injury was associ ated with lower average scores on three key factors associated with adolescent deve lopment, including attitudes toward school, belief in possibilities, and pa rent communication (see Table 24). As the frequency of self-injury increased, attitudes toward sc hool, belief in possibilities, and parent communication decreased ( p < .0001). Table 24 Frequency of Self-Injury and Development Variables Developmental Theory Correlation (r)b Attitudes Toward School -.16 Belief in Possibilitiesa -.28 Parent Communication -.12 aResults reported are for the original scale. Results from analysis conducted using the transformed scale were parallel (i.e., statistically sign ificant, negative relationship). bAll relationships reported were statistically significant ( p < .0001). A one-way between-groups ANOVA revealed a statistically significant effect for frequency of self-injury on attitudes toward school, F (2,1531) = 28.17, p < .0001, 2 = .04. Tukey’s HSD test showed that all groups differed statistically significantly from one another, on average (see Figure 6). 3.83 3.64 3.26 1 1.5 2 2.5 3 3.5 4 4.5 5 NeverOnce2 or More Times Frequency of Self-injuryAttitudes toward School Scale Score (Average) Figure 6. Frequency of self-injury by attitudes toward school.

PAGE 134

121 A one-way between-groups ANOVA revealed a statistically significant effect for frequency of self-injury on belief in possibilities, F (2,1534) = 102.57, p < .0001, 2 = .12. Tukey’s HSD test showed that all groups diffe rently statistically significantly from one another, on average (see Figure 7). 4.65 4.41 3.88 1 1.5 2 2.5 3 3.5 4 4.5 5 NeverOnce2 or More Times Frequency of Self-injuryBelief in Possibilities Scale Score (Average) Figure 7. Frequency of self-injury by belief in possibilities. Finally, a one-way between-groups ANOVA revealed a statistical ly significant effect for frequency of self-injur y on parent communication, F (2,1539) = 12.23, p < .0001, 2 = .02. Tukey’s HSD test showed that students who had self-injured once, or more than once, differed statistically significantly fr om those who had never self-injured ( p < .05), with students who had never self -injured reporting, on average, statistically hi gher levels of parent communication than di d those who had self-injured once, or more than once. However, students who had self-injured once did not differ significan tly from those who had self-injured more than once ( p > .05). The sample mean s are displayed in Figure 8.

PAGE 135

122 1.45 1.3 1.15 0 0.5 1 1.5 2 NeverOnce2 or More Times Frequency of Self-injuryParent Communication Scale Score (Average) Figure 8. Frequency of self-injury by par ent communication scale scores. The frequency of self-injury was associat ed with all four precipitants of selfinjury studied ( p < .0001; see Table 25). Examination of the Spearman correlation coefficients suggested the fr equency of having been a vic tim of bullying demonstrated the strongest relative relationship with the fr equency of self-injur y (see Table 25). The frequency of self-injury was positively associated with having been a victim of bullying ( r = .24, p < .0001). As the frequency of bullying increased, the frequency of self-injury increased. A one-way between-g roups ANOVA revealed a statis tically significant effect for frequency of self-injury on the frequenc y of having been a victim of bullying, F (2,1742) = 50.77, p < .0001, 2 = .06. Tukey’s HSD test showed all groups differed statistically significantly fr om one another, with an average increase in bullying frequency in conjunction with the increase in frequency of se lf-injury (see Figure 9).

PAGE 136

123 0.63 0.92 1.18 0 0.5 1 1.5 2 2.5 3 3.5 4 NeverOnce2 or More Times Frequency of Self-injuryFrequency of Being a Victim of Bullying (Average) Figure 9. Frequency of self-injury by the frequen cy of having been a victim of bullying. Students who had been cyberbul lied reported self-injuring more frequently than did those who had not ever been cyberbullied ( p < .0001; see Table 25). Students who had been cyberbullied self-injured more fre quently than did those who had not, 2(N = 1,740, 4) = 43.73, p < .0001, Cramer’s V = .16. For example, whereas 4% of those who had not been cyberbullied self-injured four or more times during the past month, 9% of those who had been cyberbullied self-injured four or more times during the past month ( p < .0001). Further, as the frequency of self-injury incr eased, the frequency of having been a victim of cyberbullying increased ( r = .16, p < .0001). A one-way between-groups ANOVA revealed a statistically significant effect for frequency of self-injur y on the frequency of having been a victim of cyberbullying, F (2,1739) = 34.14, p < .0001, 2 = .04. Tukey’s HSD test showed all groups differed statistica lly significantly from one another, with an average increase in cyberbullying freque ncy in conjunction with the increase in frequency of self-injury (see Figure 10).

PAGE 137

124 0.12 0.21 0.47 0 0.5 1 1.5 2 2.5 3 3.5 4 NeverOnce2 or More Times Frequency of Self-injuryFrequency of Being a Victim of Cyberbullying (Average) Figure 10. Frequency of self-injury by freque ncy of having been a victim of cyberbullying. Finally, having been physically hurt by a boyf riend or girlfriend during the past 12 months was positively associated with frequency of self-injury ( p < .0001; see Table 25). Students who had been physically hurt by a boyfriend or girlfriend during the past 12 months self-injured more freque ntly than did those who had not, 2(N = 1,714, 4) = 57.82, p < .0001, Cramer’s V = .18. For example, whereas 4% of those who had not been physically hurt by a boyfriend or girlfriend self-i njured four or more times during the past month, 16% of those who had been physica lly hurt by a boyfriend or girlfriend selfinjured four or more times during the past month ( p < .0001). Table 25 Frequency of Self-Injury and Pr ecipitants of Self-Injury Precipitants of Self-injury Correlationa (r) Bully – Victim .24 During your lifetime, have you ever been cyberbullied? .15 During the past 30 days, how many times were you the victim of cyberbullying? .16 During the past 12 months, did your boyfriend or girlfriend ever hit, slap, or physically hurt you on purpose? .14 aAll relationships reported were statistically significant ( p < .0001).

PAGE 138

125 Three indicators of social contagion were investigated, including knowledge of peer self-injury and two t ypes of media exposure, computer and television. The frequency of self-injury was stat istically significantly associat ed with peer self-injury and time on computer or video games but not television viewing time (see Table 26). Knowing a friend who had harmed themselves on purpose was associated with a greater frequency of self-injury ( p < .0001). Among those who did not know of a friend who had harmed themselves on purpose, 83% also had never self-injured, 12% had self-injured one time during the past month, and 5% had se lf-injured two or mo re times during the past month. However, among those who did know a friend who had harmed themselves on purpose, 65% had never sel f-injured, 19% had self-injur ed one time during the past month, and 16% had self-injured two or more times during the past month, 2(N = 1,732, 4) = 88.98, p < .0001, Cramer’s V = .23. Finally, as time spent on the computer or playing video games increased, the fr equency of self-injury increased ( p < .0001). A one-way between-groups ANOVA revealed a statis tically significant effect for frequency of self-injury on time spent on the co mputer or playing video games, F (2, 1606) = 13.77, p < .0001, 2 = .02. Tukey’s HSD test showed that st udents who had self-injured once, or more than once, differed significantly fr om those who had never self-injured ( p < .05). However, students who had self-injured once did not differ significan tly from those who had self-injured more than once ( p > .05). The sample means are displayed in Figure 11. Overall, results suggested peer self-inj ury demonstrated a medium effect on the frequency of self-injury, and time on video or computer for fun during the school week demonstrated a small effect on the frequency of self-injury.

PAGE 139

126 Table 26 Frequency of Self-Injury and Social Contagion Social Contagion Correlation ( r) p-value Peer Self-injury .22 <.0001 Time on computer or video games .11 <.0001 TV hours per day .05 .07 2.13 248 2.85 0 1 2 3 4 5 6 7 NeverOnce2 or More Times Frequency of Self-injuryTime Using Computer or Video Games Figure 11. Frequency of self-injury by time spent using computer or video games for fun. The frequency of self-injury was stat istically significantly associated with abnormal eating behaviors, suicide, devian t behaviors, substance use, and sexual intercourse ( p < .01; see Table 27). The frequency of self-injury was statistic ally significantly and substantially (i.e., medium effect size) associated with abnormal eating behaviors including fasting, using diet pills, powders, or liqu ids, or using laxatives. As the number of abnormal eating behaviors increased, the frequenc y of self-injury increased ( p < .0001). A one-way between-groups ANOVA revealed a statistically significant eff ect for frequency of selfinjury, F (2, 1642) = 85.87, p < .0001, 2 = .09. Tukeys HSD test showed that students in all three frequency groupsne ver, once, and more than oncediffered statistically

PAGE 140

127 significantly ( p < .05), on average, in their self-re ported abnormal eating behaviors (see Figure 12). 0.17 0.41 0.78 0 0.5 1 1.5 2 2.5 3 NeverOnce2 or More Times Frequency of Self-injuryAbnormal Eating Behavior Scale Score (Average) Figure 12. Frequency of self-injury by abnormal eating scores. The frequency of self-injury was stat istically significantly and substantially associated with suicide scale scores ( p < .0001; see Table 27). As the frequency of selfinjury increased, scores on the suicide scale increased ( p < .0001; see Table 28). A oneway between-groups ANOVA revealed a statisti cally significant effect for frequency of self-injury on suicide scale scores, F (2, 1728) = 224.84, p < .0001, 2 = .21. Tukey’s HSD test showed that students in all three frequency groups—never, once, and more than once—differed statistically significantly ( p < .05), on average, in their self-reported suicidal tendencies (see Figure 13). Results were consistent across tests—as self-injury increased in frequency, suicid al tendencies increased. Table 27 Frequency of Self-Injury and Problem Behaviors Problem Behaviors Correlation ( r ) p-value Abnormal Eating Scale* .27 <.0001 Age at first alcohol use -.06 .11 Age at first cigarette use -.10 .08

PAGE 141

128Age at first marijuana use -.13 .07 Age at first sex .03 .64 Deviant Behavior Scale* .23 <.0001 Have you ever sniffed glue, or breathed the contents of spray cans, or inhaled any paints or sprays to get high? .27 <.0001 Have you ever had sexual intercourse? .19 <.0001 Substance Use Scale* .25 <.0001 Suicide Scale .39 <.0001 With how many people have you ever had sexual intercourse? .08 .19 *Results reported are for the original scale. Results from analysis conducted using the transformed scale were parallel (i.e., statistically significant, possible relationship of same magnitude). Table 28 Frequency of Self-Injury and Suicidal Ideation, Plans, and Attempts (N=1738) Frequency of Self-injury* Never (%) Once (%) 2 or more times (%) 2 Cramer’s V Thought 14 30 66 259.71 .39 Planned 7 21 50 262.07 .39 Suicide Tried 4 7 35 211.66 .35 *All relationships were statistically significant at p < .0001 with 2 degrees of freedom. 0.25 0.58 1.52 0 0.5 1 1.5 2 2.5 3 NeverOnce2 or More Times Frequency of Self-injurySuicide Scale Score (Average) Figure 13. Frequency of self-injury by suicide scale scores. Finally, the frequency of se lf-injury was statistically significantly associated with the deviancy scores ( p < .0001; see Table 27). As self-i njury increased, deviancy scores increased ( p < .0001). The relationship was within th e small to medium effect size range. A one-way between-groups ANOVA revealed a statistically significant effect for frequency of self-injur y on deviancy scores, F (2, 1591) = 46.47, p < .0001, 2 = .06.

PAGE 142

129 Tukey’s HSD test showed that students in all three frequency gr oups—never, once, and more than once—differed statistically significantly ( p < .05), on average, in their selfreported deviancy (see Figure 14). 0.05 0.54 -0.09 -0.44 0.06 0.56 1.06 NeverOnce2 or More Times Frequency of Self-injuryDeviancy Score (Average) Figure 14. Frequency of self-injury by deviancy scores. The frequency of self-injury was associat ed with substance use scores and having ever used inhalants ( p < .0001). A one-way between -groups ANOVA revealed a statistically significant effect for fre quency of self-injury on substance use, F (2, 1704) = 82.28, p < .0001, 2 = .09. Tukey’s HSD test showed th at students in all three frequency groups—never, once, and more than once —differed statistically significantly ( p < .05), on average, in their self-reported subs tance use levels (see Figure 15).

PAGE 143

130 0.08 0.59 -0.09 -0.43 0.07 0.57 1.07 NeverOnce2 or More Times Frequency of Self-injurySubstance Use Scores (Average) Figure 15. Frequency of self-injury by substance use scores Examination of Spearman correlation coeffici ents suggested medium effects between the frequency of self-injury and substance use (s ee Table 27). Compared to 6% of students who had not ever tried inhalant s, 31% of those who had trie d inhalants had self-injured two or more times during the past month, 2(N = 1,701, 4) = 178.72, p < .0001; see Table 29. Students who reported any of the substanc es studied tended to self-injure more frequently than did those who did not ( p < .0001). For example, whereas 7% of students who had not ever tried cigarette smoking self-i njured two or more times during the past month, 20% of those who had tried cigarette smoking self-injured two or more times during the past month, 2(N = 1,737, 7) = 94.91, p < .0001; see Table 29. The frequency of self-injury was not statistic ally significantly associated wi th the age of first alcohol, cigarette, or marijuana use ( p > .01; see Table 27). The frequency of self-injury demonstrat ed a statistically significant and small relationship with having ever had sexual inte rcourse (see Table 27). Youth who had ever had sexual intercourse self-inj ured more frequently than did those who had never had sexual intercourse ( p < .0001). Among those who had never had sex, 78% also had never

PAGE 144

131 self-injured, 15% had self-injured one tim e during the past month, and 7% had selfinjured two or more times during the past month. However, among those who had had sexual intercourse, 59% had neve r self-injured, 17% had self-i njured one time during the past month, and 24% had self-injured tw o or more times during the past month, 2(N = 1,604, 4) = 80.78, p < .0001, Cramer’s V = .22. The frequency of self-injury was not statistically significantly associated with the age of first sexual in tercourse or the number of sexual partners among those who had ever had sexual intercourse ( p > .01). Table 29 Frequency of Self-Injury and Having Ever Used Substances Frequency of Self-injury* Never (%) Once (%) 2 or more times (%) N 2 Cramer’s V Cigarettes 20 36 50 1737 93.52 .23 Alcohol 32 41 61 1719 59.67 .19 Marijuana 12 15 26 1708 23.15 .12 Inhalants 10 20 46 1701 161.46 .31 Prescription 3 7 22 1705 109.12 .25 *All relationships were statistically significant at p < .0001 with 2 degrees of freedom. Multilevel Logistic Regression Analyses Multilevel multinomial logistic regressi on was conducted to examine further the relationships between a subset of individual level predictor va riables and the frequency of self-injury in the past 30 days coded as: 0 (never), 1 (once), and 2 (more than once).7 The following comparisons were made: once versus never, more than once versus never, and once versus more than once. Standard errors were adjusted to ta ke into account the nested nature of the data. 7 For analyses purposes, the variable was reverse coded so the values being predicted would be “once” and “more than once”. Probabilities modeled were cumulate d over the lower-ordered values (i.e., “once” and “more than once”).

PAGE 145

132 The first comparison compared those who had self-injured once to those who had never self-injured in the past 30 days. Four variables statis tically significantly predicted ( p = .01) the frequency of self-injury while controlling for all other variables in the model: abnormal eating behavior s, peer self-injury, suicide, and grade level (see Table 30). Abnormal eating behaviors demonstrat ed the strongest relationship with the frequency of self-injury (see Table 30). The odds of self-injuring once compared to never in the past 30 days increased as abnormal eating behaviors increased ( p < .01; see Table 30). Peer self-injury demonstrated the second stronge st relationship. Youth were 1.72 times more likely to have self-injured onc e compared to never if they knew a friend who had harmed themselves on purpose ( p < .01; see Table 30). Suicide also was associated, with the odds of having self-inj ured once compared to never increasing as suicidal tendencies increased ( p < .01; see Table 30). Alt hough statistically significant, the magnitude of the odds ratio for grade level di d not meet criteria for a small effect size (i.e., OR = 1.50). Table 30 Multilevel Logistic Regression Analysis of Factors that Predict the Frequency of SelfInjury – Once versus Never (Past 30 Day Frequency) (N=1748) Predictor Coefficient p -value SE Odds Ratio 95% CI Femalea -0.14 0.44 0.17 0.87 0.62, 1.23 Hit by boy/girlfriendb -0.15 0.65 0.33 0.86 0.46, 1.64 Cyberbulliedb 0.22 0.26 0.19 1.24 0.85, 1.81 Peer self-injuryb 0.54 0.00 0.18 1.72 1.21, 2.44 Inhalant useb 0.43 0.07 0.24 1.54 0.96, 2.47 TV viewing time 0.01 0.79 0.05 1.02 0.91, 1.13 Sex (ever had) b 0.01 0.97 0.25 1.01 0.62, 1.64 Video/computer use 0.06 0.29 0.05 1.06 0.95, 1.18 Grades -0.02 0.66 0.06 0.98 0.87, 1.09 Grade levelc,e -0.24 0.01 0.09 0.78 0.66, 0.93 Attitudes toward -0.03 0.76 0.10 0.97 0.80, 1.17

PAGE 146

133school Belief in possibilities -0.32 0.05 0.16 0.73 0.53, 1.00 Parent communication 0.05 0.73 0.15 1.05 0.79, 1.40 Bully (victim) frequency 0.10 0.02 0.04 1.10 1.01, 1.20 Abnormal eating behaviors 1.31 0.00 0.40 3.69 1.70, 8.05 Substance use 0.15 0.39 0.18 1.16 0.82, 1.65 Suicide 0.49 0.00 0.11 1.63 1.33, 2.01 Deviant behavior -.30 0.02 0.13 0.74 0.57, 0.95 Blackd 0.68 0.02 0.28 1.99 1.14, 3.46 Hispanicd -0.08 0.76 0.27 0.92 0.55, 1.56 Other ethnicityd 0.25 0.45 0.33 1.28 0.67, 2.45 aMale is the reference category. bNo is the reference category. cSixth grade is the reference category. dWhite is the reference category. e The inverse of the odds ratio (1/.78 or 1.28) was used to judge the magnitude (i.e., Cohen’s Rule of Thumb). The second comparison compared those w ho had self-injured more than once to those who had never self-injured in the pa st 30 days. Four variables statistically significantly predicted ( p = .01) the frequency of self-injury while controlling for all other variables in the model: suicide, having ever tr ied inhalants, belief in possibilities, and the frequency of having been a vi ctim of bullying. Suicide demonstrated the strongest relationship. As suicidal tendencies increas ed, the odds of having self-injured twice or more compared to never increased ( p < .01; see Table 31). Having ever tried inhalants demonstrated the second strongest relationshi p. Youth who had ever tried inhalants were 1.54 times more likely than their non-inhaling c ounterparts to have se lf-injured twice or more during the past 30 days ( p < .01; see Table 31). The ex tent to which youth believed in their possibilities was associated with the frequency of self-injury. As belief in possibilities increased, the odds of having self-i njured twice or more compared to never decreased (see Table 31). Although the relatio nship between bullying (victim frequency)

PAGE 147

134 was statistically significant, the magnitude of the odds ratio did not meet minimal criteria for a small effect size (i.e., OR = 1.50). Table 31 Multilevel Logistic Regression Analysis of Factors that Predict the Frequency of SelfInjury – More than Once versus Never (Past 30 Day Frequency) (N=1748) Predictor Coefficient p -value SE Odds Ratio 95% CI Femalea 0.54 0.03 0.25 1.72 1.06, 2.78 Hit by boy/girlfriendb 0.74 0.03 0.34 2.11 1.08, 4.10 Cyberbulliedb 0.17 0.48 0.25 1.19 0.73, 1.93 Peer self-injuryb 0.64 0.01 0.26 1.89 1.14, 3.12 Inhalant useb 0.92 0.00 0.28 2.52 1.47, 4.31 TV viewing time 0.07 0.32 0.07 1.08 0.93, 1.24 Sex (ever had) b 0.60 0.04 0.29 1.82 1.03, 3.22 Video/computer use 0.13 0.05 0.07 1.14 1.00, 1.30 Grades -.07 0.34 0.07 0.93 0.81, 1.07 Grade levelc -0.18 0.13 0.12 0.83 0.66, 1.06 Attitudes toward school -0.00 0.98 0.13 1.00 0.78, 1.28 Belief in possibilitiese -0.49 0.01 0.19 0.61 0.42, 0.88 Parent communication 0.29 0.14 0.20 1.33 0.91, 1.96 Bully (victim) frequency 0.15 0.00 0.05 1.16 1.04, 1.29 Abnormal eating behaviors 0.93 0.05 0.47 2.53 1.00, 6.38 Substance use -0.14 0.50 0.20 0.87 0.59, 1.30 Suicide 1.04 0.00 0.11 2.84 2.27, 3.55 Deviant behavior -0.14 0.30 0.14 0.87 0.66, 1.14 Blackd 0.19 0.65 0.43 1.21 0.52, 2.81 Hispanicd 0.03 0.94 0.34 1.03 0.52, 2.01 Other ethnicityd -0.01 0.98 0.47 0.99 0.37, 2.49 aMale is the reference category. bNo is the reference category. cSixth grade is the reference category. dWhite is the reference category. e The inverse of the odds ratio (1/.61 or 1.64) was used to judge the magnitude (i.e., Cohen’s Rule of Thumb). The final comparison was made between those who had self-injured more than once and those who had self-injured once in the past 30 days. Only one variable, suicide, statistically significan tly distinguished ( p = .01) the two groups. As suicidal tendencies

