USF Libraries
USF Digital Collections

The contribution of the neighborhood context to social disparities in access to health care among sexually experienced a...

MISSING IMAGE

Material Information

Title:
The contribution of the neighborhood context to social disparities in access to health care among sexually experienced adolescent females
Physical Description:
Book
Language:
English
Creator:
Nearns, Jodi
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla
Publication Date:

Subjects

Subjects / Keywords:
Disparities
Racial segregation
Poverty
Neighborhoods
Health care
Adolescent health
Multilevel modeling
Dissertations, Academic -- Public Health -- Doctoral -- USF
Genre:
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Abstract:
ABSTRACT: Access to health care is an important resource for sexually experienced adolescent females in the prevention of unintended pregnancy and sexually transmitted infections, including HIV. However, a paucity of research exists regarding the extent to which social disparities in access to health care exist among this vulnerable population of adolescents, including the potential contribution of the neighborhood context. Therefore, the primary aims of this dissertation were to examine (1) the extent to which racial and socioeconomic disparities in access to health care exist among sexually experienced adolescent females, (2) the extent to which access to health care among sexually experienced adolescent females varies across neighborhoods, and (3) the extent to which the neighborhood racial and socioeconomic context contribute to racial and socioeconomic disparities in access to health care among sexually experienced adolescent females.A multilevel design was employed for this di ssertation utilizing secondary data from Wave I of the National Longitudinal Study of Adolescent Health (Add Health). Analyses included hierarchical generalized linear modeling to examine the receipt of a routine physical, the receipt of contraceptive services, and reported unmet health needsamong the dissertation sample of 1,526 sexually experienced Non-Hispanic Black and Non-Hispanic White adolescent females between 15 years to 19 years of age who were dispersed across 546 neighborhoods. After adjusting for a variety of factors that may influence access to health care, the findings revealed no racial disparities and few socioeconomic disparities in access to health care among this sample of adolescents. No significant relationship was noted between the neighborhood racial and socioeconomic context and access to health care or social disparities in access to health care among this sample of adolescents. However, the findings revealed that access to health care among this sample of sex ually experienced adolescent females varied across neighborhoods, above and beyond the individual composition of the neighborhood. Further studies are indicated to explore the underlying factors that contribute to socioeconomic disparities in access to health care among sexually experienced adolescent females, and the potential neighborhood characteristics that may contribute to differential access to health care across neighborhoods among this vulnerable population of adolescents.
Thesis:
Dissertation (Ph.D.)--University of South Florida, 2006.
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 Jodi Nearns.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 216 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 - 001801024
oclc - 162131774
usfldc doi - E14-SFE0001645
usfldc handle - e14.1645
System ID:
SFS0025963:00001


This item is only available as the following downloads:


Full Text

PAGE 1

The Contribution of the Neighborhood Context to Social Disparities in Access to Health Care among Sexually Experien ced Adolescent Females by Jodi Nearns A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Community and Family Health College of Public Health University of South Florida Major Professor: Melinda S. Forthofer, Ph.D. Co-Major Professor: Julie A. Baldwin, Ph.D. Ellen M. Daley, Ph.D. Ann L. Abbott, J.D., Ph.D. Jason W. Beckstead, Ph.D. Date of Approval: May 4, 2006 Keywords: disparities, racial segregation, poverty, neighborhoods health care, adolescent health, multilevel modeling Copyright 2006, Jodi Nearns

PAGE 2

Acknowledgements There are many people who have guided a nd supported me through this journey. First, and foremost, thank you Dr. Melinda Fort hofer, my advisor, mentor, and friend, for your continual support, guidance, and met hodological expertise you bestowed to me throughout my doctoral program, and this dissert ation. I am eternally grateful to you for shaping me into the researcher and scholar th at I am today, and for fueling my passion in the field of social epidemiology. Dr. Julie Baldwin, thank you for your willi ngness to serve as co-chair of my dissertation committee, your continual positiv e energy and support, and your guidance in the field of health disparities. Dr. Ellen Daley, thank you for expert a dvice on issues related to adolescent reproductive health, as well as your guidance in helping me see the big picture of this dissertation. I truly enjo yed our discussions. Dr. Jason Beckstead, thank you for your expe rt statistical assi stance during this dissertation, and your ability to engage me in thinking critically about the methodological, statistical, and concep tual aspects of my dissertation. Dr. Ann Abbott, thank you for guidance rela ted to the health policy implications of this dissertation, and for your insi ght into social justice issues. On a more personal note, thank you Gino, my husband, for encouraging me to pursue my goals, as well as reminding me to stop and smell the roses from time to time. A special thank you to my mom, Judie, and my sister, Dana, words cannot express my gratitude for all the emotional and tangibl e support you have unselfishly given me throughout my doctoral studies, this dissert ation, and not to mention, my life.

PAGE 3

Thank you, Bethany, for all your help and support throughout my doctoral program, this dissertation, and in my life. You are a true and forever friend. Thank you, as well, to my “inner circle” of friends, for all your support and for remaining my friend even when you didn’t see me for months on end. And fina lly, to my dad, Dan, although you have not been with me physically for nearly 10 years now, you are always with me in spirit and in my heart. Thank you for all your support and encouragement throughout my life.

PAGE 4

i Table of Contents List of Tables................................................................................................................. .. iv List of Figures................................................................................................................ .. vii Abstract....................................................................................................................... .....viii Chapter One: Introduction...................................................................................................1 Statement of the Problem.........................................................................................1 Purpose of the Study................................................................................................2 Rationale for the Study............................................................................................3 Overview of the Study Design................................................................................ 5 Data Sources............................................................................................................9 Public Health Significance.....................................................................................10 Hypotheses.............................................................................................................12 Delimitations..........................................................................................................16 Limitations ...........................................................................................................17 Definitions..............................................................................................................19 Chapter Two: Literature Review.......................................................................................23 Introduction............................................................................................................23 Reproductive Health of Adolescent Females ........................................................23 Social Disparities in the Reproduc tive Health of Adolescent Females.................25 Neighborhood Context and Social Di sparities in Reproductive Health................27 Access to Health Care: An Institutional Resource.................................................29 Social Disparities in Adoles cents’ Access to Health Care.....................................33 Theoretical Frameworks........................................................................................33 Social Stratification....................................................................................33 The Political Economy of Health Care......................................................35 The Neighborhood Context and Acce ss to Institutional Resources.......................37 The Neighborhood Context in the U.S......................................................36 The Neighborhood Context and Access to Health Care............................39 Conclusions............................................................................................................46 Chapter Three: Methodology.............................................................................................51 Purpose of the Study .............................................................................................51 Hypotheses.............................................................................................................53 Study Design .........................................................................................................57 Add Health Study Design and Sampling Procedures ...........................................58

PAGE 5

ii Study Sample.........................................................................................................62 Power ....................................................................................................................62 Measures................................................................................................................65 Dependent Variables..................................................................................65 Access to Health Care....................................................................65 Independent Variables...............................................................................66 Adolescents’ Sociodemographic Characteristics...........................66 Adolescent Control Variables........................................................69 Neighborhood Racial and Socioeconomic Context ......................77 Neighborhood Control Variables ..................................................79 Data Analysis.........................................................................................................79 Data Management......................................................................................80 Univariate and Bivariate Analyses.............................................................82 Multivariate Analyses................................................................................82 Study Limitations...................................................................................................86 Chapter Four: Results........................................................................................................89 Study Sample.........................................................................................................89 Univariate Analyses...............................................................................................92 Adolescents’ Access to Health Care .........................................................92 Characteristics of Adoles cents in the Study Sample..................................93 Distribution of Adolescents’ Socioeconomic Position by Race................95 Characteristics of the Ne ighborhoods in the Study Sample.......................96 Bivariate Analyses.................................................................................................98 Adolescents’ Characteristics and Access to Health Care..........................99 Neighborhood Characteristics and Access to Health Care......................104 Social Disparities in Adolescents’ Health................................................106 Multivariate Analyses..........................................................................................108 Hypothesis I.............................................................................................109 Receipt of a Receipt Physical......................................................110 Unmet Health Needs....................................................................115 Receipt of Contraceptive Services...............................................119 Hypothesis II............................................................................................124 Receipt of a Receipt Physical......................................................125 Unmet Health Needs....................................................................131 Receipt of Contraceptive Services...............................................136 Hypothesis III...........................................................................................141 Receipt of a Receipt Physical......................................................142 Unmet Health Needs....................................................................149 Receipt of Contraceptive Services...............................................155 Summary of Overall Findings..............................................................................161 Social Disparities in Access to Health Care............................................161 The Neighborhood Context and Access to Health Care..........................164 The Neighborhood Context and Social Disparities in Access to Health Care ...................................................................................165

PAGE 6

iii Post-Hoc Analyses: Associations be tween Type of Health Care Place and Service Receipt .......................................................................................165 Univariate Analyses.................................................................................166 Type of Health Care Place for Adolescents’ Receipt of a Routine Physical......................................................................166 Bivariate Analyses...................................................................................167 Associations between Adoles cents’ Characteristics and the Type of Health Care Place for their Receipt of a Routine Physical......................................................................167 Associations between Adolescents’ Receipt of Contraceptive Services and the Type of Health Care Place for their Receipt of a Routine Physical..........................170 Chapter Five: Summary of Findings................................................................................171 Social Disparities in Access to Health Care........................................................171 The Neighborhood Context and Access to Health Care......................................177 Limitations of the Study.......................................................................................179 Implications for the Field.....................................................................................182 Directions for Future Research............................................................................184 Conclusions..........................................................................................................186 References..................................................................................................................... ...188 Appendices .................................................................................................................... ..208 Appendix A: Measures........................................................................................209 Appendix B: Summary of Model Building Process............................................214 About the Author...................................................................................................End Page