PAGE 148

135 increased, the odds of have self-injured more than once compared to once increased ( p < .01; see Table 32). Table 32 Multilevel Logistic Regression Analysis of Factors that Predict the Frequency of SelfInjury – More than Once versus Once (Past 30 Day Frequency) (N=1748) Predictor Coefficient p -value SE Odds Ratio 95% CI Femalea 0.67 0.01 0.27 1.96 1.16, 3.31 Hit by boy/girlfriendb 0.90 0.02 0.38 2.45 1.17, 5.14 Cyberbulliedb -0.05 0.86 0.27 0.95 0.57, 1.61 Peer self-injuryb 0.09 0.74 0.28 1.10 0.63, 1.91 Inhalant useb 0.49 0.10 0.30 1.64 0.91, 2.93 TV viewing time 0.06 0.46 0.08 1.06 0.91, 1.24 Sex (ever had) b 0.59 0.07 0.32 1.80 0.96, 3.36 Video/computer use 0.07 0.30 0.07 1.08 0.94, 1.24 Grades -0.04 0.57 0.08 0.96 0.83, 1.11 Grade levelc 0.06 0.64 0.13 1.06 0.82, 1.38 Attitudes toward school 0.03 0.85 0.14 1.03 0.79, 1.34 Belief in possibilities -0.18 0.37 0.20 0.84 0.57, 1.23 Parent communication 0.24 0.27 0.22 1.27 0.83, 1.94 Bully (victim) frequency 0.05 0.37 0.06 1.05 0.94, 1.18 Abnormal eating behaviors -.038 0.43 0.48 0.68 0.27, 1.75 Substance use -0.29 0.18 0.21 0.75 0.50, 1.14 Suicide 0.55 0.00 0.12 1.74 1.37, 2.21 Deviant behavior 0.16 0.32 0.16 1.17 0.86, 1.60 Blackd -0.49 0.28 0.46 0.61 0.25, 1.49 Hispanicd 0.11 0.78 0.38 1.12 0.53, 2.35 Other ethnicityd -0.26 0.60 0.50 0.77 0.29, 2.05 aMale is the reference category. bNo is the reference category. cSixth grade is the reference category. dWhite is the reference category. CHAID Analyses In addition to multinomial logistic regression, CHAID was used to explore interactions between predictors of the frequency of self-injur y (i.e., never, once, or more

PAGE 149

136 than once in the past 30 days ) with the intent of identifying mutually exclusive, meaningful subgroups or segments (see Figures 16 and 17). The anal ysis began with a total training sample of 900 cases (72% never, 16% one time, and 12% two or more times). CHAID analyses identified multiple interactions between predictors, with suicide, belief, Hispanic ethnic ity, and inhalant use emerging as the best predictors of the frequency of self-injury (see Figure 16). All relationships were within the small effect size range (see Table 33). The best predictor of the frequency of se lf-injury, according to CHAID, was suicide ( p < .0001, Cramer’s V= 0.31; see Fi gure 16). Suicide was further divided into three distinct groups: (a) t hose who had not thought about, planned, or attempted suicide ( 0); (b) those who had a low level of suicidal tendencies (> 0 to 1); and (c) those who had moderate to high levels of suicidal tende ncies (> 1; see Figure 16). As seen in Figure 16, the segment at greatest ri sk of frequent self-i njury comprised those who had a moderate to high level of suicid al tendencies, were non-Hispanic, and had used inhalants (66%, n = 47). In contrast, the largest proportion of youth who had never self-injured in the past 30 days were those students w ho had not thought about, planned, or attempted suicide ( 0), used the computer or played video games, on average, for less than 1 hour per day, and had not used inhalants (90%, n = 339). There was a positive relationship between suicide and the frequenc y of self-injury; as suicidal tendencies increased, the proportion of youth who had self -injured at least once in the past month increased (see Figure 16). Low suicidal tend encies statistically significantly interacted with time spent playing video games or on th e computer for fun; however, examination of the effect size (Cramer’s V = .09) suggest ed this relationship did not meet minimal criteria for a small effect size. Thus, th e decision was made to stop growing the low

PAGE 150

137 suicidal tendency branch at Node 1 (see Fi gure 16). Belief statistically significantly interacted with moderate suicidal tendencies ( p < .01, Cramer’s V = .20). Lower belief ( 4) placed youth with moderate suicidal tend encies at risk for greater frequencies of self-injury compared to those with stronger belief (> 4). For example, whereas, among those with moderate suicidal tendencies ( n = 115), 10% of those with higher belief had self-injured more than once in the past 30 da ys, over 26% of those with relatively lower belief had self-injured more than once in the past 30 days ( p < .01, Cramer’s V = .20). High suicidal tendencies statistically significan tly interacted with Hispanic ethnicity ( p < .01, Cramer’s V = .29). Among youth w ith high suicidal tendencies ( n = 126), Hispanic youth (56%) when compared to youth of ot her ethnicities (26%) were statistically significantly more likely to have never self-injured. Among non-Hispanic youth (i.e., White, Black, Other) with high suicidal tendenci es, those who had used inhalants were at increased risk for self-injuring more frequently than those who had never tried inhalants. For example, 66% ( n = 110) of those who had used inha lants had self-injured more than once in the past 30 days compared to 38% ( n = 110) of those who had not ever tried inhalants ( p < .01, Cramer’s V = .29). The overa ll model resulted in a classification accuracy of approximately 75% within the trai ning sample (i.e., risk estimate = .25) and 75% within the test sample (i .e., risk estimate = .25).

PAGE 151

138 Category%n Never72.22650 1 time16.22146 2 or more11.56104 Total(100.00)900 Node 0 Category%n Never60.006 1 time 10.001 2 or more30.003 Total(1.11)10 Node 4 Category%n Never30.1638 1 time25.4032 2 or more44.4456 Total(14.00)126 Node 3 Category%n Never56.259 1 time37.506 2 or more6.251 Total(1.78)16 Node 10 Category%n Never26.3629 1 time23.6426 2 or more50.0055 Total(12.22)110 Node 9 Category%n Never14.897 1 time19.159 2 or more65.9631 Total(5.22)47 Node 14 Category%n Never34.9222 1 time26.9817 2 or more38.1024 Total(7.00)63 Node 13 Category%n Never60.0069 1 time25.2229 2 or more14.7817 Total(12.78)115 Node 2 Category%n Never71.6058 1 time18.5215 2 or more9.888 Total(9.00)81 Node 8 Category%n Never32.3511 1 time41.1814 2 or more26.479 Total(3.78)34 Node 7 Category%n Never82.74537 1 time12.9484 2 or more4.3128 Total(72.11)69 Node 1 Category%n Never76.16214 1 time17.4449 2 or more6.4118 Total(31.22)281 Node 6 Category%n Never87.77323 1 time9.5135 2 or more2.7210 Total(0.89)368 Node 5 Category%n Never65.5219 1 time20.696 2 or more13.794 Total(3.22)29 Node 12 Category%n Never8968304 1 time85529 2 or more1776 Total(3767)339 Node 11 FREQUENCY OF SELF-NJURY (Training Sample) SUICIDE Adj. P-value=00000, Chi-square=174.9629, df=3 >1 HISPANIC Adj. P-value=0.0033, Chi-square=10.6414, df=1 Yes No, NHALE A dj. P-value=0.0079, Chi-square=90515, df=1 Yes No, >0 to <=1 BELIEF Adj. P-value=0.0025, Chi-square=13.2308, df=1 >4, <=4 <=0 VIDEO Adj. P-value=0.0015, Chi-square=14.5608, df=1 >2, <=2 INHALE Adj. P-value=0.0010, Chi-square=12.9649, df=1 Yes, No Figure 16. Segmentation of frequency of self-injury with suicide included in the model.

PAGE 152

139 Table 33 Effect Size Values for Segmentation of Fre quency of Self-Injury – Suicide Included Relationship Node Chi-square Cramer’s V Frequency of self-injury with suicide 0 174.9629 .31 Video with suicide 1 14.5608 .09 Belief with suicide 2 13.2308 .20 Hispanic with suicide 3 10.6414 .29 Inhale with video 5 12.9649 .19 Inhale with Hispanic 9 9.0515 .29 Comparison of CHAID analys es conducted with the original versus transformed variables suggested the model including suic ide was sensitive to nonnormality. Overall, the models containing the origin al and transformed variables we re similar. However, the transformed model introduced two new ‘best pr edictors’ of the freque ncy of self-injury: substance use and bullying (v ictim). Whereas belief in possibilities statistically significantly interacted with low suicidal tendencies in the model using the original variables (see Figure 16), it did not do so in the model using transf ormed variables (see Figure 17 and Table 34). Instead, in the model using the transformed variables, substance use significantly interacted with low suicidal tendencies ( p < .01, Cramer’s V = .20). Among youth with low suicidal tend encies, the largest proportion of youth who self-injured most frequently comprised yout h who had higher relative substance use and a higher relative frequency of being a victim of bullying ( p < .01, Cramer’s V = .11). The overall transformed model resulted in a cl assification accuracy of approximately 76% within the training sample (i.e., risk estimate = .24) and 76% within the test sample (i.e., risk estimate = .24).

PAGE 153

140 Category%n Never72.22650 1 time16.22146 2 or more11.56104 Total(100.00)900 Node 0 Category%n Never60.006 1 time10.001 2 or more30.003 Total(1.11)10 Node 4 Category%n Never30.1638 1 time 25.4032 2 or more44.4456 Total(1400)126 Node 3 Category%n Never56.259 1 time37.506 2 or more6.251 Total(1.78)16 Node 10 Category%n Never26.3629 1 time23.6426 2 or more50.0055 Total(12.22)110 Node 9 Category%n Never14897 1 time19.159 2 or more659631 Total(522)47 Node 16 Category%n Never34.9222 1 time26.9817 2 or more38.1024 Total(7.00)63 Node 15 Category%n Never600069 1 time252229 2 or more147817 Total(1278)115 Node 2 Category%n Never45.4530 1 time33.3322 2 or more21.2114 Total(7.33)66 Node 8 Category%n Never15003 1 time35007 2 or moe500010 Total(222)20 Node 14 Category%n Never58.7027 1 time32.6115 2 or more8.704 Total(5.11)46 Node 13 Category%n Never79.5939 1 time14.297 2 or more6.123 Total(5.44)49 Node 7 Category%n Never82.74537 1 time12.9484 2 or more4.3128 Total(72.11)649 Node 1 Category%n Never76.16214 1 time17.4449 2 or more6.4118 Total(31.22)281 Node 6 Category%n Never87.77323 1 time9.5135 2 or more2.7210 Total(40.89)368 Node 5 Category%n Never65.5219 1 time20.696 2 or more13.794 Total(3.22)29 Node 12 Category%n Never8968304 1 time85529 2 or more1776 Total(3767)339 Node 11 Category%n Never91.72277 1 time7.2822 2 or more0.993 Total(33.56)302 Node 18 Category%n Never72.9727 1 time18.927 2 or more8.113 Total(4.11)37 Node 17 FREQUENCY OF SELF-INJURY (Training Sample) SUICIDE Adj. P-value=0.0000, Chi-square=1749629, df=3 >1 HISPANIC Adj. P-value=0.0033, Chi-square=10.6414, df=1 Yes No, NHALE A dj. P-value=0.0079, Chi-square=9.0515, df=1 Yes No, >0 to <=1 SUBSTANCE USE TRANSFORMED Adj. P-value=0.0028, Chi-square=13.6840, df=1 -1.1272465585119151, BULLY VICTM Adj. P-value=0.0002, Chi-square=17.1556, df=1 >12 <=1.2 <=-1.1272465585119151 <=0 VIDEO Adj. P-value=0.0015, Chi-square=14.5608, df=1 >2, <=2 NHALE A dj. Pvalue=00010, Chi-square=12.9649, df=1 Yes, No BELIEF TRANSFORMED Adj. P-value=0.0096, Chi-square=110769, df=1 <=-0.41614683654714241 >-0.41614683654714241, Figure 17. Segmentation of frequency of self-injury with suicide included in the model (transformed variables).

PAGE 154

141 Table 34 Effect Size Values for Segmentation of Fr equency of Self-Injury – Suicide Included (transformed variables). Relationship Node Chi-square Cramer’s V Frequency of self-injury with suicide 0 174.9629 .31 Video with suicide 1 14.5608 .09 Substance use with suicide 2 13.6840 .20 Hispanic with suicide 3 10.6414 .29 Inhale with video 5 12.9649 .19 Bully(victim) with substance use 8 17.1556 .11 Inhale with Hispanic 9 9.0515 .29 Belief with inhale 11 11.0769 .18 Given the strength of the relationship be tween self-injury and suicide, the CHAID analysis was conducted with suicide remove d from the model to determine whether suicide masked relationships among other pred ictors in the model and the frequency of self-injury (see Figure 18). The analysis be gan with a total training sample of 900 cases (72% Never, 16% one time, and 12% two or more times). CHAID analyses identified multiple interactions between predictors, w ith abnormal eating behaviors, peer selfinjury, belief in possibilities, having ever had sexual intercourse, and being Black emerging as the best predictors of frequency of self-injury (see Figure 18). Interestingly, once suicide was excluded from the model, Hi spanic ethnicity and inhalant use were no longer statistically significant (see Figure 18). All relationships were within the small effect size range with the exception of peer self-injury and abnormal eating behavior, which was within the medium range (see Table 35). The best predic tor of the frequency of self-injury in the reduced model, accord ing to the CHAID results, was abnormal eating behaviors ( p < .0001, Cramer’s V = .21; see Figure 18). Abnormal eating behaviors demonstrated a positive relationship with fre quency of self-injury; as abnormal eating

PAGE 155

142 behaviors increased, the proporti on of youth who self-injured more than once during the past 30 days increased (see Figure 18). A bnormal eating behaviors were further divided into three groups: (a) those who had not gone without eating for 24 hours or more to lose weight, taken diet pills, powde rs, or liquids without a doctor’ s advice to lose weight or keep from gaining weight, or vomited or take n laxatives to lose weight or keep from gaining weight ( 0); (b) those who had a low level of abnormal eating behavior (> 0 to 1); and (c) those who had moderate to high le vels of abnormal eating behavior (> 1; see Figures 18). As seen in Figure 18, the segment at greatest risk is comprised of students with moderate to high levels of abnorma l eating behaviors and who know a friend who had harmed themselves on purpose (50%, n = 48). In contrast, the segment with the largest proportion of youth w ho had never self-injured in the past 30 days comprised youth who had no abnormal eating behaviors, did not know a friend who had harmed themselves on purpose, a nd was not Black (87%, n = 348). No abnormal eating behaviors significantly interacted with know ing a peer who had harmed themselves on purpose ( p <.0001, Cramer’s V = .21). Examinati on of the branch that contained those youth who knew a friend who had harmed th emselves on purpose suggested those who had never had sexual intercourse and had high belief in their possibilities were more likely to have never self-injured in the past 30 days (see Figure 18). On the other hand, peer self-injury and having ever had sexual intercourse attenuated the effect of no abnormal eating behaviors on the frequency of self-injury (see Figure 18). Examination of the branch that contained those yout h who did not know a friend who had harmed themselves on purpose suggested when compar ed to White, Hispanic, and Other ethnic youth, a greater proportion of Black youth self-i njured more frequently. For example,

PAGE 156

143 among youth with no abnormal eating behaviors and who did not know a friend who selfharmed ( n = 389), approximately 3% of non-Black yout h self-injured more than once in the past 30 days compared to more than 12% of Black youth ( p < .01, Cramer’s V = .16). Low abnormal eating behaviors (> 0 to 1) statistically significantly interacted with belief; relatively higher belie f protected youth with modera te abnormal eating behaviors against a higher frequency of self-injury (see Figure 8; p < .01, Cramer’s V = .15).8 Moderate to high abnormal eati ng behaviors statistically sign ificantly inte racted with knowing a friend who had harmed themselves on purpose; peer self-injury placed youth with moderate to high levels of abnormal ea ting behaviors at further risk for increased frequency of self-injury ( p < .01, Cramer’s V = .40). The overall model resulted in a classification accuracy of approximately 75 % within the training sample (i.e., risk estimate = .25) and 77% within the test sample (i.e., risk estimate = .23). 8 This interaction did not occur in the CHAID analysis conducted using transformed variables.

PAGE 157

144 Category%n Never72.22650 1 time16.22146 2 or more11.56104 Total(100.00)900 Node 0 Category%n Never28.8117 1 time27.1216 2 or more44.0726 Total(6.56)59 Node 3 Category%n Never72.738 1 time9.091 2 or more18.182 Total(1.22)11 Node 9 Category%n Never18.759 1 time31.2515 2 or more50.0024 Total(5.33)48 Node 8 Category%n Never63.10106 1 time20.2434 2 or more16.6728 Total(18.67)168 Node 2 Category%n Never69.0587 1 time19.8425 2 or more11.1114 Total(14.00)126 Node 7 Category%n Never452419 1 time21.439 2 or more333314 Total(4.67)42 Node 6 Category%n Never78.31527 1 time14.2696 2 or more7.4350 Total(74.78)673 Node 1 Category%n Never85.35332 1 time11.0543 2 or more3.6014 Total(43.22)389 Node 5 Category%n Never68.2928 1 time19.518 2 or more12.205 Total(4.56)1 Node 13 Category%n Never87.36304 1 time10.0635 2 or more2.599 Total(38.67)38 Node 12 Category%n Never68.66195 1 ime18.6653 2 or more12.6836 Total(31.56)284 Node 4 Category%n Never42.8618 1 time23.8110 2 or more33.3314 Total(4.67)42 Node 11 Category%n Never73.14177 1 time17.7743 2 or more9.0922 Total(26.89)242 Node 10 Category%n Never77.09175 1 time15.4235 2 or more7.4917 Total(25.22)227 Node 15 Category%n Never13.332 1 time53.338 2 or more33.335 Total(1.67)15 Node 14 FREQUENCY OF SELF-INJURY (Training Smple) ABNORMAL EATING Adj. P-value=0.0000, Chi-square=77.857, df=2 >1 PEER SELF-INJURY Aj. P-value=0.0019, Chi-square=96668, df=1 No Yes >0 to <=1, BELIEF Aj. P-value=00092, Chi-square=10.7945, df=1 >4,missing> <=4 <=0 PEER SELF-NJURY Adj. P-value=0.0000, Chi-square=30.6411, df=1 No, BLACK Adj. P-value=0.0038, Chi-square=10.3980, df=1 Yes, No Yes SEX Aj. P-value=0.0001, Chi-square=18.4491, df=1 Female Male, BELIEF Adj. P-value=0.0001, Chi-square=20.1112, df=1 >3.3333333333333335, <=3.3333333333333335 Figure 18. Segmentation of frequency of self-injury with suicide excluded from the model.

PAGE 158

145 Table 35 Effect Size Values for Segmentation of Fr equency of Self-Injur y – Suicide Excluded Relationship Node Chi-square Cramer’s V Frequency of self-injury with abnormal eating 0 77.4857 .21 Peer self-injury with abnormal eating 1 30.6411 .21 Belief with abnormal eating 2 10.7945 .15 Peer self-injury with abnormal eating 3 9.6668 .40 Sex with peer self-injury 4 18.4491 .25 Black with peer self-injury 5 10.3980 .16 Belief with sex 10 20.1112 .29 Relationships between Peer Se lf-injury and Other Variables This section addresses the outcome vari able of knowing a friend who has harmed themselves on purpose (yes, no). Gender demo nstrated a statistically significant and small effect on knowing a friend who had injured themselves on purpose, 2(N = 1,724, 1) = 44.03, p < .0001, Cramer’s V = .16. Compared to 38% of males, 54% of females knew of friends who had injured themselves on purpose. Knowing a friend who had selfinjured was not statistically or practically associated with race or ethnicity, 2(N=1711, 5) = 9.92, p = .08, Cramer’s V = .08. Knowing a friend who had injured themselves on purpose was statistically and substantia lly associated with grade level, 2(N = 1,732, 1) = 82.54, p < .0001, Cramer’s V = .22. Compared to 35% of sixth graders, 57% of eighth graders knew of a friend who had injured them selves on purpose. School attended had a statistically significant and sm all effect on knowing a friend who had injured themselves on purpose, 2(N=1727, 7) = 29.78, p < .01, Cramer’s V = .13. The percentage of students who knew a friend who had injured themselves on purpose ranged from a low of 36.4% at School 4 to a high of 60.6% at Schoo l 5 (see Table 36). Finally, students who reported knowing a friend who had injured themselves on purpose were, on average,

PAGE 159

146 statistically signi ficantly older ( M = 12.76 years) than did those who did not ( M = 12.23 years), t (1729) = -7.87, p < .0001, Cohen’s d = 0.38. Table 36 Prevalence (%) of Knowing a Friend Who Had Self-Injured by School Attended School 1 2 3 4 5 6 7 8 Yes 44.8% 51.7% 44.0% 36.4% 60.6% 44.4% 41.8% 46.2% 95% CI 43.6 – 46.0 50.5 – 52.9 42.8 – 45.2 35.3 37.6 59.4 – 61.8 43.2 – 45.6 40.6 – 43.0 45.0 – 47.4 Knowing a friend who had harmed themselves on purpose was associated with lower average scores on two out of thr ee key factors associated with adolescent development, including attitude s toward school and belief in possibilities (see Table 37). On average, students who knew a friend who had harmed themselves on purpose reported lower attitudes toward school and lo wer beliefs in their possibilities ( p < .0001). Both relationships were in the small effect size range, with a strong er relative effect on belief in possibilities. Table 37 Developmental Theory Variables (Independent t-tests) Peer Self-injury Yes No Scale M SD M SD t p -value Cohen’s d Attitudes Toward School 3.63 0.99 3.85 0.87 4.53 <.0001 -0.24 Belief in Possibilitiesa 4.38 0.81 4.68 0.54 8.70 <.0001 -0.44 Parent Communication 1.37 0.81 1.43 0.81 1.40 0.16 -0.09 aResults reported are for the original scale. Results from analysis conducted using the transformed scale were parallel (i.e., statistically significant mean difference). All four behavioral precip itants were statistically significantly associated with knowing a friend who had harmed themselves on purpose ( p < .01; see Tables 38 and 39). However, examination of the associated effect sizes for each suggested, whereas bullying

PAGE 160

147 and cyberbullying demonstrated small effects on the frequency of se lf-injury, the effect of having been physically hurt by a boyfrie nd or girlfriend on purpose was negligible (i.e., did not meet minimum re quirements for a small effect). Among those who had ever been cyberbullied, a greater proportion knew a friend who had harmed themselves on purpose than did those who did not know a friend who had harmed themselves on purpose, 2(N = 1,725, 1) = 80.50, p < .0001, Cramer’s V = .22. Table 38 Peer Self-injury and Precipitants of Self-In jury (Chi-square tests of independence) Peer Self-Injury Precipitants of Self-injury Yes (%) No (%) N p -value Cramer’s V During your lifetime, have you ever been cyberbullied? 32 14 1725 <.0001 .22 During the past 12 months, did your boyfriend or girlfriend ever hit, slap, or physically hurt you on purpose? 10 5 1701 .00 .08 Knowing a friend who had harmed th emselves on purpose was statistically significantly associated with having been a victim of bullying and the frequency with which students had been a vi ctim of cyberbullying ( p < .0001). On average, students who knew a friend who had harmed themselves on purpose reported a gr eater frequency of being bullied than did those who did not ( p < .0001). Finally, students who knew a friend who had harmed themselves reported, on averag e, a higher frequency of being the victim of cyberbullying ( p < .0001). Table 39 Peer Self-injury and Precipitants of Self-Injury (Independent t-tests) Peer Self-injury* Precipitants of Self-injury Yes No M SD M SD t Cohen’s d Bully-Victim 1.81 1.98 1.27 1.74 -6.04 0.29 During the past 30 days, how many 0.49 1.46 0.16 0.70 -5.79 0.29

PAGE 161

148times were you the victim of cyberbullying? *All relationships reported were statistically significant ( p < .0001). Knowing a friend who had harmed th emselves on purpose was statistically significantly associated with the time spent on the computer or playing video games ( p < .01). Students who knew a friend who had harmed themselves on purpose, on average, spent more time on the computer than did those who did not ( p < .01). Neither met the minimum criteria for a small effect size (see Table 40). Table 40 Peer Self-Injury and Social Cont agion (Independent t-tests) Peer Self-injury Yes No Scale M SD M SD t p -value Cohen’s d Time on computer or video games 1.65 1.66 1.44 1.45 -2.75 .01 0.13 TV hours per day 2.26 1.53 2.11 1.57 -1.98 .05 0.10 Knowing a friend who had harmed th emselves on purpose was statistically significantly associated with multiple risk be haviors studied, including suicide, substance use, sexual intercourse, devianc y, and abnormal eating behaviors ( p < .0001; see Tables 41 and 42). There was a statistically significant and meaningful relationship (i.e., small to medium effect size) between knowing a fr iend who had harmed themselves on purpose and suicide ( p < .0001). Students who knew a friend who had harmed themselves reported higher suicide scores than did t hose who did not know a friend who had harmed themselves on purpose ( p < .0001; see Table 42). For ex ample, whereas 12% of those who reported not knowing a friend who had harmed themselves had thought about

PAGE 162

149 suicide, 32% of those who reported knowing a friend had harmed themselves had thought about suicide (Fisher’s Exact, N = 1,726, p < .0001). Knowing a friend who had harmed th emselves on purpose was statistically significantly and substantially (i.e., medium effect size) a ssociated with substance use scores and inhalant use ( p < .0001). Knowing a friend who had harmed themselves on purpose was associated with higher substan ce use scores, and a greater proportion of those who had used inhalants also were expos ed to peer self-injury (23% vs. 8%; see Table 41). Age at first al cohol use was statistically si gnificantly associated with knowing a friend who had harmed themselves on purpose ( p < .01). The magnitude of the relationship was within th e small effect size range. Y outh who had been exposed to peer self-injury were, on average, older at first alcohol use than did those who had not been exposed to peer self-injury. Knowledge of peer self-injury was not statistically significantly associated with age at first ci garette use or age at first marijuana use ( p > .01). Knowing a friend who had harmed th emselves on purpose demonstrated a statistically significant and small relationshi p with having ever had sexual intercourse (see Table 41). A greater proportion of those who knew a friend who had harmed themselves on purposes had had sexual interc ourse compared to those who did not know a friend who had harmed themselves on purpose ( p < .0001). Knowing a friend who had harmed themselves on purpose was not statistica lly significantly asso ciated with age of first sexual intercourse or the number of se xual partners among those who had ever had sexual intercourse ( p > .01).

PAGE 163

150 Table 41 Peer Self-injury and Problem Behaviors (Chi-square tests of independence) Peer Self-Injury* Problem Behaviors Yes (%) No (%) N Cramer’s V Have you ever sniffed glue, or breathed the contents of spray cans, or inhaled any paints or sprays to get high? 23 8 1689 .21 Have you ever had sexual intercourse? 21.5 14 1591 .10 *All relationships reported were statistically significant ( p < .0001). Knowing a friend who had harmed th emselves on purpose was statistically significantly associated with deviancy scores ( p < .0001; see Table 42). On average, youth who knew a friend who had harmed them selves reported statis tically significantly higher deviancy scores than did youth w ho did not know a friend who had harmed themselves on purpose ( p < .0001). Examination of eff ect sizes suggested a small association between knowing a friend who ha d self-harmed and deviancy. Knowing a friend who had harmed th emselves on purpose was statistically significantly and substantially (i.e., small to medium effect size) associated with abnormal eating behaviors ( p < .0001; see Table 42). St udents who knew a friend who had harmed themselves on purpose, on averag e, reported more abnormal eating behaviors than did those who did not know a friend who had harmed themselves on purpose ( p < .0001). Table 42 Problem Behavior Comparisons (Independent t-tests) Peer Self-injury Yes No Scale M SD M SD t p -value Cohen’s d Abnormal Eating Scale* 0.14 0.25 0.05 0.16 -8.66 <.0001 0.43 Age at first alcohol use 10.76 1.95 10.20 1.98 -3.41 .00 0.28 Age at first cigarette use 10.88 1.82 10.26 1.97 -2.53 .01 0.33 Age at first marijuana use 11.76 1.72 11.07 2.05 -2.34 .02 0.36

PAGE 164

151Age at first sex 11.46 1.98 11.41 1.90 -0.22 .83 0.03 Deviant Behavior Scale* 0.12 0.89 -0.18 0.64 -7.53 <.0001 0.39 Number of sexual partners 1.87 0.87 1.90 0.87 0.34 .73 -0.03 Substance Use Scale* 0.11 0.74 -0.21 0.42 -10.96 <.0001 0.53 Suicide Scale 0.64 0.98 0.24 0.65 -9.97 <.0001 0.48 *Results reported are for the original scale. Results from analysis conducted using the transformed scale were parallel (i.e., statistically significant mean difference). Multilevel Logistic Regression Analyses Multilevel logistic regressi on with the Bernoulli distri bution option at level-l was conducted to examine further relationships betw een a subset of individual level predictor variables and knowing a friend who had harmed themselves on purpose (see Table 43). Standard errors were adjusted to take into account the nested nature of the data. Five variables statistically significantly predicted ( p = .01) peer self-inj ury while controlling for all other variables in the model: gender, having ever been cybe rbullied, having ever tried self-injury, grade level, and substance us e (see Table 42). All relationships were in the small effect size range. In terms of de mographics, gender and grade level emerged as statistically significant. Gende r demonstrated the strongest relationship with peer selfinjury. Females were 2.25 times more likely than were males to know a friend who had harmed themselves on purpose ( p < .01). Eighth graders were 1.55 times more likely than sixth graders to know a friend who had harmed themselves on purpose ( p < .01). Youth who had ever been cyberbullied were almost twice as likely to know a friend who had harmed themselves on purpose than did those who had not ( p < .01). Youth who had themselves ever tried self-injury were 1.88 times more likely than were those who had not to know a friend who had harmed themselves on purpose ( p < .01). Finally, higher levels of substance use increased the pr obability of knowing a friend who had harmed themselves on purpose (OR = 1.51, p < .01).

PAGE 165

152 Table 43 Multilevel Logistic Regression Analysis of Factors that Predict Peer Self-Injury (N=1748) Predictor Coefficient p -value SE Odds Ratio 95% CI Femalea 0.81 .00 0.13 2.25 1.74, 2.91 Hit by boy/girlfriendb 0.54 .03 0.26 1.72 1.05, 2.84 Cyberbulliedb 0.65 .00 0.15 1.92 1.42, 2.60 Self-injuryb 0.63 .00 0.18 1.88 1.31, 2.69 Frequency of selfinjury -0.03 .70 0.07 0.97 0.85, 1.11 Inhalant useb 0.47 .03 0.21 1.61 1.06, 2.43 TV viewing time 0.04 .37 0.04 1.04 0.96, 1.13 Sex (ever had) b -0.49 .02 0.20 0.62 0.41, 0.91 Video/computer use 0.05 .31 0.05 1.05 0.96, 1.14 Grades -0.04 .45 0.05 0.97 0.88. 1.06 Grade levelc 0.44 .00 0.07 1.55 1.36, 1.76 Attitudes toward school 0.05 .55 0.07 1.05 0.90, 1.21 Belief in possibilities -0.14 .29 0.14 0.87 0.66, 1.13 Parent communication -0.09 .40 0.11 0.91 0.73, 1.13 Bully (victim) frequency 0.09 .02 0.04 1.09 1.02, 1.17 Abnormal eating behaviors 0.65 .08 0.37 1.92 0.93, 3.97 Substance use 0.41 .01 0.16 1.51 1.12, 2.05 Suicide 0.23 .02 0.10 1.26 1.04, 1.54 Deviant behavior 0.01 .94 0.10 1.01 0.83, 1.23 Blackd -0.46 .07 0.25 0.63 0.39, 1.03 Hispanicd -0.08 .68 0.20 0.92 0.62, 1.36 Other ethnicityd 0.65 .02 0.26 1.92 1.15, 3.19 aMale is the reference category. bNo is the reference category. cSixth grade is the reference category. dWhite is the reference category. CHAID Analyses In addition to multilevel logistic re gression, CHAID was used to explore interactions between predictors of the peer self-injury (i.e., knowing a friend who had harmed themselves on purpose) with the inte nt of identifying mutually exclusive, meaningful subgroups or segments (see Figure 19). The analysis began with a total

PAGE 166

153 training sample of 894 cases (48% Yes a nd 52% No). CHAID analyses identified multiple interactions between predictors, with substance use, grade level, frequency of self-injury, abnormal eating behaviors, gender, grades, and having ever been cyberbullied emerging as the best predictors of peer self -injury (see Figure 19). Interestingly, suicide did not enter the model. All relationships we re within the small effect size range, with the exception of grade level by academic grades (see Table 44). The best predictor of peer self-injury, according to CHAID, was substance use ( p < .0001, Cramer’s V= .28; see Figure 19). Substance use was furthe r divided into three distinct groups, approximately corresponding to low ( -.39), moderate (> -.39 to 27), and high (> .27, see Figure 19). As seen in Figure 9, the segmen t at greatest risk, which is large enough to be worthy of consideration, comprised youth wi th high levels of substance use (> .27) and who have self-injured at leas t once during the past 30 days (90%, n = 77). In contrast, the segment with the largest propor tion of youth who did not know a friend who had harmed themselves on purpose comprised yo uth who had low levels of substance use (< .39), were in sixth grade, and had no abnormal eating behaviors (74%, n = 236). There was a positive relationship between substance use and peer self-injury; as substance use increased, the proportion of youth who knew a friend who had harmed themselves on purpose increased (see Figure 19). Low substance use statistically significantly interacted with grade level; youth in sixth grade were more likely to not know a friend who had harmed themselves on purpose (70%, n = 464) than had youth in eighth grade (54%) ( p < .01, Cramer’s V = .16). Among sixth graders, low substance using youth, a greater frequency of abnormal eating behaviors placed them at risk for knowing a friend who had harmed themselves on purpose ( p < .01, Cramer’s V = .26).

PAGE 167

154 Among eighth graders, low substance using youth, and being a female placed them at risk for knowing a friend who had harmed themselves on purpose ( p < .01, Cramer’s V = .27). Moderate substance use statistically si gnificantly interacted w ith grade level; youth in 8th grade were more likely to know a fr iend who had harmed themselves on purpose ( p < .01, Cramer’s V = .23). Among modera te substance using youth, grade level statistically significantly and substantially (i .e., medium effect) inte racted with academic grades; however, examination of proportions su ggested a lack of a meaningful difference in peer self-injury based on academic pe rformance (see Figure 19). Among eighth graders, moderate substance using youth, havi ng ever been a vict im of cyberbullying placed them at increased risk for knowing a friend who had harmed themselves on purpose ( p < .01, Cramer’s V = .28). High substa nce use statistically significantly interacted with frequency of self-injury; youth who self-inj ured once or more during the past 30 days were more likely to know a friend who had harmed themselves on purpose (78%, n =174) than were those who had not harmed themselves (60%) (see Figure 19; p < .0001, Cramer’s V = .14). Among non-self-i njuring, high substance using youth, high grades placed them at risk for knowing a fr iend who had harmed themselves on purpose ( p < .01, Cramer’s V = .25). The overall m odel resulted in a cla ssification accuracy of approximately 64% within the training sample (i.e., risk estimate = .36) and 64% within the test sample (i.e., risk estimate = .36).

PAGE 168

155 Category%n Yes48.32432 No51.68462 Total(100.00)894 Node 0 Category%n 17241126 0275948 Total(1946)174 Node 3 Category%n 189.169 010.398 Total(8.1)77 Node 9 Category%n 158.7657 041.2440 Total(10.85)97 Node 8 Category%n 110.001 090.009 Total(112)10 Node 20 Category%n 1861131 013.895 Total(4.03)36 Node 19 Category%n 1490225 0509826 Total(570)1 Node 18 Category%n 151.95133 048.05123 Total(28.64)256 Node 2 Category%n 161.0494 038.9660 Total(17.23)154 Node 7 Category%n 183.7831 016.226 Total(414)37 Node 17 Category%n 153.8563 0461554 Total(13.09)117 Node 16 Category%n 138.2439 061.7663 Total(11.41)102 Node 6 Category%n 127.6321 072.3755 Total(8.50)76 Node 15 Category%n 169.2318 030.778 Total(2.1)26 Node 14 Category%n 137.28173 062.72291 Total(51.90)464 Node 1 Category%n 146.4692 053.54106 Total(2215)198 Node 5 Category%n 1571468 042.8651 Total(13.1)119 Node 13 Category%n 130.3824 069.6255 Total(8.84)79 Node 12 Category%n 1304581 06955185 Total(2975)266 Node 4 Category%n 163.3319 036.6711 Total(3.36)30 Node 11 Category%n 1262762 07373174 Total(2640)236 Node 10 PEER SELF-INJURY (Training Smple) SUBSTANCE USE Adj. P-value=00000, Chisquare=64.4306, df=2 >027 FREQUENCY OF SELFINJURY Adj. Pvalue=0.0000, Chisquare=204475, df=1 >0 <=0 GRADES Adj. P-value=00002, Chisquare=22.921, df=2 >8 >7 to <=8 <=7,missing> >-0.39 to <=0.27 GRADE LEVEL Adj. P-value=0.0003, Chisquare12.7824, df1 8th CYBERBULLIED Adj. P-value=0011, Chisquare=10.5941, df1 Yes No 6th GRADES Adj. P-value=00015, Chi-square141955, df1 >6, <=6 <=0.39022339226033559, GRADE LEVEL Adj. P-value=00004, Chi-square12.4480, df1 8th GENDER Adj. P-value=0.0002, Chi-square13.616, df1 Female Male 6h ABNORMAL EAING Adj. P-value=0.0002, Chisquare17.2631, df1 >0 <=0,missing> Figure 19. Segmentation of knowledge of peer self -injury with suicide included in the model

PAGE 169

156 Table 44 Effect Size Values for Segmen tation of Peer Self-Injury Relationship Node Chi-square Cramer’s V Peer self-injury with substance use 0 64.4306 .28 Grade level with substance use 1 12.4480 .16 Grade level with substance use 2 12.7824 .23 Frequency of self-injury with substance use 3 20.4475 .14 Abnormal eating with grade level 4 17.2631 .26 Gender with grade level 5 13.6716 .27 Grades with grade level 6 14.1955 .38 Cyber with grade level 7 10.5941 .28 Grades with frequency of self-injury 8 22.9221 .25 Cognitive Interviewing Cognitive interviewing with a small sample of middle school aged youth ( n = 4) was conducted as a part of the dissertation re search to identify possi ble issues with the items (e.g., problematic words) and to doc ument item validity (i.e., whether items measured what they were intended to meas ure). Four middle school aged females were interviewed. Interviews lasted approximat ely 10 minutes. Participants were asked to read the survey items to themselves and then were asked to repeat the questions in their own words. All participants were able to re peat the questions in their own words. One participant equated hurting themselves on purpose specifically w ith cutting. All participants agreed that pinching did not fit with hur ting themselves on purpose and should not be included in the definition of self-harm. Two participants mentioned specific behaviors that were left out, including ripping hair out and banging ones head into the wall. Youth were asked why people their age harm themselves on purpose. Most responses were consistent with the definiti on provided—to relieve distress, or, as one youth described it, “Because they don’t know how to let their anger exit.” One youth

PAGE 170

157 stated some may try it because “it’s cool.” Another emphasized the role of the media and “everywhere else” in encourag ing some youth to try self-injury when they “see no other option.” Youth were asked how often someone th eir age could hurt themselves on purpose in a month. Two youth stated the existing re sponse options were enough to cover (e.g., 4 or 5 times per month). The other two youth st ated you could self-injure as often as once per day, necessitating a revision of the response options. All youth were able to respond to the peer self-injury item. All stated friends tell one another when they have harmed themselv es on purpose. When asked how they tell, they stated they tell one another in person—not online, unless necessary (i.e., they cannot see one another in person). At school, one youth explained, “most people try to hide it but you can tell by the way they act.” When as ked to explain, she st ated, “they wear the same long sleeve shirts every day,” and they pul l the sleeves over their thumbs so the cuts will not show. While most youth “act like they don’t want people to see” while at school, some youth may be “attention whores” and use self-injury as a means to gain attention from their peers. These youth roll up these sleeves and show their injuries freely, and some cut during class in front of their peers (and teachers). Summary This study describes the prevalence of self-injury in the general middle school population and the relationships between self -injury and other risk behaviors during middle school. This study also identified m eaningful segments of youth who self-injure, and factors that place them at risk or protect them from self-injury (see Table 45). A