PAGE 7

iv List of Tables Table 1 Impact of Inclusion Cr iteria on Size of Study Sample..............................89 Table 2 Descriptive Analyses of Study Sample Prior to and After Exclusion of Participants Missing Data.....................................................91 Table 3 Descriptive Statistics of Adolescents’ Access to Health Care...................92 Table 4 Descriptive Statistics of Adolescents’ Characteristics...............................95 Table 5 Distribution of Adol escents’ Socioeconomic Position by Adolescent Race.........................................................................................96 Table 6 Descriptive Statistics of the Charac teristics of Adolescents ’ Neighborhoods ..........................................................................................98 Table 7 Associations between Adolescen ts’ Sociodemographic Characteristics and A ccess to Health Care..............................................101 Table 8 Associations between Adolescents ’ Predisposing Factors, Enabling Factors, and Health Needs and Their Access to Health Care..............................................................................................104 Table 9 Associations between the Character istics of Adolescents’ Neighborhoods and Access to Health Care.............................................106 Table 10 Associations between A dolescents’ Sociodemographic Characteristics and Adolescents’ Health.................................................108 Table 11 HGLM Analyses: Social Dispar ities in the Receipt of a Routine Physical....................................................................................................114 Table 12 HGLM Analyses: Social Disparities in Unmet Health Needs ................118 Table 13 HGLM Analyses: Social Disparities in the Receipt of Contraceptive Services ............................................................................123

PAGE 8

v Table 14 HGLM Analyses: Contribution of the Neighborhood Context to the Receipt of a Routine Physical..........................................130 Table 15 HGLM Analyses: Contribution of the Neighborhood Context to Unmet Health Needs..............................................................135 Table 16 HGLM Analyses: Contribution of the Neighborhood Context to the Receipt of Contraceptive Services ..................................140 Table 17 HGLM Analyses: Contribu tion of the Proportion of Black Residents in the Neighborhood to Social Disparities in the Receipt of a Routine Physical .................................................................144 Table 18 HGLM Analyses: Contribu tion of the Proportion of Low Educated Residents in th e Neighborhood to Social Disparities in the Receipt of a Routine Physical .....................................146 Table 19 HGLM Analyses: Contribu tion of the Proportion of Poor Residents in the Neighborhood to Social Disparities in the Receipt of a Routine Physical .................................................................148 Table 20 HGLM Analyses: Contributi on of the Proportion of Black Residents in the Neighborhood to Social Disparities in Unmet Health Needs ..............................................................................150 Table 21 HGLM Analyses: Contribu tion of the Proportion of Low Educated Residents in th e Neighborhood to Social Disparities in Unmet Health Needs..........................................................152 Table 22 HGLM Analyses: Contribu tion of the Proportion of Poor Residents in the Neighborhood to Social Disparities in Unmet Health Needs................................................................................154 Table 23 HGLM Analyses: Contribu tion of the Proportion of Black Residents in the Neighborhood to Social Disparities in the Receipt of Contraceptive Services ..........................................................156 Table 24 HGLM Analyses: Contribu tion of the Proportion of Low Educated Residents in th e Neighborhood to Social Disparities in the Receipt of Contraceptive Services...............................158 Table 25 HGLM Analyses: Contributi on of the Proportion of Poor Residents in the Neighborhood to Social Disparities in the Receipt of Contraceptive Services...........................................................160

PAGE 9

vi Table 26 Descriptive Statistics of the Type of Health Care Place Where Adolescents Received Their Routing Physical........................................167 Table 27 Associations be tween Adolescents’ Sociodemographic Characteristics and the Type of Health Care Place for Their Receipt of a Routine Physical........................................................169 Table 28 Associations between Adoles cents’ Receipt of Contraceptive Services and the Type of Health Care Place for Their Receipt of a Routine Physical..................................................................170

PAGE 10

vii List of Figures Figure 1. Conceptual Model......................................................................................52 Figure 2. Statistical Power of Sample.......................................................................65

PAGE 11

viii The Contribution of the Neighborhood Context to Social Disparities in Access to Health Care among Sexually Experi enced Adolescent Females Jodi Nearns ABSTRACT Access to health care is an important resource for sexually experienced adolescent females in the prevention of unintended pre gnancy and sexually transmitted infections, including HIV. However, a paucity of research exists regarding the extent to which social disparities in access to health care exist am ong this vulnerable population of adolescents, including the potential contri bution of the neighborhood contex t. Therefore, the primary aims of this dissertation were to exam ine (1) the extent to which racial and socioeconomic disparities in access to heal th care exist among sexually experienced adolescent females, (2) the extent to which access to health care among sexually experienced adolescent females varies acro ss neighborhoods, and (3) the extent to which the neighborhood racial and socioeconomi c context contribute to racial and socioeconomic disparities in ac cess to health care among sexually experienced adolescent females. A multilevel design was employed for this dissertation utilizing secondary data from Wave I of the Nationa l Longitudinal Study of Adoles cent Health (Add Health). Analyses included hierarchical generalized li near modeling to examine the receipt of a routine physical, the r eceipt of contraceptive services, and reported unmet health needs

PAGE 12

ix among the dissertation sample of 1,526 sexua lly experienced Non-Hispanic Black and Non-Hispanic White adolescent females betw een 15 years to 19 years of age who were dispersed across 546 neighborhoods. After adjust ing for a variety of factors that may influence access to health care, the findings revealed no racial disparities and few socioeconomic disparities in ac cess to health care among this sample of adolescents. No significant relationship was noted between the neighborhood racial and socioeconomic context and access to health care or social di sparities in access to health care among this sample of adolescents. However, the findings revealed that access to health care among this sample of sexually experienced a dolescent females vari ed across neighborhoods, above and beyond the individual composition of the neighborhood. Further studies are indicated to explore the underlyi ng factors that contribute to socioeconomic disparities in access to health care among sexually experien ced adolescent females, and the potential neighborhood characteristics that may contri bute to differential access to health care across neighborhoods among this vulner able population of adolescents.

PAGE 13

1 Chapter One: Introduction Statement of the Problem According to the 2002 National Survey of Family Growth (NSFG), approximately 46% of adolescent females 15 years to 19 year s of age reported they engaged in sexual intercourse at least once in their lifeti me (Abma, Martinez, Mosher, & Dawson, 2004). Furthermore, 8% reported they engaged in in tercourse with four or more partners over their lifetime, 46% reported they did not use a condom with their last intercourse, and nearly 10% reported that thei r first sexual encounter was i nvoluntary (Abma et al., 2004). These statistics are alarming as they increase adolescents’ risk for unintended pregnancy and sexually transmitted infections (STI’s), including human immunodeficiency virus (HIV). Unfortunately in the year 2000 al one, 230,000 adolescent females 15-19 years of age were diagnosed with chlamydia, over 68,000 were diagnosed with gonorrhea (Centers for Disease Control & Prev ention, CDC, 2004) and over 800,000 became pregnant, of which the majority were unint ended (Abma et al., 2004; Henshaw, 2004). Health care services play an important role in preventing unintended pregnancy and STI’s among sexually active adolescents when services include safer sex education, the provision of contraceptives, and the ea rly detection and treatment of reproductive health problems (Brindis, 2002). Recent studies indicate that adolescents who receive health care services are more knowledgeab le about safer sexual behaviors, more

PAGE 14

1 consistent contraceptive users, and more lik ely to use contraceptives that offer greater protection against unintended pregnancy (K irby, 2002; Boekeloo, Schamus, Simmens, Cheng, O’Connor, & D’Angelo, 1999; Danielson, Marcy, Plunkett, Wiese, & Greenlick, 1990; Orr, Langefeld, Katz, & Caine, 1996; Scher, 2004; Winter & Breckenmaker, 1991). However, numerous studies have found th at many adolescents have difficulty accessing health care services, particularly minority and low-income adolescents who are less likely to receive health care services and more likely to experience unmet health needs (Bartman et al., 1997; Elster, Jarosi k, VanGesst, & Fleming, 2003; Ford, Bearman, and Moody, 1999; Lieu, Newacheck & Mc Manus, 1993; Newacheck, Hung, Park, Brindis, & Irwin, 2003; Simp son et al., 2005; Shenkma n, Youngblade, & Nackashi, 2003; Stevens & Shi, 2003). These social di sparities in access to health care are particularly unjust as sexually active adol escent females who belong to disadvantaged population groups also experience dispropor tionately higher rates of unintended pregnancy and sexually transmitted infections (Abma et al., 2004; CDC, 2004; Grunbaum et al. 2004; Henshaw, 1998). A burgeoning field of research has focused on issues surrounding social disparities in health care examining the exte nt to which they exist as well as their potential contributing factors. Although these studies have been very influential in advancing our knowledge, attention has been di rected primarily to factors occurring at the level of the individual, provider, and h ealth care system with very few studies addressing the neighborhood context (Aday, Begley, Lairson, & Balkrishnan, 2004; Kirby& Kaneda, 2005; Morrison, Anderse n, & Aday, 1998; Shi & Stevens, 2005).

PAGE 15

2 Neighborhoods serve as important outlets for th e distribution of health care resources and therefore they may play an important role in shaping social disparitie s in health care. The few studies examining this relationship support this argument as neighborhood poverty and racial segregation have been associated with fewer acute and preventive health care visits and greater unmet health needs am ong the individuals residing within these neighborhoods (Andersen et al., 2002; Brooks-Gunn, McCormick, Klebanov, & McCarton, 1998; Kirby & Kaneda, 2005) as well as with fewe r health care providers and health care facilities locate d within these neighborhoods (Guagliardo, Ronzio, Cheung, Chacko & Joseph, 2004; Komaromy, et al., 1996). Healthy People 2010 declared the promoti on of responsible sexual behavior and increased screening for STI’s as priorities fo r improving the health of adolescents within the U.S. (U.S. Department of Health & Human Services, USDHHS, 2000). However, in order to achieve these goals, adolescents mu st be assured equita ble access to quality health care services. Purpose of the Study The overall aim of this study is to explore the extent to which social disparities in access to health care services exist among se xually experienced adolescent females and the potential contribution of the neighborhood racial and socioeconomic composition in shaping these disparities. Th e neighborhood is defined as a geographic unit and measured as the census tract of residence. The specific objectives of this study are: (1) to examine the extent to which access to preventive health care services varies among sexually experienced adolescent females based upon their individual ra ce and socioeconomic position, (2) to examine the extent to which access to preventive health care services

PAGE 16

3 varies between neighborhoods among all sexua lly experienced adolescent females and the extent to which the socioeconomic and racial composition of the neighborhood may account for this variation, and (3) to ex amine the extent to which racial and socioeconomic disparities in access to prev entive health care serv ices varies between neighborhoods as a function of the ne ighborhood’s racial and socioeconomic composition. Rationale for the Study Social disparities in adolescents’ reproduc tive health are pervasive throughout the United States as numerous studies have indi cated that sexually active adolescent females who are poor or low-income are more lik ely to experience an STI (CDC, 2004), HIV (CDC, 2005), and early childbearing th an those adolescents who are more socioeconomically advantaged (Alan Guttm acher Institute, AGI, 1999; Pamuk, Makuc, Heck, Reubehaven, & Lochner, 1998). Minority adolescents are also more likely than White adolescents to experience these negati ve sequelae (Abma et al., 2004; CDC, 2004; Grunbaum et al. 2004), which experts attri bute to the higher ra tes of poverty and widespread discrimination that these social groups experi ence (Geronimus, 2003; Kirby et al., 2001). Consequently, it is particularly alarming that these socially disadvantaged adolescents, who are at greater risk for uni ntended pregnancy and STI’s, also experience differential access to health care services th at are known to promote health and wellbeing. Ethically, we must examine the extent to whic h health care disparities exist, elucidating the factors that either contribu te to or preclude their existence so that the most effective interventions can be developed to am eliorate their existence.

PAGE 17

4 Although research examining social disp arities in access to health care is burgeoning, few studies have examined access issues among highly vulnerable populations, such as sexually experienced adolescents (Elster et al., 2003; Kirby & Kaneda, 2005; Morrison et al., 1998). Furthe rmore, a paucity of research exists examining the contribution of the neighborhood cont ext in shaping health care disparities. The main purpose of this study is to addre ss these gaps in knowl edge by exploring the extent to which the socioeconomic and r acial composition of the neighborhood moderates the relationship between individual race and socioeconomic position and access to preventive health care services among sexua lly experienced adolescent females. This specific population was selected for this st udy due to the numerous behavioral, biological and social factors that increase the risk for adolescent females to experience an unintended pregnancy and STI, including HIV, in addition to the perv asiveness of social disparities in reproductive health among sexua lly active adolescent females (Abma et al., 2004; Abma, Driscoll, & Moore, 1998; Grunba um et al., 2004; Manlove, Terry-Humen, Papillo, Franzetta, Williams, & Ryan, 2001; Miller, Monson, & Norton, 1995; Raj, Silverman, & Amaro, 2000; Sarigiani, Ryan, & Petersen, 1999; Shafii & Burstein, 2004; USDHHS, 2000). Methodologically, this study is designed to address several weakness in the current literature related to neighborhood and social disparities research. For example, many studies examining the impact of neighborhoods on health or on access to health care have been based on ecological or cont extual designs. Although these studies have informed our understanding of neighborhoods and social dispari ties, their primary weakness is that the data were analyzed on a si ngle level of analysis at either the group or

PAGE 18

5 individual level. In contrast, this study employs a multilevel design, which enables researchers to examine data across multiple levels of analysis and make inferences regarding the influe nce of individual and group level characteris tics on the outcome of interest (Diez-Roux, 2003). Finally, few studies examini ng social disparities have focused on the interplay between race and socioeconomic position and ho w such interrelationships may influence health outcomes or access to institutional resources. At the individual and neighborhood level, race and socioeconomic position are inte rtwined and further research is necessary to explore their interrelati onships (Acevedo-Garcia & Lochner, 2003). This study is designed to address this limitation by expl oring the extent to which racial and socioeconomic disparities in access to health care exist among sexually active adolescent females and how this relationship may vary based upon the racial and socioeconomic structure of the neighborhood in which they live. Overview of the Study Design This study employs a nonexperimental, multilevel design via secondary data obtained from the National Longitudinal Study of Adolescent Health (Add Health). The study design is nonexperimental as the randomi zation of adolescents into advantaged or disadvantaged neighborhoods through a community trial is financially and ethically not feasible (Neuman, 2003). Although several quasi -experimental studies are underway in which families who already live within an impoverished neighborhood were randomly assigned to neighborhoods of differing socioeconomic oppor tunities (Leventhal & Brooks-Gunn, 2000), a study of this nature is beyond the scope of this study.

PAGE 19

6 The multilevel design of the study enables the examination of neighborhood and individual characteristics on adolescents’ access to health care, includ ing the direct effect of each, as well as their in teractions (Diez-Roux, 2003; Hox, 1995; Subramanian, Jones, & Duncan, 2003). The multilevel analyses will also correct for th e violation of the assumption of independent observations, which is inherent to clus ter-based research designs (Diez-Roux, 2003; Hox, 1995; Subramanian et al., 2003). The sampling frame consists of those Non-Hispanic White and Non-Hispanic Black sexually active females between 15 to 19 years of age who par ticipated in Wave I of the Add Health In-Home Interview. The following inclusion criteria were imposed on the sampling frame for this study: (1) never ma rried, (2) lived in curr ent residence greater than one year, (3) only one randomly sampled sibling, and (4) complete data on the variables of interest related to this study. This sample subset contains 7 pairs of siblings who met the eligibility criter ia for the study. However, one adolescent from the pair was randomly selected via SAS random sampling procedures for inclusion in the study analyses. The analysis is lim ited to only one sibling memb er to avoid violating the assumption of independent observations and the necessity for a 3-Level model if the sibling pairs were an alyzed together. Several issues regarding the conceptualiz ation and measurement of key constructs in this study warrant further discussion. First, this study examines the impact of race, rather than race and ethnicity, due to wi despread discrimination based on phenotype. Historically and presently, Bl acks are the most residentially segregated population group in the United States. Asian and Hispanic popul ations are also moderately segregated, but they tend to become more integrated as their socioeconomic position improves unless

PAGE 20

7 they are of Black racial origin (Charles 2003; Massey & Denton, 1989). Further research is necessary to examine race and ethnicity t ogether, such as the experiences of Black Hispanics, Black Non-Hispanics, White Hispan ics, White Non-Hispan ics and so forth. At the individual level, racial cate gories are based on adolescents’ selfidentification as White or Black. At th e neighborhood level, racial composition is measured as the proportion of Blacks residing with in a census tract as restrictions in the Add Health contextual databa se prohibit the construction of segregation indices. Racial composition is frequently used as a proxy for racial segregation th roughout the literature, but this measure fails to fully describe th e severity and scope of racial segregation (Acevedo-Garcia & Lochner, 2003; Aceve do-Garcia, Lochner, Osypuk, & Subramanian, 2003). Future multilevel studies will be needed to examine the extent to which access to health care varies at both the individual and neighborhood level for other racial and ethnic categorizations. Individual socioeconomic position is m easured via two indicators: (1) household income to poverty ratio, which is the ra tio of the total 1994 a nnual household income before taxes, adjusted for household size, to the 1994 federal pove rty threshold, divided into five categories — 0-100%, 101-200%, 201-300%, 301-400% and 401% or greater (reference group), and (2) level of parental education categorized as less than high school, high school degree or GED, some colle ge or technical trai ning but no degree, and college degree or greater (reference group). Neighborhood socioeconomic composition is measured via two continuous indicators: (1 ) proportion of residents in the census tract who are below the 1990 U.S. Census poverty thre shold and (2) the proportion of residents in the census tract who have less than a high school degree.