PAGE 171

158 summary of key findings, provided in Table 45 and Appendix D, will be discussed in Chapter 5. Table 45 Study Research Questions, Proced ures, and Key Findings (N = 1748) What is the prevalence of self-injury among middle school youth? 28.4% (95% CI = 26.3,30.5; N = 1734) What is the frequency of self-injury among youth who self-injure? N =495 (had ever tried self-injury) 35% once during past month 18% two or three different times during past month 5.5% four or five different times during past month 11% six or more different times during past month What proportion of middle school youth know someone who self-injures? 46.8% (95% CI = 45.6%, 48.0%) What demographic, attitudinal, and behavioral variables are related to self-injury? Having ever tried self-injury was associated with (suicide included): self-reported poor health (bivariate test only) peer self-injury (OR = 1.84, 95% CI = 1.34, 2.54) inhalant use (OR = 2.06, 95% CI = 1.35, 3.16) belief in possibilities (OR = .64, 95% CI = 0.48, 0.87) abnormal eating behaviors (OR = 3.76, 95% CI = 1.79, 7.91) suicide scale scores (OR = 2.82, 95% CI = 2.32, 3.43) In addition, having ever tried self-injury was associated with (suicide excluded): gender (OR = 1.54, 95% CI = 1.15,2.08) having been hit or pushed by a girlfriend or boyfriend (OR = 1.95, 95% CI = 1.19, 3.21) Frequency of self-injury was associated with: self-reported poor health (bivariate test only) self-reported frequency of not going to school b ecause of feeling unsafe (bivariate test only) Frequency of self-injury – Once vs. Never (Past 30 days) was associated with: abnormal eating behaviors (OR = 3.69, 95% CI 1.70, 8.05) peer self-injury (OR = 1.72, 95% CI = 1.21, 2.44) suicide scale scores (OR = 1.63, 95% CI = 1.33, 2.01) Frequency of self-injury – More than Once vs. Never (Past 30 days) was associated with: suicide (OR = 2.84, 95% CI 2.27, 3.55) having ever tried inhalants (OR = 2.52, 95% CI = 1.47, 4.31) belief in possibilities (OR = 0.61, 95% CI = 0.42, 0.88) Frequency of self-injury – More than Once vs. Once (Past 30 days) was associated with: suicide (OR = 1.74, 95% CI 1.37, 2.21)

PAGE 172

159 Peer self-injury was associated with: gender (more common knowledge among females, OR = 2.25, 95% CI = 1.74, 2.91) grade level (more common knowledge among 8th graders, OR = 1.55, 95% CI = 1.36, 1.76) school attended (range = 36.4% 60.6%) age (more common among older youth, mean = 12.23 years) age at first alcohol use (those exposed to peer self-injury, older, on average) having ever been cyberbullied (OR = 1.92, 95% CI = 1.42, 2.60) having ever tried self-injury (OR = 1.88, 95% CI = 1.31, 2.69) Are there gender, racial or ethnic, ag e, grade, and school differences in rates of self-injury, frequency of self-injury, and knowledge of friends who self-injure? Neither having ever tried self-injury nor the frequency of self-injury were associated with gender, race/ethnicity, grade, school attende d, or age (bi/multivariate). Peer self-injury varied according to gender, grade, age, and school attended. Knowing a friend who had self-injured was more common among females, 8th graders, older youth, and students in Schools 2 (~52%) and 5 (~61%), for example. Where does self-injury fit in with other risk behaviors such as alcohol use, tobacco use, suicide, and deviance? Having ever tried self-injury was associated with: inhalant use (OR = 2.06, 95% CI = 1.35, 3.16) abnormal eating behaviors (OR = 3.76, 95% CI = 1.79, 7.91) suicide scale scores (OR = 2.82, 95% CI = 2.32, 3.43) Frequency of self-injury – Once vs. Never (Past 30 days) was associated with: abnormal eating behaviors (OR = 3.69, 95% CI 1.70, 8.05) suicide scale scores (OR = 1.63, 95% CI = 1.33, 2.01) Frequency of self-injury – More than Once vs. Never (Past 30 days) was associated with: suicide (OR = 2.84, 95% CI 2.27, 3.55) having ever tried inhalants (OR = 2.52, 95% CI = 1.47, 4.31) Frequency of self-injury – More than Once vs. Once (Past 30 days) was associated with: suicide (OR = 1.74, 95% CI 1.37, 2.21) Peer self-injury was associated with: age at first alcohol use (those exposed to peer self-injury, older, on average) having ever tried self-injury (OR = 1.88, 95% CI = 1.31, 2.69) Are there meaningful segments of youth who self-injure? If so, what characteristics are usef ul in defining each segment? Having ever tried self-injury : Segment at greatest risk : female youth who have moderate to high levels of suicidal tendencies and used substances in the past Segment at greatest risk (suicide excluded) : youth with low belief in possibilities and who know a friend who has harmed themselves on purpose Segment at greatest risk (suicide excluded, transformed variables) : youth who have tried inhalants and know a friend who has harmed themselves on purpose Segment at least risk : youth who have not thought about, planned, or attempted suicide, have high belief in

PAGE 173

160their possibilities, and have not used inhalants Segment at least risk (suicide excluded) : youth with high belief in their possibilities, who have not used inhalants, and report low levels of bullying (victim) Frequency of self-injury: Segment at greatest risk : those who have a moderate to high level of suicidal tendencies, are non-Hispanic, and have used inhalants Segment at greatest risk (suicide excluded) : youth with moderate to high levels of abnormal eating behaviors who know a friend who have harmed themselves on purpose Segment at least risk : those who have not thought about, planned, or attempted suicide Segment at least risk (suicide excluded) : youth with no abnormal eating behaviors, who do not know a friend who have harmed them selves on purpose, and of non-Black race/ethnicity Peer Self-injury : Segment at greatest risk : youth with high levels of substance use and who have self-injured at least once during the past 30 days Segment at least risk: youth with low levels of substance use, in 6th grade, and have no abnormal eating behaviors

PAGE 174

161 Chapter Five: Discussion Introduction This chapter begins with a review of the purposes of the research. An overview of study methods is provided, as is a summa ry of findings. Results are discussed in relation to theory and previ ous research. Similarities a nd differences in the results compared to previous studies are highlighte d. Limitations of the study are discussed. Plans for disseminating the results of this study are presented, along with implications for prevention and further research. Purposes of the Research This study sought to increase what is known about superfic ial/moderate selfinjury among the general adol escent population, including factor s related to the behavior, especially those amenable to change and us eful in identifying vulnerable youth (Favazza, 1998; Purington & Whitlock, 2004). This study had three purposes: (a) contribute to what is known about self-injury among early adolescents in the general middle school population, (b) identify behaviors that are co morbid with self-inj ury, and (c) identify segments of youth who self-injure. Overa ll, the study focused on moderate/superficial self-injury as a distinct behavioral phenom enon with multiple causes and functions. For the purposes of this study, self-injury was defined as the performance of a harmful behavior such as cutting, scratching, burni ng, not allowing wounds to heal, or pinching, by a person who feels upset as a way to feel better (less upset) This study provided

PAGE 175

162 general adolescent population estimates of th e prevalence, 30-day frequency rates of injury among self-injurers, a nd information about the extent to which adolescents knew a friend who self-injured. Relatio nships between self-injury an d other risk behaviors were described. Segmentation analyses were used to identify factors associated with selfinjury among middle school youth and se gments of youth who self-injure. Recommendations (e.g., Gratz, 2003) to examin e sociocultural and ge nder variations in the prevalence, frequency, and correlates of self-injury were followed. The relation between the environment (e.g., self-reported e xposure to peers who self-injure, exposure to bullying and violence in the school setti ng, social climate) and individual behavior (i.e., having ever tried self-i njury and 30-day frequency ra te of self-injury) were considered (see Dishion & Dodge, 2005). Overview of Method This study involved a secondary analysis of data gathered using the middle school YRBS from sixthand eighth-grade stude nts in eight middle schools in a large, southeastern county in Florida. The YRBS is a school-based classroom survey of risk behaviors self-reported by middle school youth. Approximately 2,350 surveys were distributed across schools. A total of 2,003 valid surveys were completed, resulting in an initial response rate of 85.23%. Only stude nts who self-reported attending one of the eight middle schools, reported being in sixt h or eighth grade, re sponded to the having ever tried self-injury item and did not report responding untruthfully were retained, resulting in a total study sa mple of 1,748 students. Thr ee items were developed to measure three aspects of self-injury: life time prevalence—Have you ever hurt yourself on purpose (cutting, scratching, burning, not allowi ng wounds to heal, pinching)?; past 30-

PAGE 176

163 day prevalence—During the past month, how often have you hurt yourself on purpose (cutting, scratching, burning, not allowing wounds to heal, pinching)?; and awareness of peer self-injury behavior—Have any of your friends hurt themselves on purpose (cutting, scratching, burning, not allowing wounds to heal, pinching)? In addition to demographic items (e.g., gr ade, gender, race), indicators of problem behavior theory, social cont agion, precipitants of self-inj ury, and developmental theory were identified in the 2005 Y RBS. Cronbach’s alpha was calculated for item sets that were designed to measure the same behavior or underlying construct (i.e., to be used as a scale). Statistical testing involved univaria te, bivariate, and multivariate analyses and was conducted using the original and transformed scales and results were compared to examine the sensitivity of the results to nonnormality. Multilevel modeling was used because st udents (Level-1) were nested within schools (Level-2). Only Level-1 predictors were used. Models were run with three outcome variables: having ever self-injured (dichotomous), the frequency of self-injury (polytomous), and peer self-injury (dichotom ous). Multinomial logistic regression was conducted with a modified version of the fre quency of self-injury outcome variable. Two models were run, allowing for the following comparisons to be made: once versus never, more than once versus never, and once versus more than once. CHAID analyses using SPSS Answertree v. 3.1 audience segmentation soft ware were used to determine whether there were meaningful segmen ts of youth who self-injure, self-injure frequently, and know a friend who has self-injured.

PAGE 177

164 Summary of Findings A substantial percentage of students survey ed (28.4%) had tried self-injury. This rate is higher than those re ported in most other studies c onducted with adolescents in community settings, with the exception of Lloyd-Richardson et al .’s finding of 46.5%. Laye-Gindhu and Schonert-Reichl (2005) re ported 15%, Muehlenkamp and Gutierrez (2004) reported 16%, and Ross and Heath (2002 ) reported 14% (see Table 1). There are numerous potential reasons for the discre pancy, including possible sample differences between studies, and cohort differences, but the most plausible would seem to be the more inclusive definition used in this st udy, which included pinching. Further research should be conducted with items that differentia te the various forms of self-injury (e.g., cutting, burning, not allowing wounds to heal), such as those used in Muehlenkamp and Gutierrez (2004) and Lloyd-Richardson et al. (2007). The prevalence of having ever tried self -injury did not vary by race or ethnicity, grade, school attended, or age but did diffe r by gender Approximately 32% of females and 25% of males had ever tried self-injury ( p < .01). Whereas the relationship between gender and self-injury was statistically signifi cant, the effect size (i.e., Cramer’s V = .07) was negligible. The difference was not of th e same magnitude as that reported in Ross and Heath (2002), but was more in keeping with that of Muehlenkamp and Gutierrez (2004), suggesting boys must be catching up with girls in usin g self-injury as a maladaptive coping behavior. This finding is consistent with Winter’s (2005) suggestion that increasing rates of self-i njury among males represents eith er an increase in distress and depression among males and/or the influe nce of media exposure to self-injury on

PAGE 178

165 males’ choices of coping behaviors. This finding is inconsistent with feminist interpretations of self-injury as a gendered phenomenon (Shaw, 2002). Youth who had tried self-injury reporte d, on average, poorer health, lower grades, and a greater tendency to stay home from school if they felt unsafe ( p < .01). At the bivariate level, results sugge sted youth who had ever tried self-injury had less positive attitudes toward school (small effect), lo wer levels of parent communication (small effect), and weaker belief in their possibil ities (medium effect). All four behavioral precipitants—frequency of bullying (victim ), having ever been cyberbullied, the frequency of having been cyberbullied, and ha ving been pushed or hit by a girl/boyfriend in the past 12 months—demonstrated weak (small) relationships w ith having ever tried self-injury; two out of the three measures of social contagion, p eer self-injury (small effect), and time spent using the computer or video games for fun (small effect), were associated with having ever tried self-injury. Consistent with problem behavior theory (Jessor & Jessor, 1977), having ever tried se lf-injury was associated with all risk behaviors studied, including su icide (large effect), subs tance use (medium effect), inhalant use (medium effect), deviant behavior (small effect ), having ever had sex (small effect), and abnormal eating behaviors (mediu m effect). Overall, the assumption that self-injury is a White, female, high-achie ving, middle-to-upper middle class issue was challenged (Abrams, 2003; Conterio & Lader, 1998; Ross & Heath, 2002). Many key variables were associated with the three self-injur y outcome variables at the bivariate level; however, these rela tionships disappeared when entered into a multivariate model. This indicates that ma ny of these variables are interrelated and represent a system of variables, rather than isolated entities. Variables that demonstrate

PAGE 179

166 relationships with the outcome variable at the bivariate level but not at the multivariate level may have either an i ndirect relationship with the outcome variable or even a spurious relationship (i .e., dependent on a third variable) with the outcome of interest. The nature of the relationships could be exam ined using structural equation modeling. When controlling for all other variables in the multivariate model including suicide, having ever tried self -injury was associated with pe er self-injury, inhalant use, belief in possibilities, abnormal eating behavior s, and suicide scale scores. Compared to youth who had never tried self-i njury, youth who had tried self-injury were more likely to know a friend who had harmed themselves on pur pose, tried inhalants, have lower belief in their possibilities, have higher levels of abnormal eating behaviors, and have higher suicide scale scores. With the amount of atten tion recently given to the impact of Internet exposure on self-injury (Teens Health, 2005; Wh itlock et al., 2006), it was surprising that the amount of time spent using the computer or video games for fun did not emerge as significant within the multivariate logistic model. This may have been due to the lack of precision of the measure (i.e., not directly aski ng about Internet usage) Further research using a more precise measure of Internet use sh ould be used to explore this relationship. When suicide was excluded from the multivariate model, two additional variables became statistically and practically significan t: gender and having ever been hit or pushed by a girlfriend or boyfriend. Compared to you th who had never tried self-injury, youth who had tried self-injury were more likely to be female and to have been hit or pushed by a boy/girlfriend. This finding is consistent with Laye-Gindhu and Schonert-Reichl’s (2005) supposition that as with gender diffe rences in other expressions of emotional distress (i.e., internalizing be haviors versus externalizing be haviors), there may be gender

PAGE 180

167 differences in self-injurious behaviors and underlying motivations. Support for this argument is found in the average developmental trajectories associated with depression, self-esteem, and anger, all of which are a ssociated with self-injury (Brown, 2001). Depression, low self-esteem, and anger peak during early adolescen ce and the gender gap between males and females is the la rgest (Galambos et al., 2006). It is because of this gender gap that gender differences in key study variables were explored at the bivariate level. In summary, attitudes toward school, belief in possibilities, and parent communi cation did not vary by gender ( p > .01). Males reported, on average, a greater frequenc y of bullying than females ( p < .01, small effect). A greater percentage of females (26%) than ma les (19%) had been cyberbullied ( p < .01); however, this relationship was negligible. Males and females did not differ statistically significantly in the frequency of havi ng been a victim of cyberbullying ( p > .01). Interestingly, however, a greater percentage of females who had b een physically hurt by a boyfriend/girlfriend (56.5%) ha d ever self-injured compared to males who had been physically hurt by a girlfriend/boyfriend (45%). Females (54%) were significantly more likely to know a friend who had harmed themselves compared to males (38%; p < .0001, small effect). Males spent significantly more time, on average, playing video games or using a computer for fun on an average school day than did females (small effect). There was no statistically significant difference between males and females on suicide scale scores ( p > .01). Substance use scores and devi ant behaviors did not differ by gender ( p > .01). Females, on average, reported higher leve ls of abnormal eating behaviors than didi males (small effect). Overall, results sugge sted a mixed picture of gender differences,

PAGE 181

168 with some evidence that males may have beco me more similar to females on suicide risk, and females more similar to males on substance use and deviance. CHAID analyses suggested large groups of youth at risk for (and not at risk) having ever tried self-injury, depending on whether suicide was included in the model. When suicide was included in the model, th e segment at greatest risk for having ever tried self-injury comprised female youth who ha ve moderate to high levels of suicidal tendencies and used substances in the past. This segment is consistent with clinical descriptions of individual w ho self-injure (i.e., White fema les with depression or other diagnoses). When suicide was excluded, the ro le of peer self-injury became apparent: the segment at greatest risk (origi nal variables) comprised youth with low belief in their possibilities and who know a friend who has harmed themselves on purpose. When suicide was excluded and transformed scales were used, the segment at greatest risk comprised youth who have tried inhalants and know a friend who has harmed themselves on purpose. In contrast, the segment at least risk for having ever tried self-injury (suicide included in the model) comprised youth who have not thought about, planned, or attempted suicide, have high belief in their possibilities, and have not used inhalants. When suicide was excluded from the model, the segment at least risk comprised youth with high belief in their possibi lities, who have not used inhalants, and report low levels of bullying (victim). During the past month, most youth had never harmed themselves on purpose. Approximately 15% had harmed themselves one time. Smaller proportions of youth had harmed themselves more frequently, including two or three different times (5%), four or

PAGE 182

169 five different times (2%), and six or more di fferent times (3%). There was a statistically significant and large relationship between havi ng ever tried self-i njury and past month frequency of self-injury. Among youth who self -reported having ever tried self-injury (N = 495), 35% had harmed themselves one time during the past month, 18% had harmed themselves two or three different times, 5.5% had harmed themselves four or five different times, and 11% had harmed themse lves six or more different times. The frequency of self-injury did not vary by gender, race or ethnicity, grade, or school attended. Although Goodman (2005) suggested repetitive self-injury may be more common among females, this study failed to suppor t this assertion. At the bivariate level, the frequency of self-injury was negatively a ssociated with attitude s toward school, belief in possibilities, and parent communication. Whereas all groups (i.e., never, once, and more than once self-injured) differed from one another in terms of their attitudes toward school and belief in possibilities, youth w ho had never tried self-injury reported significantly higher levels of parent communication than di d youth who had self-injured more frequently, but there were not signi ficant differences in parent communication between youth who had self-injured once and youth who had self-injured more than once in the past 30 days. This is consistent with the finding that co mmunication difficulties between parent and youth may place some yout h at risk for self-injury (Derouin & Bravender, 2004). The frequency of self-injury was associated with all four behavioral precipitants. The three groups differed from one another, on average, in terms of bullying frequency (victim) and cyberbullying fr equency (victim), with youth who had self-injured more than once reporting the greatest frequency of both. The frequency of self-injury was

PAGE 183

170 associated with two indicators of social c ontagion—peer self-injur y and time spent using the computer or video games for fun. Y outh who had never self-injured reported substantially lower average use of computer or video games for fun than did either youth who had self-injured once or more than once, which is consistent with emerging research on Internet use and self-injury (Teens Health, 2005; Whitlock et al., 2006). However, there was no statistically si gnificant difference, on average, between youth who had selfinjured once and those who had self-injured more than once, which suggests a threshold effect (i.e., once a certain level of Internet use is reached, a child is at risk). Consistent with problem behavior theo ry (Jessor & Jessor, 1977), the frequency of self-injury was associated with all risk behaviors studied, in cluding abnormal eating behaviors (medium effect), su icide scale scores (medium to large effect), deviant behavior scores (small to medium effect), substance use (medium effect), inhalant use (medium effect), and having ever had sex (s mall effect). All gr oups differed from one another on each risk behavior studied, with youth who had self-injured more than once reporting the highest level of each risk be havior (for continuous variables). Relationships changed substantially between predictors and the frequency of selfinjury, however, when the variables were entered into a multivariate model. The first comparison compared those who had self-inj ured once to those who had never selfinjured in the past 30 days. Three variables we re directly related to the frequency of selfinjury: abnormal eating beha viors, peer self-injury, and suicide. As abnormal eating behaviors increased, the odds of having self-injured once, compared to never, increased. Youth who knew a friend who had harmed themselves on purpose were almost twice as likely to have self-injured once in the past 30 days compared to never. Suicide also was