PAGE 21

8 The neighborhood is defined as a geogra phic unit and measured as the census tract of residence. Census tracts co mmonly serve as proxies for neighborhoods throughout the literature (Leventhal & Brooks -Gunn, 2000) and they are also considered “administrative units used by federal, st ate, and local governme nts, including public health departments, to characterize jurisd ictions, determine eligibility for diverse programs, and allocate resources” (Subram anian, Chen, Rehkopf, Waterman, & Krieger, 2005, p. 260). For the purpose of this study, access to hea lth care is measured by two indicators: (1) “realized access” or the receipt of health care services, and (2) unmet health needs, in which the adolescent reports that health care was needed, but not obt ained (Aday et al., 2004). Conceptually, access to health care is a multidimensional construct defined by the Institute of Medicine (IOM, 1993, p. 4) as “the timely use of personal health services to achieve the best possible outcomes” and is of ten referred to in the literature as an individual’s receipt of health care (Aday et al., 2004; US DHHS, 2003), the barriers they face when trying to obtain services (Ada y et al., 2004, USDHHS, 2003), and the quality of care received (Aday et al., 2004; IO M, 2003; USDHHS, 2003). Consequently, researchers have operationalized access to health care with a variety of measures and currently no recommendations exist as to which indicator is superior. The individual and neighborhood level cont rol variables were selected based on findings from the literature as well as con ceptual frameworks rela ted to access to and utilization of health care among vulnera ble populations and vulnerable communities (Aday, 2001; Aday et al., 2004; Andersen 1995; Davidson, Andersen, Wyn, & Brown, 2004). In particular, guiding constructs include d those “predisposing factors” that may

PAGE 22

9 influence the propensity to use services (hea lth beliefs, age), “enabli ng factors” that may facilitate access to and use of services (transportation, in surance) and “health needs” (Aday, 2001; Aday et al., 2004; Andersen 1995; Davidson et al., 2004; Stevens & Shi, 2005). Data Sources This study utilizes data from the Add H ealth study for secondary data analysis. Add Health is a school-base d longitudinal study with a cl uster based sampling design. Data from multiple contexts are available that enable the exploration of how the social structure and social context influence adoles cent health outcomes (H arris, Florey, Tabor, Bearman, Jones, & Udry, 2003). This study utilizes four data sources from the restricted data set of the Add Health study available through the UNC Carolina Popul ation Center – Wave I Adolescent InHome Interview, Wave I Pa rent In-Home Questionnaire, Wave I School Administrator Data, and Wave I Contextual Data. Add Health researchers collected the adolescent, parent, and school administrator data be tween September 1994 and December 1995. The contextual data are derived from a variety of administrative sources, including the Census of Population and Housing, 1990: Summary Tape File 3A (STF 3A) and are linked to the adolescent and parent data via participan t identification numbers created by the Add Health researchers for each adolescent. The variables from the Adolescent In-Home Intervie w examined in this study include: individual sociodemographics (race: White or Black) and the individual control variables of length of time lived in nei ghborhood of permanent residence health status (history of STI, pregna ncy, self-rated health), potential transportation resources

PAGE 23

10 (adolescent has a driver’s license), attitudes (susceptibility for STI/pregnancy), perceptions (perception of parental disappr oval regarding contraception) and beliefs regarding contraception (contraception is morally wrong, contraception is difficult to obtain, contraception is expensive etc.). In addition, the Adolescent In-Home Interview contains the outcome variables of interest – access to health care (receipt of routine physical exam in the past year, receipt of c ontraceptive services in the past year, and unmet health needs). The variables from the Parent Questi onnaire examined in this study include: sociodemographics (household income level and pa rental education) and control variables: family structure (two parents, single parent a nd no parental figure identified by the adolescent), and health insurance (yes or no). The variables from the Contextual Databa se examined in this study are measured at the level of the cen sus tract and include: total population median age of population (most common age of the population in the census tract), residential stability (proportion over age five years living in the same house fo r past 5 years), and the focal variables of interest – neighborhood racial composition (proportion of Bl ack population residing within the census tract) and neighborhood socioeconomic composition (proportion of families within the census tract below the U.S. Census Bureau poverty threshold for 1989 and proportion of census tract over 25 years of age with less than a high school degree). Public Health Significance Given the limited evidence regarding the contribution of the neighborhood context to racial and socioec onomic disparities in access to h ealth care, the findings from this study will enhance the rese arch literature as well as our knowledge regarding the

PAGE 24

11 extent to which neighborhood characteristics influence social disparities in access to health care among a vulnerable population of adolescents. Futu re studies are needed to examine the potential mechanisms through which the neighborhood context contributes to social disparities in acce ss to health care; the interactions between local and state policies, neighborhood conditions, and individual characteristics on access to health care; and the extent to the neighborhood context cont ributes to social disparities in access to health care among other racial and ethnic groups, the adolescent male population, and adolescents who engage in other high-risk be haviors, such as tobacco, alcohol, and drug use. Second, this study may contribute theoreti cally and methodologic ally to the field of research related to neighborhoods and healt h. For example, current theories posit that access to institutional resources is one of the pathways through which neighborhoods influence individual health outcomes, yet fe w studies have explored how access to these resources may vary between neighborhoods with differing racial and socioeconomic conditions (Leventhal & Brooks-Gunn, 2000) The multilevel design and multilevel analyses employed with this study will allo w inferences to be made regarding the independent and interactive effects of ne ighborhoods and individuals on individual access to health care. Furthermore, this study s eeks to address current weaknesses in the measurement of neighborhoods, particularly related to the interplay of race and socioeconomic position on access to health care. Most studies have analyzed race and socioeconomic position as independent constr ucts, yet they are known to be closely associated (Acevedo-Garcia & Lochner, 2003; Acevedo-Garcia et al., 2003; Kawachi, Daniels, & Robinson, 2005). Further exploratio n of how these two constructs interact

PAGE 25

12 with one another in relation to access to health care may have important implications for understanding how the neighborhood context ma y contribute to social disparities in access to health care and the types of nei ghborhoods that need to be targeted for intervention. Finally, public health professionals need to lead social change efforts in an attempt to instill more collective attitudes among polic y makers, health care professionals, health related inst itutions, and the general public. Public health is rooted in social justice, which supports collective ac tion to ensure the equ itable distribution of resources, such as health care (Beaucham p, 2003). Although hea lth care is only one determinant of health, th e inequitable distribution of this resource based upon socioeconomic position and race is a violati on of human and civil rights principles (Bambas & Casas, 2003; Braveman & Grus kin, 2003; Smith, 2005). As public health professionals we must develop and implement interventions to change the broader social forces of society that perpetuate a nd sustain racism and discrimination. Hypotheses Research Objective I : To examine the extent to which individual race and socioeconomic position influence access to hea lth care services among sexua lly experienced adolescent females and the extent to which th e disparities vary across neighborhoods. IA. Sexually experienced adolescent female s who self-identified racially as Black will be less likely to have received a routine physical exam and contraceptive services and more likely to report unmet health needs than those sexually experienced adolescent females who are White, after adjusting for socioeconomic position and individual control variables.

PAGE 26

13 IB. Sexually experienced adolescent fe males who are lower in socioeconomic position will be less likely to have received a routine physical exam and contraceptive services and more likely to report unmet health needs than those sexually experienced adolescent female s who are higher in socioeconomic position, after adjusting for i ndividual race and individua l control variables. IC. The relationship between adolescent race and access to health care will be moderated by adolescent socioeconomic position, so that sexually experienced adolescent females who self -identified racially as Bl ack and who are lower in socioeconomic position will be less likely to have received a routine physical and contraceptive services and more likel y to report unmet health needs than adolescents who are White and higher in socioeconomic posi tion, after adjusting for individual control variables. ID. There will be variation in the receip t of a routine physical, the receipt of contraceptive services and the report of unmet health needs across neighborhoods. IE. The social disparities in access to health from Hypotheses IA-IC will vary across neighborhoods. Research Objective II : To explore the relationships be tween the neighborhood racial and socioeconomic context and the average odds of access to health care among the sexually experienced adolescent females in this study as the extent to which the neighborhood racial and socioeconomic context contribute s to the variation in access to health care across neighborhoods.

PAGE 27

14 IIA. The neighborhood racial context will be a ssociated with access to health care as the average odds of having received a routine physical and having received contraceptive services will be lower wh ile the average odds of reporting unmet health needs will be higher in those neighborhoods characterized by a higher proportion of Black residents than thos e neighborhoods characterized by a lower proportion of Black residents. IIB. The neighborhood socioeconomic context will be associated with access to health care as the average odds of having received a routine physical and having received contraceptive services will be lower while the average odds of reporting unmet health needs will be higher in those neighborhoods characterized by a higher proportion of poor or lower edu cated residents than those neighborhoods characterized by a lower proportion poor or lower educated residents. IIC. The relationship between the nei ghborhood racial context and the average odds of access to health will be mode rated by the neighborhood socioeconomic context in that the averag e odds of having received a routine physical and having received contraceptive services will be lower while the average odds of reporting unmet health needs will be higher in those neighborhoods that are characterized by a higher proportion of Black and poor or lo wer educated residents than in those neighborhoods that are characterized by a lower proportion of Black and poor or lower educated residents. IID. The neighborhood racial and socioeconom ic context will contribute to the variation in the average odds of acces s to health care across neighborhoods.