PAGE 184

171 associated with the odds of having self-injur ed once (compared to never), increasing as suicidal tendencies increased. The second comparison compared those w ho had self-injured more than once to those who had never self-injured in the past 30 days. Three variable s were related to the frequency of self-injury while controlling for all other variables in the model: suicide, having tried inhalants, and belief in possibili ties. As suicidal tendencies increased, the odds of having self-injured twice or more (c ompared to never) increased. Youth who had tried inhalants were more twoand one-half times more likely to have self-injured twice or more in the past 30 days compared to never. Finally, as levels of belief in possibilities increased, the odds of having self-injured twic e or more (compared to never) decreased. The final comparison was made between those who had self-injured more than once and those who had self-injured once in the past 30 days. Only one variable, suicide, significantly distingu ished the two groups. As suicidal tendencies increased, the odds of having self-injured twice or more (compare d to once) increased. Suicidal tendencies were the most important factor in distingui shing between those who try the behavior once in the past 30 days and those who self-injure more frequently. This suggests the presence of two basic groups of youth—youth who may be catching a cultural trend (i.e., those who try the behavior once) and youth who ha ve underlying mental h ealth issues (i.e., those who self-injure more than once). CHAID analyses with and without suicide in the model were used to identify segments at greatest and least risk of frequent self-injury. When suicide was included in the model, the segment at greatest risk of frequent self-injury comprised those who have a moderate to high level of suicidal te ndencies, are non-Hispanic, and have used

PAGE 185

172 inhalants. When suicide was excluded, the segm ent at greatest risk of frequent self-injury comprised youth with moderate to high leve ls of abnormal eating behaviors who know a friend who have harmed themselves on purpose. In comparison, when suicide was included in the model, the segment at least risk of frequent self-injury comprised thos e who have not thought about, planned, or attempted suicide, use the computer or played video game, on average, for less than 1 hour per day, and have not used inhalants. Finally, when suicide was excluded, the segment at least risk comprised those w ith no abnormal eating behaviors, who do not know a friend who have harmed themselves on purpose, and of non-Black race/ethnicity. Results suggested a sizable proportion of youth are alr eady discussing self-injury and are aware of its presence among their pe ers (Fennig et al., 1995). This was not surprising because youth spend more time with their peers than ever before; they are connected 24/7 via cell phone, Internet, telepho ne, and face-to-face c ontact at school and other locations (Roberts et al ., 2005). Almost one-half of students surveyed (46.8%) knew a friend who had harmed themselves on pur pose. At the bivariate level, peer selfinjury was associated with age (small effect), attitudes toward school (small effect), belief in possibilities (small effects), all four precipitants of self -injury (small effects), suicide scale scores (small to medium effects), subs tance use scores and inhalant use (medium effects), having ever had sex (small effect), deviancy scores (small effect), and abnormal eating scores (small effect). However, peer self-injury demonstrated multivariate relationships with gender, grade, and school attended. Knowing a fr iend who had self-injured was more common among females, eighth graders, and students in Schools 2 (~52%) and 5 (~61%). Further,

PAGE 186

173 in the multivariate model, peer self-injury also was associated directly with age at first alcohol use, having ever been cyberbullied, and having ever tried self-injury. Compared to youth who were not exposed to peer sel f-injury, youth exposed were older at first alcohol use, more likely to ha ve ever been cyberbullied, a nd were more likely to have ever tried self-injury ( p < .01). Self-injury is operating via the Inte rnet: results suggest that youth who have tried self-injury and w ho have been cyberbullied may reach out to like others in cyberspace. CHAID analyses revealed the segment at greatest risk of expo sure to peer selfinjury comprised youth with hi gh levels of substance use and who have self-injured at least once during the past 30 days. This ro le of substance use is consistent with McCloskey and Berman’s (2003) finding that alcohol use ma y increase disinhibition and risk taking, setting the stage fo r self-injury. Information is not available to explain why having ever tried self-injur y would place youth at risk fo r peer self-injury, but the literature suggests some possible explanations For example, youth who self-injure may share their injuries with me mbers of their peer groups e xpecting social reinforcement (e.g., attention, sympathy), which may, in part explain the shift be tween experimentation and repetition (Nock & Prinst ein, 2004; Oliver et al., 2005) Some youth may compete with one another (i.e., comparing their inju ries) and overestimate the number of their peers who self-injure. As Dishion and Dodge (2005) expl ained, peer contagion works through competition and false consensus bias (i.e., thinking more peers are performing a behavior than actually are). Conversely, the segment at least risk of exposure to peer self-injury comprised youth with low levels of substance use, in sixth grade, and with no abnormal eating

PAGE 187

174 behaviors. These youth have not had the early adverse experiences or been socialized to gravitate toward risky peer groups (i.e., Goths; Young et al., 2006; see also Hartup, 2005). Strengths & Limitations Quantitative approaches such as those used in this dissertation offer advantages. In addition to anonymity and privacy, the re duction of a complex topic in a careful manner can provide useful information, in term s of empirical evidence, obtained from a large, representative group of individuals. In this study, th e collection of information on a wide range of demographic, attitudinal, a nd behavioral variables, combined with the use of CHAID resulted in the de velopment of typologies of y outh most at risk for selfinjury. These results have important imp lications for prevention and intervention. Although this study had many strengths, ther e were limitations that need to be kept in mind when interpreting the results. One of these limitations stems from the development of the self-injury items. Id eally, youth would have been allowed to conceptualize self-injury, which would have then informed item development. In addition, pretesting was not c onducted, which may have picked up on ambiguities in the items. On the positive side, a preliminary review of the literature was conducted and used to inform item development. Also, professionals well-versed with adolescent mental health informed the item developmen t process. Cognitive interviewing with a small sample of middle school aged youth wa s conducted as a part of the dissertation research to identify possible issues with the items (e.g., problematic words) and to document item validity (i.e., whether items measured what they were intended to measure). In summary, results suggested items represented valid measures of self-injury;

PAGE 188

175 however, the inclusion of pinching may have resulted in the over-inflation of prevalence rates. In addition to servi ng as validity evidence, cogniti ve interviewing results were used to suggest improvements to self-injur y items for future administrations of the YRBS. The current state of the l iterature made it difficult to develop or identify items appropriate for a large scale survey. Most items in the literature are qualitative or openended in nature, and, thus were not suited for la rge-scale survey research. In addition, the need to limit the number of items include d on the YRBS precluded the inclusion of multiple items designed to measure all key as pects of self-injury (e.g., preferred methods, precipitants). For example, items used in th is study were not speci fic enough to enable the determination of types of self-injury. On the other hand, the desire for information was weighed against the desire to do no harm The inclusion of multiple items seeking more in-depth information about the behavi or may have triggered the behavior among vulnerable youth. Finally, the definition provi ded to youth gave examples, which may or may not have tapped into self-injury pr eferences among males (e.g., punching). This may result in higher prevalence estimates within gender. The lack of clear distinction between self-injury and suicide within the self-injury lead in also is a limitation of this study. This lack of distincti on represents a potential source of contamination between the two behaviors. The inclusion of separate sections— one for suicide and one for self-injury—may ha ve helped to distinguish between the two behaviors. However, there remains the po ssibility that self-injury prevalence rates reported may include suicide attempts.

PAGE 189

176 The sampling approach used in this study also poses limitations. Because classrooms and participants were not randomly sampled, the segments identified in this study may not correspond to associated segm ents in the general population (Vriens, 2001). Further, the reader should consider the demographics of the county when attempted to generalize the results of this study since random sampling was not utilized. In addition, students who participated in the survey administration were nested within schools. CHAID does not offer strategies for addressing the multilevel nature of the data. However, HLM 6 multilevel software, which can handle nested data involving categorical outcome variables, was used to conduct the logis tic regressions. Another limitation stemmed from the use of existing or secondary data. The low reliability of some of the scales used in this study represents a limitation as lower reliability makes it more difficult to find relationships. The definition of self-injury used on the YRBS was broad, which limited the ability of this study to focus on specific types of self-injury such as cutting and burning. Th e use of a broad definition resulted in a higher prevalence rate, which included behavi ors such as pinching and scratching that may not be as problematic as other forms of self-harm (e.g., cutting). Further, the definition did not distinguish between repetiti ve self-injury and one -time self-injury. Also, the reliance on existing data limited the ability to ensure all key variables were included in the analysis. Th e absence of these variables along with the correlational design in this study precluded the examinati on of questions of etiology or causality. Also, relationships between self-injury and variables more useful in segmenting youth from an intervention design perspective (e.g., group affiliation), but were not included in the YRBS, could not be addressed. Further, even though theories of social contagion

PAGE 190

177 (e.g., Gladwell, 2000/2002; Marsden, 1998) info rmed this study, items specific to these theories were not available. Only three it ems, which measured whether youth knew of a friend who self-injured and media exposure, we re included that had the potential to tap into this theory. By no means did these item s enable an exhaustive test of this body of literature. Also, a measure of lifetime fre quency of self-injury was not included, which limited the ability to distinguish accurately between youth who had tr ied self-injury once and those who practiced the behavior regularl y. The measure of past month frequency of self-injury made it possible to identify those who had practi ced the behavior recently. Overall, segmentation and logistic regression models were underspecified because of the inability to include all releva nt variables (e.g., self -identification with Goth subculture; Young et al., 2006). This was demonstrated, fo r example, in the classification accuracy rates of the CHAID models. Results suggested the models for having tried self-injury performed well, within the training samples, for example, correctly classifying 78% to 80% of cases. However, the model for peer se lf-injury did not perform as well. Within the training sample, it correctly classified onl y 64% of cases (comparison studies are not available), although this proporti on still exceeds chance. As a result, the findings from this study should be considered preliminary. The use of self-report data using closed-ended questions is also a limitation. This study relied on students’ self-re ports of several risk beha viors—information that is sensitive to some. The following precautions were taken to ensure the validity of students’ self-reports: students were a ssured of the anonymity of the survey administration, identifying information was not collected, and a truthfulness item was

PAGE 191

178 included on the YRBS. Students who reported responding truthfully less than one-half of the time were excluded from the analyses. There also were limitations associated with CHAID. CHAID is a forward stepwise approach; thus, segmentation resu lts depend upon the order in which variables enter the model (The Measurement Group, 1999-2005; Vriens, 2001). Once a predictor has entered the model, it cannot be removed later in the analysis (Vriens, 2001). Fortunately, CHAID trees can be revised ma nually to reflect theoretical or applied knowledge (Vriens, 2001). Investigators can choose to ‘force’ in independent variables at different stages in the tr ee based on non-statistical crit eria (Vriens, 2001). Once a predictor variable is remove d or added to a model, the entire model changes, making CHAID results unstable. Thus, CHAID is most useful for exploring large data sets and model building. Results should be considered suggestive and need to be confirmed using some external criteria (e.g., qua litative research with members of segments identified). Finally, the lack of agreed up stopping rules s hould be addressed with future research. The approach used in this study (i.e., statis tical significance combined with effect size) represents an improvement over standard appr oaches (i.e., statistical significance alone); however, it is not without limitations. For exam ple, statistical and practical effects can occur in nodes following those that do not meet a minimum effect size value. Finally, this study relied on cross-secti onal data. Thus, prevalence estimates represent a one-time snapshot of self-injur y in a community sample of adolescents. Given the lack of baseline information avai lable for early adolescents in the general population and the methodological variation across studies conducted within general populations of adolescents, it was impossible to explain di fferences in prevalence

PAGE 192

179 estimates between this study and others or determine whether self-injury has increased among early adolescents. Finally, analyses usi ng these cross-sectional data were not able to inform issues of directionality and causality. Dissemination & Utilization Dissemination of this study’s results will occur through a brief report and presentation to peers and faculty during the dissertation defense. Study results will be summarized in a brief report that will be made available to Pupil Support Services of the county school board where data were collected. In addition, papers we re presented at the 2007 American Educational Research Asso ciation (AERA) conference and the 2007 American School Health Association Conference Finally, a journal-ready article will be prepared and submitted for possible publication in a professional journal. Journal options include Journal of Youth and Adolescence, Journal of Adolescent Health, and Journal of Counseling & Development. Efforts have been made to reach school administrators and guidance through two presentations, Best Practices in the Sch ool Setting for Children at Risk delivered in the study county. The pr esentations have been delivered to approximately 85 to 100 school counselors nurses, and interested staff. Implications for Prevention This is the first study to empirically ex amine self-injury in relation to multiple risk behaviors within a community sample of early adolescents with the goal of informing school-based prevention efforts. Th e results of this study suggest self-injury serves different functions for different youth. Self-injury operates as an expression of distress among youth with multiple risk factors (e.g., depression, abnormal eating behaviors, substance use) a nd is a “new” expression of a dolescent risk behavior among

PAGE 193

180 youth who may not have diagnosable mental illn ess that is being “labeled as risqu by adults in a particular historical and so ciocultural setting” and becoming “normative” (Rew, 2005, p. 167). A substantial proportion of youth in the general population of early adolescents have tried the behavior and an even larger proportion of youth know friends who have tried the behavior. When shared within a group setting, whethe r a clinical setting (e.g., mental health ward) or community setting (e.g., Goth subcul ture), self-injury may offer group cohesion, acceptance, and understanding (Crouch & Wr ight, 2004; Machoian, 2001; Muehlenkamp, 2005; Young et al., 2006). On campuses where th e prevalence of peer self-injury is high, schools should offer youth alternatives to gaining group cohesion, acceptance, and understanding. Further research should seek to identify character istics of schools that encourage high rates of peer self-injur y (e.g., social dynamics, environmental determinants). Among more recent cohorts, it is assumed that adolescents have been exposed to self-injury via some social venue (e.g., media, school) (A dler & Adler, 2005; Hodgson, 2004). This assumption was tested in this study and was supported. Knowing a friend who had harmed themselves on purpose (i.e., pe er self-injury) was associated with an increased risk of having ever tried self-i njury, possibly by setting the scene for some youth to experiment with self-injury when e xposed within their pe er networks. More than likely, some adolescents who self-injur e (“individual deviants ”) may be surrounded by “fellow deviants” who share their views of self-injury (i.e., the benefits, motivations) (e.g., Goths; Young et al., 2006), which may ma ke it difficult for them to cease the behavior (Adler & Adler, 2005, p. 372). Being surrounded by their “fellow deviants”

PAGE 194

181 confirms the “deviant identity” and makes it difficult for some adolescents to stop selfinjuring and adopt healthier coping behavi ors (Adler & Adler, 2005, p. 372). Although there is a relationship between these two variables, it may not be causal. Rather, it may be caused by some other variable (i.e., third variable). One possible prevention approach is to reposition self-injury as an unacceptable, pathological behavior—not romantic, desirabl e, or positive (Suyemoto, 1998), a behavior that goes against the goal of adolescence (e.g., self-injury is an imitative behavior) (Taiminen et al., 1998; Walsh & Rosen, 1985), and a behavioral choice (Saxe et al., 2002). Repositioning self-injury in such a way may discourage social reinforcement for the behavior (e.g., attention, sympathy), whic h may, in turn, discour age the shift between experimentation and repetition. Providing yout h with materials that coach them on how to deal with a friend who has self-injured and addressing the role of competition and overestimation in spreading the behavior woul d be essential in addr essing self-injury on school campuses. This study informs the growing literature on self-injury among males, suggesting gender differences may be negligible. Males are understudied due to their underrepresentation within clini cal settings (Gratz, 2003; Laye-Gindhu & Schonert-Reichl, 2005). Thus, prevention programming should targ et males as well as females. Further research, however, should seek to identif y differential motivations for self-injury, settings, and expressions of the behavior. Fo r example, females were more likely than males to know a friend who had harmed themse lves on purpose. This may suggest that males are more private about their self-injury than are females.

PAGE 195

182 Interestingly, youth who had never tried self-injury reported that significantly higher levels of parent comm unication than did youth who had self-injured once or more than once. Recall that the communication ite ms were, “My parents have talked to me about their feelings toward me smoking cigarettes” and “My parents have talked to me about their feelings toward me drinking alc ohol.” Conceptualizing self-injury as a new risk behavior would mean needing to educate parents about the need for talking to their child about self-injury. Parents should be info rmed of the current cultural trend, the risks associated with self-injury, and resources av ailable to help youth and families who are dealing with self-injury, associat ed behaviors, and traumas, if relevant. Future research should seek to identify familial influences on the initiation and maintenance of self-injury (e.g., family systems theory). In addressing self-injury, one would need to identify aspe cts of individuals transmitting the self-injury message that make them attractive sources of information. Not having a measure of group affiliation was a limitation of this study. Knowing whether a student self-identified with certa in groups (e.g., Goths, Skaters, Preps) prevalent in middle schools would have allowed for more powerful and informative segmentation strategies. For example, Young et al. (2006) fo und that identification with the Goth subculture was the best predictor of having self-injured or attempted suicide (Young et al., 2006). It would be interesting to know the extent to which the Goth identity overlapped with the at risk segments identified in this study. Further research conducted with early adolescents should incl ude a measure of group identification such as that used in Y oung et al. (2006).

PAGE 196

183 Primary prevention is defined as any type of intervention designed to prevent a behavior or negative outcome before it occurs Primary prevention efforts are geared to general populations. Although childhood sexual abuse was not measured in this study, it should be considered an invisi ble third variable linked to ma ny of the risk behaviors at play, including suicidal tendencies, abnormal eating behaviors, substance use, deviance, and self-injury (Darkness to Light, 2001-2005; Favaro, Ferrara, & Santonastaso, 2007; Gratz, Conrad, & Roemer, 2002; Muehlenkamp & Gutierrez, 2004). Within clinical settings, sexual abuse has been identified as the single best predictor of self-injury, and a recent study conducted among adults supported th e association (Favaro et al., 2007). Approximately 21% of adults report having e xperienced sexual abuse as children (CDC, 1995/1997). One in four girls and one in six boys are sexually abused before the age of 18 (Darkness to Light, 2001-2005). The median age for reported a buse is 9 years of age—if the abuse is reported (Darkness to Light, 2001-2005). Most (80%) initially deny the abuse or tentatively disclose, and, of those who do come forward, most recant (Darkness to Light, 2001-2005). Most children do not disclose sexual abuse even if directly asked (Darkness to Light, 2001-2005). Self-injury, substance use and abuse, deviance, and suicidal thoughts, planning, and attempts offer these youth who have been harmed by the adults in their lives maladaptive ways to cope with th e trauma. Self-injury in particular offers a unique way to communicat e distress, one that seems to operate quite effectively in peer and online settings. A lthough not explored in this study, it would seem one of the most critical means of preventing self-injury would be through the prevention of child sexual abuse through such public health approaches as Stop It Now! ( http://www.stopitnow.com/ ).

PAGE 197

184 Prevention efforts should address current adverse experience in the adolescent’s life, including bullying online and on school campuses, and dating violence. For example, although boys are more likely to experience dating violence, girls who had experienced this were more likely to re port self-injury. Prev ention programming that addresses dating violence coul d also address maladaptive co ping behaviors such as selfinjury. Also, schools should implement ev idence-based bullyi ng prevention programs and make sure that every stude nt is ensured a safe learning environment. Finally, schools and community-based agencies need to partne r together to address cyberbullying. There is a need for further research and development in this area. Belief in possibilities redu ces the risk for self-injur y. Youth who believed they could choose not to use substances even if they were going through difficult times, believed their future held many possibilities, and believed they had better things to do than use substances such as cigarettes or alcoho l, were much less likely to self-injure. On the other hand, youth who had relatively low leve ls of belief in their possibilities were more likely to have tried se lf-injury. Prevention and inte rvention efforts should offer youth who have had adverse experiences (i.e., children at risk) alte rnatives to using substances and self-injury for dealing with pa in and other emotions that stem from these experiences. Efforts to inspire these youth to continue to believe in their possibilities despite what they have faced should be made (i.e., building resiliency). Engaging children at risk in community youth deve lopment activities or other prevention programming such as Teen Theater are possibilities. Substance use, including inhalant us e, plays a role in the initiation and maintenance of self-injury. Although this study was not able to shed light on this role

PAGE 198

185 because of the limitations discussed previousl y, the literature suggests substance use, in and of itself, is a form of self-abuse (Favar o et al., 2007) and may set the stage for selfinjury to occur through the disinhibiti on process (McCloskey & Berman, 2003). Prevention efforts should target all substances ; however, the results of this study suggest that particular attention shoul d be paid to the prevention of inhalant use, particularly when seeking to prevent expe rimentation with self-injur y and increasing frequency of self-injury among those who have already tried the behavior. Secondary prevention, or prevention that occurs among those at risk for performing a behavior or deve loping a disease, could focus on peer prevention. The initial reaction to the behavior is a key time point for intervention—some youth will cut once and move on, whereas others cut once and find it works. Since youth gravitate more toward their peers at this age, they are more likely to disclose their first attempt—if at all—to a close friend. Equipping peers with the right things to say at the right time (i.e., when a peer discloses self-inj ury) to prevent their friends from self-injuring again could prevent some youth from developing a chr onic, maladaptive beha vioral condition. The results of this study suggest self-injury is asso ciated with time spent using the computer for fun (i.e., bivariate results); however, this relati onship is outweighed by many other aspects in the child’s life (Whitlo ck et al., 2006). Further research using more sensitive and comprehensive measur es of Internet usage may find stronger relationships between Internet exposure and se lf-injury and shed light on the nature of this relationship. Given the role of cyber bullying and peer self-injury, it would seem wise to follow segmentation results that sugge sted the most protected youth were those who spent less than one hour per day using th e computer or playing video games for fun.

PAGE 199

186 Schools and parents should be made aware of this recommendation. One logical placement of self-injury prevention information would be in Internet safety training for students and parents. It is important to note the relationship may not be causal; some other variable (e.g., social skills) may account for the relationship be tween self-injury and time spent on the computer. Tertiary prevention, or prevention effort s targeted at those who have already adopted a behavior, should focus on reduci ng the frequency of the behavior while simultaneously increasing the individual’s adaptive coping skills. Results suggested selfinjury, for some youth, is part of a probl em (risk) behavior syndrome that includes substance and inhalant use, deviance, abnorma l eating behaviors, and suicidal tendencies (Jessor, 1991). Jessor (1991) argued that youth who dem onstrate such a syndrome may be in need of interventions that focus at th e lifestyle level rather than at the level of individual problem or risk behaviors. Y outh who tried self-injury exhibited multiple problems and reported poorer health, lower gr ades, and a tendency to stay home from school if they felt unsafe. This is a group in need of attention. Interesting, youth who self-injured in this study di ffered from those described in Fennig et al. (1995). These youth were described as high functioning soci ally and academically but who exhibited internalizing traits (e.g., anxi ety)—not severe emotional di sturbance. Focusing on the early identification of vulnerable youth and teaching/modeling adaptiv e coping skills may be a more effort-, time-, and cost-effective approach than a universal approach (Gladwell, 2000/2002). Yip (2005) advocated for a multid imensional intervention with emphasis on the social environment, including supportive pa rents and peers, teaching youth to handle

PAGE 200

187 frustration and anger and regulate emotions in positive ways, and nurturing youth with the goal of developing thei r self-image and promoti ng their competence. In practice, self-injury shoul d be considered comorbid w ith other risk behaviors. Screening for one behavior should include scr eening for self-injury. For example, if a student exhibited a pattern of visiting the school nurse to be weighed on a frequent basis, the student should be screened for self-injur y. Another example of combining prevention approaches would be includi ng a self-injury component w ith suicide screening and prevention programming. Implications for Further Research Much continues to be learned about self-i njury during early adol escence. Several recommendations for further research were already made and will not be repeated here. One area needing further resear ch is understanding how youth conceptualize and attribute meaning to self-injury. To achieve this understanding, both qualitative (e.g., phenomenological) and mixed methods approaches are needed. In public health research, typically mixed methods designs result in th e best information necessary for designing interventions that will be most responsive to the target audience and, thus, achieve behavior change. In this study, it was not possible to conduct extensive qualitative research with students. To complete the description and devel op an intervention to address self-injury in the st udy county’s schools, further research would need to be conducted with students, staff, and parents. For example, in-dep th interviews with individuals who fell into selected segments could be conducted to gather information needed to design an interven tion (e.g., peer communication). Focus groups or interviews could be conducted with parents to gather information needed to develop a social

PAGE 201

188 marketing campaign targeted at increasing awareness of the behavior and seeking resources for their child if needed. Peer groups could be observe d and both individual and group interviews could be c onducted. Clinical skills, give n the nature of the topic, may be needed when conducting qualitative re search with youth. Supporting youth (i.e., peer research) in conducting research in this area would provide a novel means of learning more about self-injury and culturally appropriate interventions (see Alfonso (2003, 2004) for an overview of working with youth researchers). Finally, to my knowledge this study is the first to investigate empirically the extent of peer self-injury (i .e., the frequency of self-injur y among their friends). Much work remains to be undertaken in this area. Early adolescents are very much aware of each other’s behavior and may encourage one another to adopt and continue a behavior that places them at risk for negative outco mes. Some questions for future research include: Are there some youth who try self -injury during middle school or beyond for attention (“fakes”, “attention whores”; Taiminen et al., 1998 ) and some who self-injure ‘legitimately’ (Crouch & Wright, 2004)? Wh at are youths’ reactions to other youth who self-injure (e.g., social reinforcement, is olation)? Should schools remain quiet (“reluctant”) about the issue a nd isolate those who self-injur e to prevent c ontagion (e.g., Derouin & Bravender, 2004; Lieberman, 2004)? What can schools do to address peer contagion without making it worse?

PAGE 202

189 References Abrams, L. S. (2003). Sociocultura l variations in adolescent gi rls’ expressions of distress: What we know and need to know. Child and Adolescent Social Work Journal, 20 135-150. Adler, P. A., & Adler, P. (2005). Self-injur ers as loners: The social organization of solitary deviance. Deviant Behavior, 26 345-378. Aldenderfer, M. S., & Bl ashfield, R. K. (1984). Cluster analysis Beverly Hills, CA: Sage. Alfonso, M. L. (2003). Involving youth in re search: When should we? How should we? Social Marketing Quarterly, 9 (4), 71-78. Alfonso, M. L. (2004). Involving youth in the formative research process. The Health Education Monograph Series, 21 (1), 52-58. American Psychiatric Association Work Group on Eating Disorders. (2000). Practice guideline for the treatment of patie nts with eating disorders (revision). American Journal of Psychiatry 157, 1-39. Austin, L., & Kortum, J. (2004). Self-injury: Th e secret language of pain for teenagers. Education, 124 517-527. Ayton, A., Rasool, H., & Cottrell, D. ( 2003). Deliberate self-harm in children and adolescents: Association with social deprivation. European Child & Adolescent Psychiatry, 12 303-307.

PAGE 203

190 Baumeister, R. F., & Leary, M. R. (1995). The need to belong: Desire for interpersonal attachments as a fundamental human motivation. Psychological Bulletin, 117 497-526. Becker, H. J. (2000). Who’s wired and w ho’s not: Children’s access to and use of computer technology. Future of Children, 10 44-75. Bongers, I. L., Koot, H. M., van der Ende, J., & Verhulst, F. C. (2003). The normative development of child and adolescent problem behavior. Journal of Abnormal Psychology, 112 179-192. Boyce, P., Oakley-Browne, M. A., & Hatcher, S. (2001). The problem of deliberate selfharm. Current Opinion in Psychiatry, 14 107-111. Bramlett, M. D., & Mosher, W. D. (2002). Cohabitation, marriage, divorce, and remarriage in the United States. Nati onal Center for Health Statistics. Vital and Health Statistics, 23 Briere, J., & Gil, E. (1998). Self-mutilation in clinical and general population samples: Prevalence, correla tes, and functions. American Journal of Orthopsychiatry, 68 620. Brooks-Gunn, J. (1988). Commentary: Developmen tal issues in the transition to early adolescence. In M. R. Gunnar & W. A. Collins (Eds.), Development during the transition to adolescence The Minnesota Symposia on Child Psychology, (Volume 21, pp. 189-208). Hillsdale, NJ: Lawrence Erlbaum Associates. Brown, M. (2001). The behavioral treatment of self-mutilation Self-mutilation, treatment & research symposium. XVI Congress of the World Association for Social Psychiatry, Vancouver B.C., Cana da. Retrieved June 23, 2006, from

PAGE 204

191 http://lists.w3.org/Archives/Publ ic/www-amaya/2001JanMar/att-0235/01DBT.html Carlson, C. I., Cooper, C. R., & Spradling, V. Y. (1991). Developmental implications of shared versus distinct perceptions of the family in early adolescence. In R. L. Paikoff (Ed.), Shared views in the fam ily during adolescence. New Directions for Child Development, no. 51 (pp. 13-32). San Francisco, CA: Jossey-Bass. Carlson, L., DeGeer, S. M., Deur, C., & Fent on, K. (2005). Teachers’ awareness of selfcutting behavior among the adolescent population. PRAXIS, 5 22-29. Centers for Disease Contro l and Prevention. (1995/1997). Adverse childhood experiences study Data and statistics. Retr ieved February 1, 2007, from http://www.cdc.gov/nccdphp/ace/prevalence.htm# Centers for Disease Contro l and Prevention. (1998). Prevention research at CDC Atlanta, GA: Department of Media Relations. Claes, L., Vandereycken, W., & Vertommen, H. (2005). Self-care versus self-harm: Piercing, tattooing, and self-inj uring in eating disorders. European Eating Disorders Review, 13 11-18. Claes, L., & Vandereycken, W. (2007). Self-injurious behavi or: Differential diagnosis and functional differentiation. Comprehensive Psychiatry in press. Clarke, L., & Whittaker, M. (1998). Self-m utilation: Culture, contexts and nursing responses. Journal of Clinical Nursing, 7 129-137. Conterio, K., & Lader, W. (1998). Bodily harm: The breakthrough healing program for self-injurers NY: Hyperion.

PAGE 205

192 Cooper, C. R., & Cooper, R. G. (1992). Li nks between adolescents’ relationships with their parents and peers: Models, eviden ce, and mechanisms. In R. D. Parke & G.W. Ladd (Eds.), Family-peer relationshi ps: Models of linkage (pp. 135-158). Hillsdale, NJ: Lawrence Erlbaum Associates. Crouch, W., & Wright, J. (2004). Deliberate se lf-harm at an adolescent unit: A qualitative investigation. Clinical Child Psychology and Psychiatry, 9 185-204. d’ Acremont, M., & Van der Linden, M. (2005) Adolescent impulsivity: Findings from a community sample. Journal of Youth and Adolescence, 34 427-435. Darkness to Light (2001-2005). Statistics surrounding ch ildhood sexual abuse Retrieved February 1, 2007, from http://www.darkness2light.or g/KnowAbout/statistics_2.asp Derouin, A., & Bravender, T. (2004). Livi ng on the edge: The current phenomenon of self-mutilation in adolescents, American Journal of Ma ternal and Child Nursing, 29 1220. Dieter, P. J., Nicholls, S. S., & Pearlman, L. A. (2000). Self-injur y and self capacities: Assisting an indi vidual in crisis. Journal of Clinical Psychology, 56 1173–1191. Dillon, W. R., & Kumar, A. (1994). Latent structure and other mixture models in marketing: An integrative survey and overview. In R. Bagozzi’s Advanced methods of marketing research (pp. 295– 351). Cambridge, MA: Blackwell Publishers. Dishion, T. J., & Dodge, K. A. (2005). Peer contagion in interventions for children and adolescents: Moving towards an unders tanding of the ecolo gy and dynamics of change. Journal of Abnormal Child Psychology, 33 395-400.

PAGE 206

193 Dishion, T. J., Spracklen, K. M., Andrews, D. W., & Patterson, G. R. (1996). Deviancy training in male adolescent friendships. Behavior Therapy, 27 373-390. Donovan, J. E., & Jessor, R. (1985). Structur e of problem behavior in adolescence and young adulthood. Journal of Consulting and Clinical Psychology, 53 890-940. Dowdy, B. B., & Kliewer, W. (1998). Dati ng, parent-adolescent conf lict, and behavioral autonomy. Journal of Youth and Adolescence, 27 473-492. Elliott, G. R., & Feldman, S. S. (1990). Capturing the adolescent experience. In S. S. Feldman & G. R. Elliott (Eds.), At the threshold: The developing adolescent (pp. 1-13). Cambridge, MA: Harvard University Press. Evans, E., Hawton, K., & Rodham, K. (2005). In what ways are adolescents who engage in self-harm or experience thoughts of self-harm different in terms of helpseeking, communication a nd coping strategies. Journal of Adolescence, 28 573587. Favaro, A.; Ferrara, S., & Santonastaso, P. (2007). Self-injurious behavior in a community sample of young women: Relationship with childhood abuse and other types of self-damaging behaviors. The Journal of clin ical psychiatry, 68 (1), 122-131. Favazza, A. R. (1998). The coming of age of self-mutilation. Journal of Nervous and Mental Disorders, 186 259-268. Favazza, A., & Conterio, K. (1989). Female habitual self-mutilators. Acta Psychiatrica Scandinavica, 79 283-289. Favazza, A., DeRosear, L., & Conterio, K. (1989). Self-mutilation and eating disorders. Suicide and Life-threatening behaviors, 19 352-361.

PAGE 207

194 Fennig, S., Gabrielle, C., & Fennig, S. (1995). Contagious self-mutilation. Journal of the American Academy of Child & Adolescent Psychiatry, 34 402-403. Froeschle, J., & Moyer, M. (2004). Just cu t it out: Legal and ethical challenges in counseling students who self-mutilate. ASCA, 7 231-235. Forthofer, M. S., & Bryant, C. A. (2000). Us ing audience-segmentation techniques to tailor health behavi or change strategies. American Journal of Health Behavior, 24 36-43. Galambos, N. L., Barker, E. T., & Krahn, H. J. (2006). Depression, self-esteem, and anger in emerging adulthood: Seven-year trajectories. Developmental Psychology, 42 350-365. Galley, M. (2003). Student self-h arm: Silent school crisis. Education Week, 23 (1), 14-15. Gilligan, C. (1991). Women's psychologi cal development: Implications for psychotherapy. In C. Gilligan, A. Rogers, & D. Tolman (Eds.), Women, girls, & psychotherapy: Reframing resistance. Binghamton, NY: Harrington Park Press. Gladwell, M. (2000/2002). The tipping point: How little things can make a big difference NY, NY: Little, Brown and Company. Goodman, D. (2005). Youth in child welf are care and self-harming behaviors: Preliminary descriptive findings. OACAS Journal, 49 5-8. Gould, M. S. (2001). Suicide and the media. Annals of the New York Academy of Sciences, 932 200-224. Gratz, K. L. (2003). Risk factors for and func tions of deliberate self -harm: An empirical and conceptual review. Clinical Psychology: Science and Practice, 10 192-205.

PAGE 208

195 Gratz, K. L., Conrad, S. D., & Roemer, L. (2002). Risk factors fo r deliberate self-harm among college students. American Journal of Orthopsychiatry, 72 (1), 128-140. Gross, E. F. (2004). Adolescent Internet us e: What we expect, what teens report. Journal of Applied Developmental Psychology, 25 633-649. Gunnell, D. J., Peters, T. J., Kammerling, R. M., & Brooks, J. (1995). Relation between parasuicide, suicide, psychiatric, admissions, and socioeconomic deprivation. British Medical Journal, 311 226-230. Harrison, D. (1997). Cutting the ties. Feminism & Psychology, 7 438-440. Hartup, W. W. (2005). Peer inte raction: What causes what? Journal of Abnormal Child Psychology, 33 387-394. Hauser, S. T. (1991). Adolescents and their families: Paths of ego development New York, NY: The Free Press. Hawton, K., Arensman, E., Townsend, E., Br emner, S., Feldman, E., Goldney, R., Gunnell, D., Hazell, P., van Heeringen, K., House, A., Owens, D., Sakinofsky, I., & Traskman-Bendz, L. (1998). Deliberat e self-harm: Systematic review of efficacy of psychosocial and pharmacologi cal treatments in preventing repetition. British Medical Journal, 317 441-447. Hawton, K., Fagg, J., Simkin, S., Bale, E., & Bond, A. (1997). Trends in deliberate selfharm in Oxford, 1985-1995. Implications fo r clinical services and the prevention of suicide. The British Journal of Psychiatry, 171 556-560. Hawton, K., Harriss, L., Hall, S., Simkin, S., Bale, E., & Bond, A. (2003). Deliberate self-harm in Oxford, 1900-2000: A time of change in patient characteristics. Psychological Medicine, 33 987-995.

PAGE 209

196 Hawton, K., Harriss, L., & Zahl, D. (2006). Deat hs from all causes in a long-term followup study of 11583 deliberate self-harm patients. Psychological Medicine, 26 397-405. Hawton, K., Rodham, K., Evans, E. & Weathe rall, R. (2002). Deliberate self-harm in adolescents: Self report survey in schools in England. British Medical Journal, 325 1207-1211. Hawton, K., Zahl, D., & Weatherall, R. (2003) Suicide following deliberate self-harm: long-term follow-up of patients who presented to a general hospital. The British Journal of Psychiatry, 182 537-542. Hemme, C. (2001). Farber, Sharon Klayman. When the body is the target: Self-harm, pain, and traumatic at tachments. Book Review. The Journal of Nervous and Mental Disease, 189 646-647. Herpertz, S. (1995). Self-injurious beha vior: Psychopathological and nosological characteristics in subtypes of self-injurers. Acta Psychiatrica Scandinavica, 91, 57-68. Hilt, L. M., Borelli, J. L., Nock, M. K., & Prinstein, M. J. (2004, April). Interpersonal models of adolescent self-mutila tive behavior: A longitudinal study Poster presented at the National Conference on Child Health Psychology, Charleston, SC. Hoare, R. (2004). Using CHAID for classification problems Paper presented at New Zealand Statistical Association Conference, Wellington, New Zealand. Hodgson, S. (2004). Cutting through the silenc e: A sociological c onstruction of selfinjury. Sociological Inquiry, 74 162-179.

PAGE 210

197 Huband, N., & Tantam, D. (2004). Repeated self-wounding: Women’ s recollection of pathways to cutting and of the value of different interventions. Psychology and Psychotherapy: Theory, Re search and Practice, 77 413-428. Jensen, P. S. (2003). Comorbidity and child psychopathology: Recommendations for the next decade. Journal of Abnormal Child Psychology, 31 293-300. Jessor, R. (1991). Risk behavior in a dolescence: A psychosocial framework for understanding and action. Journal of Adolescent Health 12, 597-605. Jessor, R., & Jessor, S. L. (1977). Problem behavior and psyc hosocial development: A longitudinal study of youth NY, NY: Academic Press. John, J., & Miaoulis, G. (1992). A model fo r understanding benefit segmentation in preventive health care. Health Care Management Review, 17 21-32. Johnstone, L. (1997). Self-injury and the psychiatric response. Feminism & Psychology, 7 421-426. Joiner, T. E. (2003). Contagion of suicidal sy mptoms as a function of assortive relating and shared relationship st ress in college roommates. Journal of Adolescence, 26 495-504. Jones, I., & Daniels, B. (1996). An ethological approach to self-injury. The British Journal of Psychiatry, 169 263-267. Kann, L., Kinchen, S. A., Williams, B. I., Ross, J. G., Lowry, R., Hill, C., Grunbaum, J., Blumson, P., Collins, J. L., & Kolbe, L., State and Local YRBSS Coordinators. (1998). Youth risk behavior su rveillance-United States, 1997. MMWR, 47 1-89. Kass, G. (1980). An exploratory technique for investigating la rge quantities of categorical data. Applied Statistics, 29 119-127.

PAGE 211

198 Kilpatrick, D. G., Ruggiero, K. J., Acierno, R., Saunders, B. E., Resnick, H. S., & Best, C.L. (2003). Violence and risk of PTSD, major depression, substance abuse/dependence, and comorbidity: Results from the National Survey of Adolescents. Journal of Consulting and Clinical Psychology, 71 692-700. Kleinman, A. (1988). The illness narratives: Sufferi ng, healing and the human condition NY: Basic Books. Klonsky, E. D. (2004). The functions of deliberate self-harm in college students Unpublished doctoral dissertation, Univer sity of Virginia Charlottesville, Virginia. Klonsky, E. D., Oltmanns, T. F., & Turkheimer, E. (2003). Deliberate self-harm in a nonclinical population: Prevalence and psychological correlates. American Journal of Psychiatry, 160 1501-1508. Laye-Gindhu, A., & Schonert-Reichl, K. A. (2005). Nonsuicidal self-harm among community adolescents: Understanding the “Whats” and “Whys” of self-harm. Journal of Youth and Adolescence, 34 447-457. Lenhart, A., Madden, M., & H itlin, P. (2005, July 27). Teens and technology: Youth are leading the transition to a fully wired and mobile nation Retrieved June 18, 2006, from http://www.pewInternet.org/pdfs/PIP_Teens_Tech_July2005web.pdf Levine, M. (2006). The price of privilege: How parental pressure and material advantage have created a generati on of disconnected and unhappy kids NY, NY: Harper Collins. Lieberman, R. (2004). Understanding and responding to students who self-mutilate. Principal Leadership, 4 10-13.