PAGE 28

15 Research Objective III : To explore the extent to which the neighborhood racial and socioeconomic context moderates the re lationship between individual race, socioeconomic position and access to health care among sexually experienced adolescent females. IIIA1. Adolescents who live in neig hborhoods characterized by a higher proportion of Black residents will be less likely to have received a routine physical or contraceptive se rvices and more likely to report unmet health needs than those adolescents who live in a neighborhood characterized by lower proportions of Black resident s but the negative effect of living in a neighborhood characterized by a higher pr oportion of Black resident s will be greatest among Black adolescents, after adjusting for the individual control va riables, individual socioeconomic position, neighborhood socioeconomic context and the neighborhood control variables. IIIA2. Adolescents who live in a ne ighborhood characterized by a higher proportion of Black residents will be less likely to have received a routine physical or contraceptive se rvices and more likely to report unmet health needs than those adolescents who live in a neighborhood charac terized by a lower proportion of Black resident s but the negative effect of living in a neighborhood characterized by a higher pr oportion of Black resident s will be greatest among adolescents who are of lower socioec onomic position, after adjusting for the neighborhood racial context and th e neighborhood control variables.

PAGE 29

16 IIIB1. Adolescents who live in a neighborhood lower in socioeconomic position will be less likely to have received a routine physical and contraceptive services and more likely to report unmet health need s than those adolescents who live in a neighborhood higher in socioeconomic position but the negative effect of living in a neighborhood lower in socioeconomic position will be greatest among Black adolescents, after adjusting for the neighborhood control variables. IIIB2. Adolescents who live in a neighborhood lower in socioeconomic position will be less likely to have received a routine physical and contraceptive services and more likely to report unmet health need s than those adolescents who live in a neighborhood higher in socioeconomic position but the negative effect of living in a neighborhood lower in socioeconomic position will be greatest among adolescents who are of lower socioeconomic position. IIIC. The cross-level relationships between IIIA and IIIB will vary across neighborhoods. Delimitations 1. The study sample is limited to those sexua lly active adolescent females who are at least 15 years of age by Wave I who have never been married. 2. The study is restricted to individual racial categories: Black or White. 3. The measurement of the neighborhood racial composition is restricted to the proportion of Blacks residing within the census tract.

PAGE 30

17 4. The individual and contextual measures of race are based upon socially constructed ideals that force individual s into specific ca tegories based upon phenotype. 5. The study examines only those selected access to health care measures. 6. The study is limited to adoles cents who lived in their census tract for at least one year. 7. The sample contains only those adolescen ts who had responses to the items of interest. 8. The neighborhood is defined as the census tract of residence. Limitations 1. The study employs a non-experimental design, thus causality cannot be established. 2. Mobility of residents (not respondent s) from neighborhoods between 1990 and 1994 could have influenced the racial a nd socioeconomic composition as well as the social context of the neighborhood over these four years, which are not captured in this study. 3. The individual and contextual data were collected in 1990-1994, t hus variations in the racial and socioeconomic compositi on of the United States, health care (insurance changes such as SCHIP, HMO penetration etc.) and social welfare policies that have occurred since this tim e could have an impact on access to health care, which are not captured in this study. 4. The investigation is restricted to variable s that were available in the Add Health data set, which is a limitation to any secondary data analysis.

PAGE 31

18 5. No information is available in the Add Health data set rega rding the number or type of health care providers at the level of the census tract. 6. No information is available from the A dd Health data set regarding the attitudes or behaviors of health care providers or participant-pr ovider communication. 7. Quality of health care received cannot be ascertained from the available data. 8. The measurement of the neighborhood r acial and socioeconomic context is limited to the racial and socioeconomic composition of the census tract. Ideally racial and socioeconomic segregation i ndices would be uti lized, but the Add Health data set does not c ontain the necessary informa tion related to contiguous census tracts needed for cal culation of these indices. 9. No information is available from the Add Health data set regarding social attitudes towards differing racial and socioeconomic groups in each neighborhood. 10. In the Add Health study, the questions related to sexuality are only asked to those adolescents who were at least 15 years of age, thus the experiences of sexually active adolescents under 15 years of ag e were not captured by this study. 11. Variables from the Adolescent In-Home In terview and the Parent Questionnaire were collected via self-report, of which some were based on recall of events over the year preceding the survey. 12. The residence of the adolescent refers to her permanent address as no information is available regarding dual resi dence between family members.

PAGE 32

19 Definitions Access to health care : A multidimensional construct that is measured for this study as “realized access” or utilization of health care (Aday et al., 2004) and unmet health needs. Black: A socially constructed ca tegory referring to the sk in color and phenotype of individuals who have origins in Afri ca (Smelser, Wilson, & Mitchell, 2001). Contraceptive services : A preventable health care servic e in which patients are provided contraceptives from a physician or nurse (Harri s et al., 2003). Discrimination : A social process through which memb ers of a social group are treated differently because they are member s of that group (Krieger, 2001). Gini Coefficient : Standard index of income in equality (Krieger et al., 1997). Institutional resources : Social and material assets of neighborhoods and communities that promote health and well being. Mean Statistical Area (MSA): Refers to a geographic area “with at least one urbanized area that has a population of at least 50,000.” The MSA contai ns the primary county or a group of counties that socially and economically share the urbanized area (Office of Management and Budget, 2000, p.12). Multilevel design : A type of study design that examines data across multiple levels of analyses. Multilevel analysis : An analytic technique that is usef ul to examine data that are nested within one another, such as individuals within neighborhoods. Multilevel analysis controls for the non-independence of observati ons that occurs due to this nesting as individuals who belong to a group (i.e. neighbor hood) are likely to be more similar to one another resulting in correlated data. Furthe rmore, multilevel analysis allows for the

PAGE 33

20 examination of the variability within and between individuals and groups as well as their interactions (Diez-Roux, 2003; H ox, 1995; Subramanian et al., 2003). Neighborhood: A geographic area where individuals reside. For this study, the census tract was used to opera tionalize the neighborhood. Political economy theory : A theoretical framework that can be used to examine the processes through which societies produce and distribute their social and material resources (Goldsmith & Gunn, 2003; Raphael, 2003). Race : A socially constructed category that is based upon the shared an cestry or skin color of social groups that is dictated by the dominant social group (Krieger, 2001). Racism : “Institutional and individual practices that create and reinforce oppressive systems of race relations” (Krieger, 2001, p. 696). Residential segregation : A social process that arises fr om social stratif ication in which population groups are sorted into specific residential areas base d upon characteristics such as race, ethnicity or so cial class resulting in the se paration of the population groups Residential segregation refers to “the composition and spatial distribution of the population of a metropolitan area among neighbo rhoods”. (Acevedo-Garcia & Lochner, 2003, p. 267). Routine physical exam: A preventive health care visit that includes a physical examination, screening for risk behaviors and health concerns, and developmentally appropriate anticipatory guidance. Sexually active adolescent females : Refers to adolescent females who have had vaginal intercourse at least once in their lifetime.

PAGE 34

21 Social class : Refers to social groups that are “determined by a society’s forms of property, ownership, and labour, and their conne ctions through produc tion, distribution, and consumption of goods, services, and information” (Krieger, 2001, p. 697). Often measured in the United States via occ upational class (Krieg er et al., 1997). Social disparities in health : The differences in health betw een social groups of differing levels of advantage (Braveman & Gruskin, 2003). Social disparities in access to health care : The differences in access to health care that occur between social groups of differing levels of advantage (Braveman & Gruskin, 2003). Socioeconomic position: Consists of resourcebased and prestige-based measures. Resource-based measures include material and social resources such as income, wealth, and education. “Prestige-based m easures refer to individuals’ ra nk or status in the social hierarchy, typically evaluated with referen ce to people’s access to and consumption of goods, services, and knowledge” (Krieger, 2001, p. 697). Social stratification : A process in which population groups are ranked into certain social positions based upon socially constructed ideals, su ch as race/ethnicity, social class, age, and/or gender. Social structure: Refers to the way a society orga nizes relationships between social groups, often based upon race or social class. Unmet health needs : Health problems that individuals thought needed medical attention but no health care or inadequate health care was received (Shi & Stevens, 2005).

PAGE 35

22 White : A socially constructed ca tegory referring to the sk in color and phenotype of individuals who have origins in Europe, North Africa, or the Middle East (Smelser et al., 2001).

PAGE 36

23 Chapter Two: Literature Review Introduction The following literature review provides an epidemiological overview of the reproductive health concerns that sexually experienced adolescent females encounter and the extent to which social disparities in a dolescents’ reproductive he alth exists. Next, the role of health care as an institutional res ource is discussed incl uding what is known about social disparities in access to health care among adolescents. An overview of the theoretical frameworks, social stratification and political economy, that guide this proposed study is then provided. Finally, the empirical evidence exam ining the influence of the neighborhood context on access to hea lth care is discussed, including a critical review of the gaps in the literature and th e methods through which this study will attempt to address them. Reproductive Health of Adolescent Females According to the 2003 Youth Risk Beha vior Survey, over 45% of adolescent females reported having sexual intercourse at least once in their lifetime (Grunbaum et al., 2004). Of particular concern is that many of these adolescents reported engaging in high-risk sexual behaviors. For example, among the 34% of adolescent females who reported being currently sexually active, over 42% reported not using a condom with their last sexual intercourse, 11% reported ha ving four or more se xual partners over their

PAGE 37

24 lifetime, and over 21% had been under the influe nce of either alcohol or illicit drugs with their last sexual intercours e (Grunbaum et al., 2004). Adolescent females who are sexually ac tive and engaging in high risk sexual behaviors are particularly vulnerable for e xperiencing an unintended pregnancy or STI, including HIV (Abma et al., 2004; Grunbaum et al., 2004). In the year 2000 alone, 230,000 adolescent females 15-19 years of age were diagnosed with chlamydia, over 68,000 were diagnosed with gonorrhea (CDC, 2004) and over 800,000 became pregnant, of which the majority were unintende d (Abma et al., 2004; Henshaw, 2004). Biological and social factors also increas e adolescent females’ risk for unintended pregnancy and sexually transmitted infections. Females are at greater biological risk for STI’s than males due to the nature of di sease transmission, but adolescent females are even more vulnerable than older females due to physiological differences in the cells lining their cervix (USDHHS, 2000; Shafii & Burstein, 2004). Females are also more likely than males to have asymptomatic inf ections that can result in delayed treatment and increased sequelae, such as pelvic inflam matory disease, infertility, cervical cancer, liver cancer, ectopic pregnancy, a nd even death (USDHHS, 2000). Socially, adolescent females are more likel y than males to be coerced into sexual intercourse or to be sexually abused. Vi ctimization increases the risk for STI and unintended pregnancy due to the inability to use protective measures (Abma, Driscoll, & Moore, 1998; USDHHS, 2000; Sarigiani et al., 1999). Furthermore, adolescent females who have been sexually abused are more likel y to engage in voluntary sexual intercourse at earlier ages, have a greater number of se xual partners, and experience an unintended pregnancy (Manlove et al., 2001; Mill er et al., 1995; Raj et al, 2000).