PAGE 212

199 Liebling, H., Chipchase, H., & Velangi, R. (1997). Why do women harm themselves? Surviving special hospitals. Feminism & Psychology, 7 427-737. Linehan, M. M. (1993). Cognitive-behavioral treatment for borderline personality disorder. NY: Guildford Press. Lloyd-Richardson, E. E., Perrine, N., Dierker, L., & Kelley, M. L. (2007). Characteristics and functions of non-suicidal self-injur y in a community sample of adolescents. Psychological Medicine in press. Machoian, L. (2001). Cutting voices: Self -injury in three adolescent girls. Journal of Psychosocial Nursing and Mental Health Services, 38 22-29. Maczewski, M. (2002). Exploring identities through the Internet: Youth experiences online. Child & Youth Care Forum, 31 111-129. Magidson, J. (1990). CHAID, LOGIT, and log-linear modeling. Danapro Reports on Marketing Information Systems. Delran, NJ: McGraw-Hill Inc. Magidson, J. (1994). The CHAID approach to segmentation modeling: Chi-squared automatic interaction detec tion. In R. Bagozzi’s (Ed.) Advanced methods of marketing research (pp. 119– 159). Cambridge, MA: Blackwell Publishers. Magidson, J., & Vermunt, J. K. (2005). An extension of the CHAID tree-based segmentation algorithm to multiple dependent variables Retrieved February 8, 2006, from www.statisticalinnovations.com Malhotra, N. K. (1989). Segmenting hosp itals for improved management strategy. Journal of Health Care Marketing, 9 45-52.

PAGE 213

200 Malone, K. M., Waternaux, C., Haas, G. L, Cooper, T. B., Li, S., Mann, J. J. (2003). Cigarette smoking, suicidal behavior, and serotonin function in major psychiatric disorders. American Journal of Psychiatry, 160 773-779. Mark, M. M. (2006, June). Principled discovery: Learning from your data, with reason American Evaluation Association and Cent er for Disease Control and Prevention Summer Evaluation Institute 2-Hour Training Workshop. Atlanta, GA. Marsden, P. (1998). Memetics & social cont agion: Two sides of the same coin? The Journal of Memetics: Evolutionary Models of Information Transmission Retrieved June 18, 2006, from http://www.viralculture.c om/pubs/socialcontagion.htm Marsden, P. (2005). Social contagion Retrieved June 18, 2006, from http://www.viralculture.com/contagion.htm McCloskey, M. S., & Berman, M. E. (2003). Alcohol intoxication and self-aggressive behavior. Journal of Abnormal Psychology, 112 306-311. McElroy, A., & Sheppard, G. (1999). The asse ssment and management of self-harming patients in an accident and emergency department: An action research project. Journal of Clinical Nursing, 8 66-72. Milligan, R. J., & Waller, G. (2001). Ange r and impulsivity in non-clinical women. Personality and Individu al differences, 30 1073-1078. Menninger, K. A. (1935). A psychoanalytic study of the significance of self-mutilations. Psychoanalytic Quarterly, 4, 408-466. Muehlenkamp, J. J. (2005). Self-injurious be havior as a separate clinical syndrome. American Journal of Orthopsychiatry, 75 324-333.

PAGE 214

201 Muehlenkamp, J. J., & Gutierrez, P. M. (2004) An investigation of differences between self-injurious behavior and suicide at tempts in a sample of adolescents. Suicide and Life-Threatening Behavior, 34 12-23. Muthn, L. K. & Muthn, B. O. (1998-2006). Mplus user’s guide (4th ed.). Los Angeles: Muthn & Muthn. National Institute of Mental Health. (2001). Teenage brain: A work in progress National Institutes of Health. Retrieved June 16, 2006, from http://www.nimh.nih.gov/ publicat/teenbrain.cfm Narrow, W. E. (1998). One-year prevalence of depressive disorders among adults 18 and over in the U.S.: N IMH ECA prospective data Population estimates based on U.S. Census estimated residential population age 18 and over on July 1, 1998. Unpublished. Nock, M. K., & Prinstein, M. J. (2004). A f unctional approach to the assessment of selfmutilative behavior. Journal of Consulting and Clinical Psychology, 72 885-890. Nock, M. K., & Prinstein, M. J. (2005). Cont extual features and behavioral functions of self-mutilation among adolescents. Journal of Abnormal Psychology, 114 140146. O’Keefe, L. (2002). Media exposure feeding ch ildren’s violent acts, AAP policy states. AAP NEWS: The Official Newsmagazi ne of the American Academy of Pediatrics. Retrieved June 15, 2006, from http://www.aap.org/advocacy/OKeefeMediaViolence.htm

PAGE 215

202 Oliver, C., Hall, S., & Murphy, G. (2005). Th e early development of self-injurious behavior: Evaluating the role of social reinforcement. Journal of Intellectual Disability Research, 49 591-599. Olfson, M., Gameroff, M. J., Marcus, S. C., Greenberg, T., & Shaffer, D. (2005). National trends in hospitalization of youth with intentional self -inflicted injuries. American Journal of Psychiatry, 162 1328-1335. Onacki, M. (2005). Kids who cut: A protocol for public schools. Journal of School Health, 75 400-401. Paikoff, R. L., & Brooks-Gunn, J. (1991). Do parent-child relationships change during puberty? Psychological Bulletin, 110 47-66. Papini, D. R., & Micka, J. C. (1991). Synchr onization in ratings of pubertal maturity and faulty beliefs about family relationships. In R. L. Paikoff (Ed.), Shared views in the family during adolescence. New Directions for Child Development, no. 51 (pp. 33-49). San Francisco, CA: Jossey-Bass. Patton, G. C., Harris, R., Carlin, J. B., Hi bbert, M. E., Coffey, C., Schwartz, M., & Bowes, G. (1997). Adolescent suicidal behaviours: A population-based study of risk. Psychological Medicine, 27 715-724. Pipher, M. (1994). Reviving Ophelia: Saving the selves of adolescent girls NY: Ballantine Books. Powers, S. I., Hauser, S. T., & Kilner, L. A. (1989). Adolescent mental health. American Psychologist, 44 200-208. Prinstein, M .J., & Wang, S. S. (2005). False consensus and adolescent peer contagion: Examining discrepancies between per ceptions and actual re ported levels of

PAGE 216

203 friends’ deviant and health risk behaviors. Journal of Abnormal Child Psychology, 33 293-306. Purington, A., & Whitlock, J. (2004). Self-injury fact sheet ACT for Youth Update Center of Excellence: Research Facts and Findings. Ithaca, NY: Cornell University, Family Life Development Center. Regier, D. A., Narrow, W. E., Rae, D. S., et al. (1993). The de facto mental and addictive disorders service system. Epidemiologic Catchment Area prospective 1-year prevalence rates of disorders and services. Archives of General Psychiatry, 50 85-94. Reis, H. T., & Shaver, P. (1988). Intimacy as an interpersonal process. In S. Duck (Ed.), Handbook of personal relationships: Theory, research, and interventions (pp. 367-389). Chichester, UK: Wiley. Rew, L. (2005). Adolescent health: A multidisciplinary approach to theory, research, and intervention Thousand Oaks, CA: Sage. Roberts, D. F., Foehr, U.G., & Rideout, V. (2005). Generation M: Media in the lives of 8 – 18 year olds Washington, DC: Henry J. Kaiser Family Foundation. Rosen, P. M., & Walsh, B. W. (1989). Patterns of contagion in self-mutilation epidemics. American Journal of Psychiatry, 146, 656–658. Ross, S., & Health, N. L. (2002). A study of the frequency of self-mutilation in a community sample of adolescents. Journal of Youth and Adolescence, 31 67-77. Ross, S., & Heath, N. L. (2003). Two models of adolescent self-mutilation. Suicide and Life-Threatening Behavior, 33 277-287.

PAGE 217

204 Saxe, G.N., Chawla, N., & Van Der Kolk, B. (2002). Self-destructive behavior in patient with dissociative disorders. Suicide and Life-Threatening Behavior, 32 313-320. Shaw, S. N. (2002). Shifting conversations on gi rls’ and women’s self-injury: An analysis of the clinical literature in historical context. Feminism & Psychology, 12 191219. Simmons, R. G., & Blyth, D. A. (1987). Moving into adolescence: The impact of pubertal change and school context New York, NY: Al dine De Gruyter. Simmons, R. G., Burgeson, R., & Reef, M. J. (1988). Cumulative change at entry to adolescence. In M.R. Gunnar & W.A. Collins (Eds.), Development during the transition to adolescence The Minnesota Symposia on Child Psychology, (Volume 21, pp. 123-150). Hillsdale, NJ : Lawrence Erlbaum Associates. Slater, M. D., & Flora, J. A. (1991). Health lifestyles: Audience segmentation analysis for public health interventions. Health Education Quarterly, 18 221-233. Smetana, J. G. (1988). Concepts of self and social convention. Adolescents’ and parents’ reasoning about hypothetical and actual family conflicts. In M. R. Gunnar & W.A. Collins (Eds.), Development during the tr ansition to adolescence The Minnesota Symposia on Child Psychology, (Volume 21, pp. 79-122). Hillsdale, NJ: Lawrence Erlbaum Associates. Smetana, J. G. (1991). Adolescents’ and moth ers’ evaluations of justifications for conflicts. In R.L. Paikoff (Ed.), Shared views in the fam ily during adolescence. New Directions for Child Development, no. 51 (pp. 71-86). San Francisco, CA: Jossey-Bass.

PAGE 218

205 Solano, R., Fernandez-Aranda, F., Aitken, A ., Lpez, C., & Vallejo, J. (2005). Selfinjurious behaviour in pe ople with eating disorders. European Eating Disorders Review, 13 310. Solomon, Y., & Farrand, J. (1996). “Why don’t you do it properly?” Young women who self-injure. Journal of Adolescence, 19 111-119. Spear, L. P. (2000). The adolescent brain a nd age-related behavior al manifestations. Neuroscience and Biobehavioral Reviews, 24 417-463. Stanley, B., Gameroff, M., Michalsen, V., & Ma nn, J. J. (2001). Are suicide attempters who self-mutilate a unique population. American Journal of Psychiatry, 158 427432. Steinberg, L. (1990). Autonomy, conflict, a nd harmony in the family relationship. In S.S. Feldman & G.R. Elliott (Eds.), At the threshold: The developing adolescent (pp. 255-276). Cambridge, MA: Harvard University Press. Strong, M. (1998). A bright red scream: Self-mutilation and the language of pain NY: Penguin Group. Sullivan, H. S. (1953). The interpersonal th eory of psychiatry NY, NY: Norton. Suyemoto, K. L. (1998). The functions of self-mutilation. Clinical Psychology Review, 18 531-554. Suzuki, L. K., & Calzo, J. P. (2004). The s earch for peer advice in cyberspace: An examination of online teen bulletin boards about health and sexuality. Applied Developmental Psychology, 25 685-698.

PAGE 219

206 Taiminen, T. J., Kallio-Soukainen, K., NoksoKoivisto, H., Kalionen, A., & Helenius, H. (1998). Contagion of deliberate se lf-harm among adolescent inpatients. American Academy of Child and Adolescent Psychiatry, 37 211-217. Teens Health. (2005). Cutting Nemours Foundation. Retrieved April 8, 2006, from http://www.kidshealth.org The Measurement Group. (1999-2005). CHAID Retrieved April 26, 2006, from http://www.themeasurementgroup.com/Definitions/CHAID.htm Tiefenbacher, S., Novak, M. A., Lutz, C. K ., & Meyer, J. S. (2005). The physiology and neurochemistry of self-injurious behavior: A nonhuman primate model. Frontiers in Bioscience, 10 1-11. Vriens, M. (2001). Market segmentation: Analytic al developments and application guidelines Technical Overview Series. Mi llward Brown IntelliQuest. Wakschlag, L. S., Pittman, L. D., ChaseLansdale, P. L., & Brooks-Gunn, J. (1996). Mama, I’m a person too: Individuati on and young African-American mothers’ parenting competence In A.M. Cauce & S. Hauser (Eds.), Adolescence and beyond: Family processes and development. Mahway, NJ: Lawrence Erlbaum. Walsh, B. W., & Rosen, P. (1985). Self-mutilation and contagion: An empirical test. American Journal of Psychiatry, 142 119-120. Walsh, B. W., & Rosen, P. (1988). Self-mutilation: Theory, research, and treatment NY: The Guilford Press. West, M., Rose, S., Spreng, S., Sheldon-Ke ller, A., & Adam, K. (1998). Adolescent Attachment Questionnaire: A brief asse ssment of attachment in adolescence. Journal of Youth and Adolescence, 27 661-673.

PAGE 220

207 White, V. E., Trepal-Wollenzier, H., & Nola n, J. M. (2002). College students and selfinjury: Intervention st rategies for counselors. Journal of College Counseling, 5 105-113. Whitlock, J. L., Eckenrode, J., & Silverman, D. (2006). Self-injurious behaviors in a college population. Pediatrics, 117 1939-1948. Whitlock, J. L., Powers, J., & Eckenrode, J. E. (2006). The virtual cutting edge: Adolescent self-injur y and the Internet. Developmental Psychology, 42 1-12. Winters, C. A. (2005). Self-mutilation in a male adolescent inpatient population. Unpublished thesis, Ohio State University. Wocjik, D. (1995). Punk and neo-tribal body art Jackson, MI: University Press of Mississippi. Woldorf, G. M. (2005). Clinical implicati ons of the paradox of deliberate self-injury. JSPN, 10 196-2000. Wright, R. E. (1998). Logistic regression. In L. G. Grimm & P. R. Yarnold’s (Eds.), Reading and understanding multivariate statistics (pp. 217-244). Washington, DC: American Psychological Association. Yankelovich, D., & Meer, D. (2006). Rediscovering market segmentation. Harvard Business Review, February 122-131. Yates, T.M. (2004). The developmental ps ychopathology of self-injurious behavior: Compensatory regulation in posttraumatic adaptation. Clinical Psychology Review, 24 35-74. Yip, K. (2005). A multi-dimensional pers pective of adoles cents’ self-cutting. Child and Adolescent Mental Health, 10 80-86.

PAGE 221

208 Young, R., Sweeting, H., & West, P. (2006). Pr evalence of deliberate self harm and attempted suicide within contemporar y Goth youth subculture: Longitudinal cohort study. British Medical Journal, 332 1058-1061. Zila, L. M., & Kiselica, M. S. (2001). Unde rstanding and counseling self-mutilation in female adolescents and young adults. Journal of Counseling and Development, 79 46-52.

PAGE 222

Appendices

PAGE 223

210 Appendix A 2005 YOUTH RISK BEHAVIOR SURVEY MIDDLE SCHOOL QUESTIONNAIRE This survey is about health behavior. It has been developed so you can tell us what you do that may affect your h ealth. The information you give will be used to develop better health education for young people like yourself. DO NOT write your name on this survey The answers you give will be kept private. No one will know what you write. Answer the questions based on what you really do. Completing the survey is volunt ary. Whether or not you answ er the questions will not affect your grade in this cla ss. If you are not comfortable answering a question, just leave it blank. The questions that ask about your bac kground will be used only to describe the types of students completing this survey. The information will NOT be used to find out your name. No names will ever be reported. Make sure to read every question. Use a #2 pencil only. Fill in the ovals completely. When you are finished, follow the in structions of the person giving you the survey. Thank you very much for your help.

PAGE 224

211 1. How old are you? A. 10 years old or younger B. 11 years old C. 12 years old D. 13 years old E. 14 years old F. 15 years old G. 16 years old or older 2. What is your sex? A. Female B. Male 3. In what grade are you? A. 6th grade B. 7th grade C. 8th grade D. Other 4. How do you describe yourself? A. American Indian or Alaska Native B. Asian C. Black or African American D. Hispanic or Latino E. Native Hawaiian or Other Pacific Islander F. White 5. What school do you go to? A. Middle School 1 B. Middle School 2 C. Middle School 3 D. Middle School 4 E. Middle School 5 F. Middle School 6 G. Middle School 7 H. Middle School 8 I. None of the above

PAGE 225

212 6. What school do you go to? A. Other School 1 B. Other School 2 C. Other School 3 D. Other School 4 E. Other School 5 F. Other School 6 G. None of the above 7. How do you describe your health in general? A. Excellent B. Very good C. Good D. Fair E. Poor The next 8 questions ask about personal safety and violence-related behaviors. 8. How often do you wear a seat belt when riding a car? A. Never B. Rarely C. Sometimes D. Most of the time E. Always 9. When you ride a bicycle, how often do you wear a helmet? A. I do not ride a bicycle B. Never wear a helmet C. Rarely wear a helmet D. Sometimes wear a helmet E. Most of the time wear a helmet F. Always wear a helmet 10. When you rollerblade or ride a sk ateboard, how often do you wear a helmet? A. I do not rollerblade or ride a skateboard B. Never wear a helmet C. Rarely wear a helmet D. Sometimes wear a helmet E. Most of the time wear a helmet F. Always wear a helmet

PAGE 226

213 11. Have you ever ridden in a car dr iven by someone who had been drinking alcohol? A. Yes B. No C. Not sure 12. During the past 30 days, have you ever carried a weapon, such as a gun, knife, or club to school? A. Yes B. No 13. During the past 30 days, have you ev er been in a physical fight at school? A. Yes B. No 14. Have you ever been in a physical fight at school in which you were hurt and had to be treated by a doctor or nurse? A. Yes B. No 15. During the past 12 months, did your boyfri end or girlfriend ever hit, slap, or physically hurt you on purpose? A. Yes B. No The next 12 questions ask about bullying at school during the past 30 days. Definition of Bullying : Bullying is anything from teas ing, saying mean things, writing mean notes, or leaving someone out of th e group, to physical atta cks (hitting, pushing, kicking) where one person or a group of people picks on another person over and over again. Kids who are bullied have a hard time defending themselves. 16. During the past 30 days, how many tim es did another student tease or call you names? A. Never B. 1 or 2 times C. 3 to 5 times D. 6 to 9 times E. 10 or more times

PAGE 227

214 17. During the past 30 days, how many times did another student threaten to hit or hurt you? A. Never B. 1 or 2 times C. 3 to 5 times D. 6 to 9 times E. 10 or more times 18. During the past 30 days, how many times di d another student spread rumors about you? A. 0 times B. 1 or 2 times C. 3 to 5 times D. 6 to 9 times E. 10 or more times 19. During the past 30 days, how many times did other students not let you join in what they were doing? A. 0 times B. 1 or 2 times C. 3 to 5 times D. 6 to 9 times E. 10 or more times 20. During the past 30 days, how many times did another student push, shove, slap, hit, or kick you on purpose? A. 0 times B. 1 or 2 times C. 3 to 5 times D. 6 to 9 times E. 10 or more times 21. During the past 30 days, how many tim es did you tease or call another student names? A. 0 times B. 1 or 2 times C. 3 to 5 times D. 6 to 9 times E. 10 or more times

PAGE 228

215 22. During the past 30 days, how many times did you threaten to hit or hurt another student? A. 0 times B. 1 or 2 times C. 3 to 5 times D. 6 to 9 times E. 10 or more times 23. During the past 30 days, how many times did you spread rumors about another student? A. 0 times B. 1 or 2 times C. 3 to 5 times D. 6 to 9 times E. 10 or more times 24. During the past 30 days, how many times did you keep another student from joining in what you were doing? A. 0 days B. 1 or 2 days C. 3 to 5 days D. 6 to 9 days E. 10 or more times 25. During the past 30 days, how many times di d you push, shove, slap, hit, or kick another student on purpose? A. 0 times B. 1 or 2 times C. 3 to 5 times D. 6 to 9 times E. 10 or more times 26. Have you been taught about not bullying at school? A. Yes B. No C. Not sure

PAGE 229

216 27. During the past 30 days, how many days did you not go to school because you felt you would be unsafe at school or on your way home from school? A. Never B. 1 day C. 2 or 3 days D. 4 or 5 days E. 6 or more days The next 4 questions are about "cyberbullying". Cyberbullying is "using the Inte rnet or cell phone to send or pos t harmful or cruel text or images to bully others." Examples of cybe rbullying include sending cruel or threatening messages, creating websites that ridicule othe rs, posting pictures of classmates online and asking students to rate them, morphing photos, taking a picture of a person in a locker room or bathroom using a digital phone camera and sending to others, or engaging someone in instant messaging (IM) to trick them into revealing sensitive information for the purpose of sending on to others. 28. During your lifetime, have you ever been cyberbullied? A. Yes B. No 29. Have you ever cyberbullied someone else? A. Yes B. No 30. During the past 30 days, how many times were you the victim of cyberbullying? A. 0 times B. 1 or 2 times C. 3 to 5 times D. 6 to 9 times E. 10 or more times 31. During the past 30 days, how many tim es did you cyberbully someone else? A. 0 times B. 1 or 2 times C. 3 to 5 times D. 6 to 9 times E. 10 or more times

PAGE 230

217 The next 3 questions ask about attempted suicide. Sometimes people feel so depressed about the future that they may consider attempting suicide or killing themselves. 32. Have you ever seriously thought about killing yourself? A. Yes B. No 33. Have you ever made a plan about how you would kill yourself? A. Yes B. No 34. Have you ever tried to kill yourself? A. Yes B. No The next 3 questions ask about self-harm (c utting, scratching, bu rning, not allowing wounds to heal, pinching). Sometimes pe ople who feel upset hurt themselves on purpose as a way to feel better (less upset). 35. Have you ever hurt yourself on purpos e (cutting, scratching, burning, not allowing wounds to heal, pinching)? A. Yes B. No 36. During the past month, how often ha ve you hurt yourself on purpose (cutting, scratching, burning, not allowi ng wounds to heal, pinching)? A. Never B. 1 time C. 2 or 3 different times D. 4 or 5 different times E. 6 or more different times 37. Have any of your friends hurt them selves on purpose (cutting, scratching, burning, not allowing wounds to heal, pinching)? A. Yes B. No

PAGE 231

218 The next 10 questions ask about tobacco use. 38. Have you ever tried cigarette smoking, even one or two puffs? A. Yes B. No 39. How old were you when you smoked a whole cigarette for the first time? A. I have never smoked a whole cigarette B. 8 years old or younger C. 9 years old D. 10 years old E. 11 years old F. 12 years old G. 13 years old H. 14 years old or older 40. During the past 30 days, have you smoked cigarettes even one or two puffs? A. Yes B. No 41. During the past 30 days, on how many days did you smoke cigarettes? A. 0 days B. 1 or 2 days C. 3 to 5 days D. 6 to 9 days E. 10 to 19 days F. 20 to 29 days G. All 30 days 42. During the past 30 days, on the days you smoked, how many cigarettes did you smoke per day? A. I did not smoke cigarettes during the past 30 days B. Less than 1 cigarette per day C. 1 cigarette per day D. 2 to 5 cigarettes per day E. 6 to 10 cigarettes per day F. 11 to 20 cigarettes per day G. More than 20 cigarettes per day

PAGE 232

219 43. During the past 30 days, how did you usually get your own cigarettes? (Select only one response) A. I did not smoke cigarettes during the past 30 days B. I bought them in a store, such as a conveni ence store, super market, or gas station C. I bought them from a vending machine D. I gave someone else money to buy them for me E. I borrowed (or bummed) them from someone else F. A person 18 years or older gave them to me G. I took them from a store or family member H. I got them some other way 44. When you bought or tried to buy cigarettes in a store during the past 30 days, were you ever asked to show proof of age? A. I did not try to buy cigarettes in a store during the past 30 days B. Yes, I was asked to show proof of age C. No, I was not asked to show proof of age 45. Have you ever smoked cigarettes daily, th at is, at least one ci garette every day for 30 days? A. Yes B. No 46. During the past 30 days, on how many days did you use chewing tobacco or snuff, such as Redman, Levi Garrett Beechnut, Skoal Bandits, or Copenhagen? A. 0 days B. 1 or 2 days C. 3 to 5 days D. 6 to 9 days E. 10 to 19 days F. 20 to 29 days G. All 30 days 47. During the past 30 days, on how many days did you smoke cigars, cigarillos, or little cigars? A. 0 days B. 1 or 2 days C. 3 to 5 days D. 6 to 9 days E. 10 to 19 days

PAGE 233

220 F. 20 to 29 days G. All 30 days The next 6 questions ask about drinking alco hol. This includes drinking beer, wine, wine coolers, and liquor such as rum, gin, vodka, or whiskey. For these questions, drinking alcohol does not include drinking a few sips of wine for religious purposes. 48. Have you ever had a drink of alcohol, other than a few sips? A. Yes B. No 49. How old were you when you had your first drink of alcohol othe r than a few sips? A. I have never had a drink of al cohol other than a few sips B. 8 years old or younger C. 9 years old D. 10 years old E. 11 years old F. 12 years old G. 13 years old H. 14 years old or older 50. In the past 30 days have you had any alcohol to drink, other than a few sips? A. Yes B. No 51. In the last year, have you had five or more drinks of alcohol in one day? A. Yes B. No 52. During the past 30 days, how many times ha ve you had 5 or more drinks in one day? A. 0 days B. 1 to 2 days C. 3 to 5 days D. 6 to 9 days E. 10 to19 days F. 20 to 29 days G. All 30 days

PAGE 234

221 53. During the past 30 days, how did you get alcohol? A. I did not drink alcohol during the past 30 days. B. I bought alcohol in a store such as a ga s station, super market, or convenience store. C. I took alcohol from my house. D. I had a person 21 years or older buy alcohol for me. E. I had a stranger buy alcohol for me. F. I was with a group that was drinking alcohol. The next 4 questions ask about marijuana us e. Marijuana also is called grass or pot. 54. Have you ever used marijuana? A. Yes B. No 55. During the past 30 days, how often have you used marijuana? A. 0 days B. 1 to 2 days C. 3 to 5 days D. 6 to 9 days E. 10 to 19 days F. 20 to 29 days G. All 30 days 56. How old were you when you tried marijuana for the first time? A. I have never tried marijuana B. 8 years old or younger C. 9 years old D. 10 years old E. 11 years old F. 12 years old G. 13 years old H. 14 years old 57. During the past 30 days how did you get marijuana? A. I did not use marijuana in the past 30 days. B. I took marijuana from my house. C. I was with a group that was using marijuana. D. I bought it at school.

PAGE 235

222 E. I bought it outside of school. The next 4 questions ask about other drug use. 58. Have you ever used any form of cocaine, including powder, crack, or freebase? A. Yes B. No 59. Have you ever sniffed glue, or breathed the contents of spray cans, or inhaled any paints or sprays to get high? A. Yes B. No 60. Have you ever used prescription drugs or over the counter medicine (cough/cold medicine) to get high? A. Yes B. No 61. Have you ever used a needle to inject any illegal drug into your body? A. Yes B. No The next 7 questions ask about body weight. 62. How do you describe your weight? A. Very underweight B. Slightly underweight C. About the right weight D. Slightly overweight E. Very overweight 63. Which of the following are you trying to do about your weight? A. Lose weight B. Gain weight C. Stay the same weight D. I am not trying to do anything about my weight 64. Have you ever exercised to lose weight or to k eep from gaining weight?

PAGE 236

223 A. Yes B. No 65. Have you ever eaten less food, fewer calories, or foods low in fat to lose weight or to keep from gaining weight? A. Yes B. No 66. Have you ever gone without eating for 24 hours or more (also called fasting) to lose weight or to k eep from gaining weight? A. Yes B. No 67. Have you ever taken any diet pills, powders, or liquids without a doctor’s advise to lose weight or to keep from gaining weight? (Do not include meal replacement products such as Slim Fast.) A. Yes B. No 68. Have you ever vomited or taken laxatives to lose weight or to keep from gaining weight? A. Yes B. No The next 9 questions ask about physical activity. 69. On how many of the past 7 days did you exercise or participate in physical activity for at least 20 minutes that made you sweat and breathe hard, such as basketball, soccer, running, swimming laps, fast bicycling, fast dancing, or similar aerobic activities? A. 0 days B. 1 day C. 2 days D. 3 days E. 4 days F. 5 days G. 6 days H. 7 days

PAGE 237

224 70. On an average school day, how many hours do you watch TV? A. I do not watch TV on an average school day B. Less than 1 hour per day C. 1 hour per day D. 2 hours per day E. 3 hours per day F. 4 hours per day G. 5 or more hours per day 71. Do you play on any sports teams? (Include any teams r un by your school or community groups.) A. Yes B. No 72. In an average week when you are in school, on how many days do you go to physical education (PE) classes? A. 0 days B. 1 day C. 2 days D. 3 days E. 4 days F. 5 days 73. In the last 2 months, did you try a ne w game or sport (rock climbing, roller blading, or other fun thing) that you've never done before? A. Yes B. No 74. Have you ever seen, read, or h eard any messages or ads about VERB? A. Yes B. No 75. Have you ever seen, read, or heard any messages or ads about VERB Summer Scorecard ? A. Yes B. No

PAGE 238

225 76. Think about an average week during th is school year. How many days of the week do you do a physical activity or play a sport, NOT including PE? A. 0 days B. 1 day C. 2 days D. 3 days E. 4 days F. 5 days G. 6 days H. 7 days 77. If I did physical activitie s on most days it would be fun. A. Really Agree B. Sort of Agree C. Sort of Disagree D. Really Disagree The next question asks about AIDS education. 78. Have you ever been taught about AIDS or HIV infection in school? A. Yes B. No C. Not sure The next 4 questions ask about sexual intercourse. 79. Have you ever had sexual intercourse? A. Yes B. No 80. How old were you when you had sexual intercourse for the first time? A. I have never had sexual intercourse B. 8 years old or younger C. 9 years old D. 10 years old E. 11 years old F. 12 years old G. 13 years old H. 14 years old or older

PAGE 239

226 81. With how many people have you ever had sexual intercourse? A. I have never had sexual intercourse B. 1 person C. 2 people D. 3 or more people 82. The last time you had sexual intercourse, di d you or your partner use a condom? A. I have never had sexual intercourse B. Yes C. No The next 2 questions are about health-related behaviors. 83. How often do you wear sunscreen or sun block when you are outside for more than an hour? A. Never B. Rarely C. Sometimes D. Most of the time E. Always 84. On an average school day, how many hours do you spend playing video games or using a computer for fun? (Include activ ities such as Nintendo, Game Boy, Play Station, and computer games.) A. I do not play video games or use a computer for fun B. Less than 1 hour C. 1 hour D. 2 hours E. 3 hours F. 4 hours G. 5 hours H. 6 or more hours The next 4 questions are about delinquent behaviors. 85. Since school started this year how many times have you skipped school? A. Never B. 1 time

PAGE 240

227 C. 2 times D. 3 times E. More than 3 times 86. During the past 12 months, how often have you shoplifted (stolen something from a store)? A. 0 times B. 1 time C. 2 or 3 times D. 4 or 5 times E. 6 or more times 87. During the past 12 months, have you been a member of a gang? (A group of people who identify themselves with the same symbol, color, and/or name and participate in criminal activity.) A. Yes B. No 88. Do you think you will be involved in a gang in the future? A. Yes B. No The next question asks about the Believe in All Your Possibilities campaign. 89. Have you ever heard, seen, or read anything about the Believe in All Your Possibilities campaign (BELIEVE)? A. Yes B. No C. Not sure The next 2 questions ask about SOURCE Teen Theatre performances. High school students from SOURCE Teen Theatre have performed plays about underage smoking and drinking (“End of Summer”), bullying (“Surviving Lunch”), and other topics (“Read My Lips”). 90. Have you seen a SOURCE Teen Theatre performance? A. Yes B. No C. Not sure

PAGE 241

228 91. Was the SOURCE Teen Theatre play talked about in your classroom either before or after the performance? A. I have not seen a SOURCE Teen Theatre play B. Yes, we talked about the play. C. No, we did not talk about the play. D. Not sure The next question asks about the Welcome Everybody or Where Everybody Belongs (WEB) program. 92. Have you participated in WEB activities such as the 6th grade back to school assembly? A. Yes B. No C. Not sure The next two questions ask about your parents. 93. My parents have talked to me about thei r feelings toward me smoking cigarettes. A. Yes B. No C. Not sure 94. My parents have talked to me about thei r feelings toward me drinking alcohol. A. Yes B. No C. Not sure The next several questions ask about your f eelings about your future, substance use, and your family. 95. I believe my future holds many possibilities. A. Strongly Agree B. Agree C. Neither Agree nor Disagree D. Disagree E. Strongly Disagree 96. I believe I have better things to do th an smoke cigarettes or drink alcohol.

PAGE 242

229 A. Strongly Agree B. Agree C. Neither Agree nor Disagree D. Disagree E. Strongly Disagree 97. My parents stick by what they believe is best for me even if I disagree. A. Strongly Agree B. Agree C. Neither Agree nor Disagree D. Disagree E. Strongly Disagree 98. I believe I can choose to not smoke cigarettes or drink alcohol, even if I’m going through tough times. A. Strongly Agree B. Agree C. Neither Agree nor Disagree D. Disagree E. Strongly Disagree 99. There is at least one teac her or adult at this school I can talk w ith if I have a problem. A. Strongly Agree B. Agree C. Neither Agree nor Disagree D. Disagree E. Strongly Disagree

PAGE 243

230 100. People at my school notice when I am good at something. A. Strongly Agree B. Agree C. Neither Agree nor Disagree D. Disagree E. Strongly Disagree 101. I participate in activities (clubs, sp orts, WEB, etc.) at this school. A. Strongly Agree B. Agree C. Neither Agree nor Disagree D. Disagree E. Strongly Disagree 102. How would you describe the grades you usually get on school assignments? A. Mostly A’s B. Mostly A’s and B’s C. Mostly B’s D. Mostly B’s and C’s E. Mostly C’s F. Mostly C’s and D’s G. Mostly D’s H. Mostly D’s and F’s I. Mostly F’s The next questions ask about your answers on this survey. 103. In general, how often did you tell the truth in answering the questions on this survey? A. All of the time B. Most of the time C. About half of the time D. Less than half the time E. None of the time 104. I read this survey carefully A. All of the time B. Most of the time

PAGE 244

231 C. About half of the time D. Less than half the time E. None of the time Thank you very much for your help!

PAGE 245

232 Appendix B Exploratory Factor Analysis Results Variable Promax Factor Loadings 1 2 3 4 Thought about suicide -0.037 0.901 -0.020 0.068 Planned suicide -0.040 0.913 -0.001 0.064 Tried suicide 0.081 0.822 -0.096 0.117 Carried weapon to school 0.063 0.124 -0.069 0.759 Fight at school -0.015 00044 0.010 0.616 Hurt in a fight 0.006 0.184 0.046 0.383 Hit or pushed by girl/boyfriend 0.134 0.165 0.091 0.231 Ever been cyberbullied -0.092 -0.071 1.059 0.010 Ever tried cigarettes 0.835 0.077 0.099 -0.074 Smoked cigarettes in past 30 days 0.907 0.205 0.039 -0.257 Ever tried alcohol 0.735 -0.098 0.057 0.163 Frequency of 5 or more drinks in one da y in past 30 days 0.743 -0.094 -0.063 0.306 Drank alcohol in past 30 days 0.674 -0.059 0.019 0.298 Ever had five or more drinks of alcohol 0.748 -0.142 -0.079 0.247 Ever tried marijuana 0.919 -0.065 -0.037 -0.006 Used marijuana in the past 30 days 0.814 0.020 -0.095 0.154 Ever used inhalants 0.267 0.192 0.083 0.371 Ever used OTC or prescription medications to get high 0.641 0.074 0.012 0.233 Ever had sex 0.516 -0.009 -0.051 0.306 Frequency of skipping sc hool 0.278 -0.015 0.009 0.383 Frequency of shoplifting 0.457 0.015 0.009 0.383 Frequency of cigarette smoking during past 30 days 0.890 0.209 -0.017 -0.091 TV viewing hours 0.042 -0.034 -0.045 0.088 Video game and computer use fo r fun – hours -0.173 -0.075 0.195 0.329 Peer self-injury 0.267 0.249 0.242 -0.108 Frequency of having been the victim of cyberbullying 0.055 0.008 0.789 -0.062 Inter-Factor Correlations Factor 1 2 3 4 1 1.000 2 0.522 1.000 3 0.420 0.436 1.000 4 0.618 0.431 0.347 1.000

PAGE 246

233 Appendix C Relationships among Predictor Variables 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 1 1.00 2 -.13 1.00 3 .16 -.05 1.00 4 -.22 .44 -.13 1.00 5 .18 -.07 .16 -.13 1.00 6 .30 -.21 .16 -.34 .14 1.00 7 .16 -.09 .11 -.13 .22 .19 1.00 8 .22 -.11 .23 -.24 .10 .29 .11 1. 00 9 .28 -.16 .13 -.28 .23 .23 .19 .14 1.00 10 .29 -.19 .15 -.32 .16 35 .19 .28 .30 1.00 11 .18 -.08 .11 -.14 .21 .16 .50 .11 14 .11 1.00 12 .27 -.16 .14 -.28 .24 .23 .15 .19 65 .28 .16 1.00 13 .12 -.02 -.10 -.02 -.06 -.02 .07 -.09 .07 .01 .03 .04 1.00 14 .08 -.09 .01 -.13 .02 .17 .10 .19 .01 .07 .08 .01 .01 1.00 15 -.12 .23 -.05 .32 -.11 -.22 -.09 -.16 -.16 -.17 -.11 -.17 09 -.08 1.00 16 .23 -.10 .08 -.22 .19 .23 22 .10 .25 .21 .14 .22 .16 .22 -.14 1.00 17 -.10 .17 -.03 .21 -.02 -.11 -. 04 -.03 -.13 -.12 -.06 -.12 -.04 04 .14 -.04 1.00 18 .00 -.07 -.02 .01 .01 -.01 06 -.09 .00 -.01 .03 -.03 .02 -. 01 .08 .01 .09 1.00 19 .01 -.03 -.01 .01 -.08 -.04 .03 -.07 -.04 .02 .01 -.04 .00 -.12 .05 .00 -.00 .24 1.00 20 .31 -.18 .19 -.40 .17 .41 .19 .43 .26 .40 .19 .25 -.06 .29 -. 25 .30 -.10 -.02 -.01 1.00 21 .35 -.19 .17 -.30 .26 .26 .19 .23 .48 .34 .17 .39 .06 .11 -. 18 .26 -.11 -.00 -.03 .32 1.00 22 .02 -.02 .00 -.07 .07 .04 -.03 02 .03 .03 -.04 .05 .11 .06 -.11 .06 -.05 -.14 -.11 .03 .01 1.00 23 .02 -.08 .05 -.05 .13 .09 .15 06 .10 .06 .09 .11 -.19 .04 -.06 05 -.02 -.07 -.07 .06 .07 .26 1.00

PAGE 247

234 Appendix C Continued Variable Key # in Correlation Matrix Variable Description Middle School YRBS Item # 1 Abnormal eating behavior scale 66 – 68 2 Attitudes toward school scale 99 – 101 3 Been physically hurt by girlfriend or boyfriend 15 4 Belief in possibilities scale 95, 96, 98 5 Bully scale 16 – 20 6 Deviant Behaviors 85 – 86 7 Ever been cyberbullied 28 8 Ever had sexual intercourse 79 9 Ever harmed themselves on purpose 35 10 Ever tried inhalants 59 11 Frequency of being a victim of cyberbullying 30 12 Frequency of self-injury during past 30 days 36 13 Gender 2 14 Grade level 3 15 Grades – self-reported academic performance 102 16 Knowledge of peer self-injury 37 17 Parent communication scale 93 – 94 18 Race or ethnicity 4 19 School 5 20 Substance Use 38, 40 – 42, 48, 50 – 52, 54, 55, 60 21 Suicide 32 – 34 22 TV viewing – amount per school day 70 23 Video/computer use – amount per school day 84

PAGE 248

235 Appendix D Summary of Bivariate and Multivariate Results Self-injury Frequency of SI Peer Self-injury Predictor Bivariate Multivariate Bivari ate Multivariate Bivariate Multivariate Female S+ NS NS NS S+ S+ Hit by boy/girlfriend S+ NS S+ NS S+ NS Cyberbullied S+ NS S+ NS S+ S+ Tried self-injury NA NA NA NA S+ S+ Peer self-injury S+ S+ S+ S+ S+ NS Inhalant use S+ S+ M+ M+ S+ NS TV viewing time NS NS NS NS NS NS Sex (ever had) S+ NS S+ NS S+ NS Video/computer use S+ NS S+ NS S+ NS Grades SNS SNS SNS Grade level NS S+ NS SS+ S+ Attitudes toward school SNS S_ NS SNS Belief in possibilities MSMSSNS Parent communication SNS SNS SNS Bully (victim) frequency S+ NS S+ S+ S+ NS Abnormal eating behaviors M+ M+ M+ M+ S+ NS Substance use M+ NS M+ NS M+ S+ Suicide L+ M+ M+ S+ & M+ S+ NS Deviant behavior S+ NS S+ NS S+ NS Black NS NS NS NS NS NS Hispanic NS NS NS NS NS NS Other ethnicity NS NS NS NS NS NS NS = non-statistically significant S = small M = moderate/medium L = large + = positive = negative

PAGE 249

About the Author Moya has spent her professional caree r conducting community-based research and evaluation. She is known for her ability to help community-based institutions such as schools identify prevention needs and effective strategi es and conduct evaluations. She is a staff member of the Methods a nd Evaluation Unit of the Florida Prevention Research Center at the University of S outh Florida (USF). Currently, she provides evaluation support for three phys ical activity social mark eting campaigns building on CDC’s national VERBTM campaign for “tweens” and Believe in All Your Possibilities, a community-based alcohol and t obacco prevention program. Moya Lynn Alfonso has an interd isciplinary background in psychology, anthropology, public health, a nd education. She obtained he r master’s of science in public health from the USF, College of Public Health. She received her doctoral degree from USF in Educational Measurement and Research. Moya plans to continue her research related to adolescent and family health and teach at the university level.