PAGE 38

25 While all sexually active adolescent fema les are at risk for experiencing an unintended pregnancy or cont racting a STI, adolescents who belong to socially disadvantaged groups are even more vulnerable. Social processes, such as racism and discrimination, impede access to social and ma terial resources that promote health and wellbeing (Beeghley, 2005; Thomas, 2003), poten tially contributing to worse health and social outcomes among those already socia lly disadvantaged (Leventhal & Brooks-Gunn, 2000). Social Disparities in the Reproducti ve Health of Adolescent Females Social disparities in the reproductive health of sexua lly active adolescent females are pervasive in the United States (Abma et al., 2004; CDC, 2004). For example, according to recent national data, Black adolescent females 15 years to 19 years of age were six times more likely than White female s to be diagnosed with chlamydia, three times more likely to be diagnosed with gonorrhea (CDC, 2004), and more likely to be diagnosed with HIV (CDC, 2005). Furthermore, recent data from the 2002 NSFG reported that 20% of Black adolescent fema les gave birth before the age of 20 years compared to only 8% of White adolescent females (Abma et al., 2004). A significant weakness of the above data is that the interplay between race and socioeconomic position is not explored. Research has noted that Black adolescents are more likely than White adolescents to live in poverty (DeNavas-Walt, Proctor, & Mills, 2005) and that adolescents who belong to lower socioeconomic population groups are more likely to experience STI’s (CDC, 2004) HIV (CDC, 2005), and early childbearing (Abma et al., 2004; AGI, 1999) than thos e adolescents who belonged to more socioeconomically advantaged population gr oups. Specifically, the 2002 NSFG reported

PAGE 39

26 that 23% of sexually active adolescent fema les whose mother had not completed high school or obtained a GED had a child before the age of 20 years compared to only 7% of those whose mother had some college edu cation (Abma et al., 2004). In addition, the Alan Guttmacher Institute (1999) reported that 83% of all adolescent females who gave birth were from poor or low income familie s. Thus, if studies do not adjust for socioeconomic position, the relationship between race and the prevalence of STI, HIV and early childbearing among Black a dolescents may be inflated. Historically, early childbearing has been a concern due to higher rates of high school drop out and lower educational at tainment among pregnant teens and young mothers, which often led to fewer opportunitie s in the labor market and ultimately, lower incomes and poverty (Moore, Myers, Morri son, Nord, Brown, & Edmonston, 1993). In addition, some studies have found a rela tionship between early childbearing and decreased cognition, knowledge and language skills (Moore, Morrison, & Greene, 1997; Terry-Humen, Manlove & Moore, 2005) among the children of adolescent mothers. However, the negative sequelae of adol escent childbearing has been called into question recently as research has indicated that many teenage mothers return to school, complete their education, and are financially si milar to their peers or siblings who had no children (Geronimus, 2003). Methodologically, this research compared the outcomes of adolescents who are of similar social pos ition rather than comparing those who are disadvantaged to those who are advantag ed (Geronimus, 2003). Geronimus (2003) theorizes that early childbearing among disa dvantaged adolescents may be adaptive, as the structural barriers and limited social mobility that disadvantaged social groups experience across their lifecourse are associated with higher morbidity and mortality rates

PAGE 40

27 at earlier ages (Geronimus, 2003). Conseque ntly, disadvantaged parents may not be healthy enough or even alive to raise their children into adulthood (Geronimus, 2003). According to a recent study though, the majority of all adolescent females who gave birth before 20 years of age reported th at their pregnancy was unintended either unwanted or mistimed (81.5 White, 76% Hi spanic, 80% Black) (Abma et al., 2004). Thus, prevention of early childbearing remain s vital for adolescents, but interventions must target the structural barriers and social forces that perpetuate social disadvantage and limit the resources and life opportunities fo r those disadvantaged groups (Geronimus, 2003). Neighborhood Context and Social Dispar ities in Reproductive Health A burgeoning field of research has demons trated that early childbearing and STI’s often cluster in neighborhoods that are ch aracterized by social disadvantage. For example, even after controlling for indivi dual characteristics, adolescent females who reside in socioeconomically disadvantaged neighborhoods are more likely to experience premarital childbirth than those adolescen t females who live in more affluent neighborhoods (Brooks-Gunn, Duncan, Klebano v, & Sealand, 1993; Kirby et al., 2001; South & Baumer, 2000; South & Crowder, 1999). In addition, South & Baumer (2000) reported that nearly two-thirds of the racial differences in adolescent childbearing rates were due to the greater likelihood that Black adolescents lived in more socioeconomically disadvantaged neighborhoods than White a dolescents (South & Baumer, 2000). Sucoff & Upchurch (1998) found that regardless of the neighborhood socioeconomic structure, living in a racially segreg ated neighborhood increased the risk for adolescent

PAGE 41

28 childbearing by 50%, which suggest s that racial residential segregation also directly impacts adolescent outcomes. Neighborhood socioeconomic disadvantage ha s also been associated with higher rates of STI’s and HIV among adults in severa l statewide studies. Sp ecifically, research has indicated that living in poor or low income neighborhoods increases the likelihood of contracting chlamydia, syphilis, and gonorrh ea (Krieger, Waterman, Chen, Soobader, & Subramanian, 2003). Furthermore, another st udy reported the cumulative incidence of AIDS was almost 7 times higher among those ad ults residing in block groups where 40% or more of the population was below the povert y line compared to those residing in block groups where less than 2% of the population wa s below the poverty line (Zierler et al., 2000). Neighborhoods are hypothesized to influe nce individual health through a variety of mechanisms, such as the social organizati on of the neighborhood, avai lability of social networks, and access to institutional resour ces (Jencks & Mayer, 1990; Leventhal & Brooks-Gunn, 2000). However, few studies have explored how the neighborhood context may influence these potential pathways, particularly access to institutional resources (Leventhal & Brooks-Gunn, 2000). Neighborhood resource models focus on the “availability”, “accessibility”, “affordability” and “quality” of resources in the community that promote health and well being (Leventhal & Brooks-Gunn, 2000). Exam ples of resources include “learning, recreational, and social activities; child care; schools; medical facili ties; and employment opportunities” (Leventhal & Brooks-Gunn, 2000, p. 322). Resear chers hypothesize that socially disadvantaged neighborhoods have fe wer of these resources than those more

PAGE 42

29 affluent neighborhoods and this inequitable di stribution can influenc e residents’ health and social outcomes (Leventhal & Brooks-Gunn, 2000). Health care is one of the many institutional resources that plays a role in shaping adolescent health (Leventhal & Broo ks-Gunn, 2000). However for sexually active adolescents, the receipt of health care se rvices may be even more critical for the prevention of unintended pregna ncy and STI’s, including HIV. Access to Health Care: An Institutional Resource Health care is a vital resource for sexu ally active adolescents when services include education about sexuality and safe se x practices, the provision of contraceptives, and early detection and management of pre gnancy and STI’s (Brindi s, 2002). According to clinical practice guideline s established by the American Academy of Pediatrics (AAP) and the American Medical Association (A MA), all adolescents should receive a confidential, yearly preventive health care vi sit that includes screening for sexual activity and counseling on safer sex practices. Furthermor e, if sexually active, adolescents should be tested at least yearly for STI’s and more frequently if clinical signs or sexual history warrant (AAP, 2000; AMA, 1997). Sex education and counseling are key compone nts of adolescent health care that should be provided to adolescen ts at their yearly physical at focused visits when prescribing contraceptives, when screening or treating STI’s, and at general illness visits if targeted preventive counseling is need ed (AAP, 2001). Health care services that provide safer sex education or contraceptives to adolescents are frequently the topic of political debate due to concerns that in creased access will encourage sexual activity among adolescents (Brindis, 2002). Although re search examining these concerns has

PAGE 43

30 found that counseling in the cl inic setting neither increases nor decreases sexual activity, studies have found that they do increase adolescents’ use of condoms and other contraceptives (Kirby, 2002; Boekeloo et al., 1999; Danielson et al ., 1990; Orr et al., Scher, 2004; Winter & Breckenmaker, 1991). Fu rthermore, researchers have attributed the recent decline in adolescen ts’ unintended pregnancy rates over the past decade to be due in part to adolescents’ use of more eff ective contraceptives as well as their increased consistency in using contraceptives (Boonstr a, 2002; Darroch & Singh, 1999; Santelli et al., 2004). Thus, health care services may play an important role in preventing unintended pregnancy and STI’s, including HIV, among adolescents, particularly for those adolescents who are already sexually active. Adolescents may receive health care se rvices through a variety of settings, including private physicians’ offices, co mmunity health centers, public health departments, school-based health clinics, family planning clinics, walk-in clinics, academic medical centers, and emergency de partments (Brindis, Park, Ozer, & Irwin, 2002; Society for Adolescent Medicine, [SAM], 2004). Ideally adoles cents would receive health care services in the cont ext of a medical home, in whic h they have a regular source of primary health care that o ccurs either by the same provider or at the same physical place (Starfield & Shi, 2004). Evidence has demons trated that the receip t of services from the same provider results in earlier and more accurate diagnosis, better monitoring, fewer unmet health needs, and increased patient satisfaction than the receipt of services from the same place (Starfield & Shi, 2004). Howe ver, health care services need to be adolescent-oriented with health care profe ssionals providing devel opmentally appropriate

PAGE 44

31 services, open patient-provider communication and confidentiality assurances (Brindis et al., 2002; SAM, 2004). Access to health care has been define d by the Institute of Medicine (IOM, 1993, p. 4) as “the timely use of personal health se rvices to achieve the best possible outcomes” and refers to an individual’s ability to obtain health care a nd the barriers he or she faces when trying to obtain services (USDHHS, 2003) Access to health care is a broad and multidimensional construct that has been operationalized in a variety of ways, such as, (1) the actual utilization or receipt of h ealth care services, including the number of hospitalizations that could have been avoi ded if timely health care was received, (2) factors that influence entry into the heal th care system, such as health insurance, having a usual source of care or regular provide r, or the individual’s perceptions of their health needs, (3) structural barriers within the health care system, including difficulty in obtaining care due to transpor tation issues, inconvenient o ffice hours, waiting times, length of time before able to make an a ppointment, (4) patientprovider communication and relationship quality, and (5) quality of health care (Aday et al., 2004; IOM, 2003; USDHHS, 2003). Research examining adolescents’ access to health care has indicated inadequate access exists across a variety of measures. Fo r example, adolescents are less likely than any other age group of children to receive a ny health care, to have health insurance (Elixhauser et al., 2002) or to receive preventive hea lth care (Yu, Bellamy, Kogan, Dunbar, Schwalberg, & Schuster, 2002) and more likely to report having unmet health needs (Newacheck, Hughes, Hung, Wong, & Stodda rd, 2000). Health insurance plays a vital role in improving access to health care in the United States (Elixhauser et al., 2002;

PAGE 45

32 Shi & Stevens, 2005) and adolescents who have health insurance are more likely to receive preventive health care (Elixhauser et al., 2002) a nd less likely to report unmet health needs (Ford et al.,1999). Adolescent females who are sexually active are particularly vulnerable for having inadequate access to preventive reproductive he alth care based on concerns of or actual breaches in confidentiality (B rindis, 2002; Brindis et al., 2002; Finer & Zabin, 1998; Ford et al., 1999; Zimmer-Gembeck, Alexander, & Nystrom, 1997). Numerous studies have indicated that many adolescents report they would be less likely to seek health care services for reproductive health concerns if they thought their parents disapproved of their sexual activity (Cheng, Sava geau, Sattler, & DeWitt, 1993; Scher, 2004) or if their parents would be notified that they were seek ing reproductive health care services (Jones, Purcell, Singh, & Finer, 2005; Reddy, Fleming, & Swain, 2002). Confidential reproductive health care services allow adolescents to seek out and receive counseling and/or treatment related to their sexual a nd reproductive health without the consent or notification of their pa rents. Although parental involveme nt is ideal, many adolescents are unable or do not believe they can involve their parents in decisions related to their sexual health and thus, they may delay or forgo necessary health care (Cheng, et al., 1993; Jones et al., 2005; Reddy et al., 2002). Although many adolescents have limited acces s to health care services, social disparities in accessing health care are pervasiv e (Bartman et al., 1997; Elster et al., 2003; Ford et al., 1999; Lieu et al., 1993; New acheck et al., 2003; Shenkman et al., 2003; Simpson et al., 2005; Stevens & Shi, 2003). The inequitable distribution of this resource among adolescents who belong to socially disadvantaged population groups is

PAGE 46

33 particularly alarming and unjust due to their greater vulnerability for STI’s, HIV and unintended pregnancy (Abma et al ., 2004; Grunbaum et al., 2004). Social Disparities in Adolescents’ Access to Health Care Numerous studies have demonstrated that si gnificant social disparities in access to health care exist among adolescents in the Unite d States (Bartman et al., 1997; Elster et al., 2003; Ford et al., 1999; Lieu et al., 1993; Newacheck et al., 2003; Shenkman et al., 2003; Simpson et al., 2005; Stevens & Shi, 2003; Wilson & Klein, 2000). For example, adolescents from low income families are less li kely to be insured, to receive any health care (Newacheck et al., 2003; Simp son et al., 2005), and to have a preventive health care visit. In addition, they are more likely to fo rgo health care due to costs (Newacheck et al., 2003) and to utilize the emergency departme nt as their usual s ource of care (Wilson & Klein, 2000). Similar racial and ethnic disparities in access to health care have also been reported even after controlling for family so cioeconomic position (E lster et al., 2003). Specifically, Black adolescents are less likel y than White adolescents to receive any health care (Bartman et al., 1997; Lieu et al ., 1993) as well as pr eventive health care (Shenkman et al., 2003) and they are more likel y to forgo health car e (Ford et al., 1999) and to utilize the emergency department as their usual source of care (Wilson & Klein, 2000).

PAGE 47

34 Theoretical Frameworks Social Stratification Social stratification refers to how we di fferentiate, evaluate, and reward social groups within our societya process that ultim ately leads to the inequitable distribution of resources across the social groups (Beeghl ey, 2005; Thomas, 2003). Societies stratify their members based on ascrip tion and achievement processes (Beeghley, 2005; Thomas, 2003). Ascription refers to stratification based on those social characteristics or conditions we are born into, su ch as our race, ethnicity, so cial class, or gender while achievement refers to stratification based on the attainment of some social ideal, such as higher education or a prestigious occupati on (Beeghley, 2005; Thomas, 2003). A primary difference between the two is the latter as sumes people have a personal choice regarding their social position (e.g. if people work hard enough they can achieve wealth, prestige and status) while the former relegates us into a social position on th e basis of criteria in which we have no control (Beeghley, 2005; Thom as, 2003). Social stratification occurs in all societies, but the ideals that serve as th e basis for stratificati on may differ depending on what attributes are considered to be of subjective importan ce to that society (Beeghley, 2005; Thomas, 2003). Social stratification implies social inequa lity and results in discrimination and the inequitable distribution of social and materi al resources, such as education, employment, housing and health care among the social groups (Beeghley, 2005; Thomas, 2003). Accordingly those social groups considered to be higher in social position accumulate

PAGE 48

35 more resources, furthering their economic and political power in the social structure (Beeghley, 2005; Thomas, 2003). Social stratification often results in reside ntial segregation in which specific social groups are “distributed” into disparate neighborhoods thr ough unfair housing practices and interpersonal discrimination. based upon th eir race, ethnicity, an d/or social class (Beeghley, 2005; Williams & Collins, 2001) Typically those social groups who are considered to be lower in social position reside in more disa dvantaged neighborhoods, limiting their access to institutional resources even further due to private and public disinvestments in the social and public infr astructure of these communities (Andrulis & Duchon, 2005; Beeghley, 2005; Whiteis, 1997). While many laws have been enacted in the United States prohibiting racism, disc rimination, and segregation, these social processes are still pe rvasive and often institutionalized through political, economic and social policies that restrict opportunities and resources for more disadvantaged social groups (Beeghley, 2005; Hofricht er, 2003; Krieger, 2000). The Political Economy of Health Care Political economy theories examine the processes through which societies produce and distribute their social and ma terial resources (Goldsmith & Gunn, 2003; Raphael, 2003). Thus, they are pa rticularly relevant for examining social disparities in health care as they center on the economic and political power structures that control the distribution of resources be tween the different social groups (Goldsmith & Gunn, 2003; Raphael, 2003).

PAGE 49

36 In the United States, the economic market politics and institu tional power are the underlying forces driving the health care sy stem (Putsch & Pololi, 2004; Whiteis, 1997, 1998). Consequently the health care system has become increasingly privatized and dominated by large corporations limiting ac cess to health care for those social groups who lack the economic means to purchase se rvices (Putsch & Pololi, 2004; Whiteis, 1997; 1998). For example, the United States is the only Western i ndustrialized country without a national health insurance plan (Quadagno, 2004). Consequently 15.7% of Americans are without health insurance (DeNavas-Walt et al., 2005), which significantly limits their access to health care. Scholars a nd advocates attribute th e lack of a national health insurance plan to the successful lobbyi ng of powerful stakeholde rs, such as health insurance companies, whose primary concern is the loss of economic gain if a national plan was instated (Qua dagno, 2004; McCanne, 2004). Powerful corporations lobby policym akers through communication and/or contributions to political campaigns in an e ffort to sway U.S. health policy (Kushel & Bindman, 2004; Landers & Sehgal, 2004). In th e year 2000, pharmaceutical and health product companies spent nearly twice as much money as physician’s and other health professionals, and eight times more money th an advocacy and public health organizations lobbying U.S. politicians (Kushel & Bi ndman, 2004; Landers & Sehgal, 2004). The United States also spends more m oney per capita on health care than any other country belonging to the Organizati on for Economic Cooperation and Development (OECD), yet the utilization of health care se rvices among Americans is actually lower than the OECD median (Anderson, Reinhard t, Hussey, & Petrosyan, 2003). This means that we are paying much more for health car e services than the other OECD countries

PAGE 50

37 (Anderson et al., 2003). Unfortunately, this has not translated into better health (Institute of Medicine, 2003) and social disparities in health and health care are pervasive throughout the country. The Neighborhood Context and Access to Institutional Resources The Neighborhood Racial and Socioeconomic Context in the U.S. The neighborhood context can be define d as the racial and socioeconomic composition or the heterogeneity of the social groups that reside within a neighborhood (Acevedo-Garcia & Lochner, 2003; Aceve do-Garcia et al., 2003). The neighborhood racial and socioeconomic composition often serv es as a proxy for residential segregation, but conceptually and operationally they are different (Acevedo-Garc ia & Lochner, 2003; Acevedo-Garcia et al., 2003). Specifically, the former refe rs to only the racial or socioeconomic composition within a neighborhood while residential segregation refers to the racial or socioeconomic composition within a neighborhood and how this composition compares to the racial or so cioeconomic distribution of the surrounding metropolitan area (Acevedo-Garcia & Lochner, 2003; Acevedo-Garcia et al., 2003). Thus the latter allows for a more detailed exam ination of how different social groups are distributed across an entire metropolitan area and how this may cont ribute to inequality (Acevedo-Garcia & Lochner, 2003; Acevedo-Garcia et al., 2003). The neighborhood context is often conceptu alized and examined as two distinct concepts either racial segregation or income segregation (Acevedo et al., 2003; Acevedo-Garcia & Lochner, 2003; Jargow sky, 1996). However, the two are closely intertwined and by exploring their interplay a better unde rstanding of how neighborhood

PAGE 51

38 segregation influences health and access to health care may be achieved and potentially strengthen current policy initiatives (Jargowsky, 1996). Empirical evidence over the past decade has demonstrated that residential segregation based upon race or et hnicity is a stronger force separating social groups than segregation based upon social cl ass or income with Blacks remaining highly segregated from Whites, regardless of their income le vel (Acevedo-Garcia et al., 2003; AcevedoGarcia & Lochner, 2003; Charles, 2003; Masse y & Fischer, 1999). Hispanics and Asians, especially new immigrants, are also moderate ly segregated from Whites, but much less than Blacks, and as their socioeconomic positioning increases they tend to become more integrated (Charles, 2003). However, Hisp anics of Black or mixed race origin are reported to experience pattern s of residential segregation similar to Blacks, which supports the hypothesis that skin color remains a dominant discrimina tory characteristic in the United States (Charles, 2003; Massey & Denton, 1989). Interpersonal and institutional racism has led to discrimina tory practices in the government and housing industry (real estate, mortgage construction) constraining th e options for Blacks to move out of less advantaged neighborhoods (AcevedoGarcia et al., 2003; Acevedo-Garcia & Lochner, 2003; Charles, 2003). Although race is the primary characteris tic by which neighborhoods are stratified, segregation also occurs along socioeconom ic lines (Acevedo-Garcia et al., 2003; Acevedo-Garcia & Lochner, 2003; Charle s, 2003; Jargowsky, 2003; Jargowsky, 1996). According to recent data, the number of high poverty neighborhoods in the U.S. declined significantly from 1990-2000 due to the ec onomic gains experienced throughout the decade (Jargowsky, 2000). High poverty nei ghborhoods can be defined as those

PAGE 52

39 neighborhoods in which “40% or more of its residents are classified as poor using the federal poverty standard” (Jargowsky, 2003, p. 3) The number of persons living in high poverty neighborhoods decreased among all racial and ethn ic groups with Blacks experiencing the greatest decline, although 19% of all Blacks continue to reside in high poverty neighborhoods (Jargowsky, 2003). Ce ntral cities and rural areas both experienced declines in the number of high poverty neighborhoods, but suburban areas remained unchanged (Jargowsky, 2003). However, researchers express concern that the recent economic recession may lead to a re versal in some of these positive gains (Jargowsky, 2003). Residential segregation has significant consequences fo r the residents of those neighborhoods (Acevedo-Garcia et al., 2003; A cevedo-Garcia & Lochner, 2003; Charles, 2003). Disinvestments in the social and public infrastructure of se gregated neighborhoods leads to fewer institutional resources that promote health and well being, such as employment opportunities, safe housing, qual ity schools, healthy foods, recreational facilities and health care services (Charles, 2003; Leventhal & Brooks-Gunn, 2000). Without access to these resources, social mobility is limited and individual and population health are compromised (Cha rles, 2003; Leventhal & Brooks-Gunn, 2000). The Neighborhood Context and Access to Health Care Social stratification and political economy theories posit that access to social and material resources are distri buted inequitably among the soci al groups with those more advantaged groups accumulating a greater number of resources (Beeghley, 2005; Goldsmith & Gunn, 2003; Raphael, 2003; Thomas 2003). This inequitable distribution of resources occurs across geographic areas such as the community and neighborhood, as

PAGE 53

40 well as at the level of the individual. Res earchers have speculated that the neighborhood may impact health outcomes through the inequitable distribution of institutional resources, including health care, but few have actually explored this relationship (Leventhal & Brooks-Gunn, 2000). However, the limited evidence that does exist suggests that residential areas characterized by higher proportions of Black and /or poor populations have fewer health care providers (Guagliardo et al., 2004; Komaromy, et al., 1996). Furthermore, individuals residing in disadvantaged areas report fewer acute and preventive health care visits and more unm et health needs (Andersen et al., 2002; Brooks-Gunn et al., 1998; Kirby & Kaneda, 2005). According to a recent ecological stu dy, census tracts characterized by higher proportions of Black children (r = -0.68, p < 0.0001) and lower median incomes (r = 0.44, p < 0.0001) were independently a ssociated with fewer pediatri c primary care providers in the census tract despite a greater number of pediatric providers in the overall metropolitan area than the national average (Guagliardo et al., 2004). Similar findings were found in another ecological study (Kom aromy et al., 1996) in which communities with higher proportions of Black residents had fewer primary care physicians than those communities that were more heterogene ous even after controlling for neighborhood poverty. Furthermore, when the researchers c ontrolled for community race and ethnicity, neighborhood poverty was not associated w ith the number of physicians in the community (Komaromy et al., 1996). The rese archers also found that Black physicians were more likely to practice in neighborhoods in which there were a higher proportion of Black and poor residents and fewer primary care providers per capita as compared to White physicians (p<0.001).

PAGE 54

41 Unfortunately due to the ecological nature of these studies inferences about how the neighborhood’s availability of health ca re providers may influence individual access to health care is unknown. Furthe r research is needed to expl ore how the availability of health care providers may moderate th e relationship between the racial and socioeconomic composition of communities and neighborhoods and individuals’ access to health care. National datasets, such as A dd Health, typically provi de data regarding the availability of health care pr oviders at the level of the count y, but this larger geographic area may be unable to detect the inequitable di stribution of providers at smaller levels of analysis, such as the neighborhood. Evidence from several contextual analyses also indicated that the neighborhood context may influence individuals’ access to health care. For example, living in a socioeconomically disadvantaged neighborhood was reported to be associated with an increased utilization of the emergency depa rtment among formerly premature children (Brooks-Gunn et al., 1998) as well as a decreased likelihood of having a usual source of care and blood pressure screening and an in creased likelihood of reporting unmet health needs among adults (Kirby & Kaneda, 2005). Furthermore, another study reported low income children who resided in higher income metropolitan statistical area’s (MSA) were more likely to receive health care over the previous year, but so were children who resided MSA’s characterized by greater in come inequality (Andersen et al., 2002). Perhaps health care resources are still availa ble in MSA’s with greater income inequality increasing the likelihood that lower income children will be able to access them. However, the effect size for this last study was low and the research ers did not report if they statistically controlled for the violat ion of independence due to the nesting of

PAGE 55

42 individuals within communities. Thus the st andard errors may be biased downward and the findings falsely significant (Diez-Roux, 2003). One contextual analysis examined the im pact of the county’s racial composition on access to health care, controlling for the c ounty socioeconomic structure (Haas et al., 2005). The study was more complex than the pr eceding studies as th e researchers also examined the moderating influence of indivi dual race (controlling for individual income) on the relationship between the county racial composition and access to health care. The researchers reported that Blacks experienced fewer problems in obtaining health care and fewer financial barriers when they lived in a county that had a higher percentage of Blacks than those who lived in counties with a lower percentage of Blacks. There was no difference in access to health care between Whites who lived in c ounties with higher or lower proportions of Black re sidents (Haas et al., 2004). Since the study controlled for individual income, health status, health insurance, as well as county income and education levels the findings may indica te that living in a more homogenous county that is similar to one ’s own racial identity increases access to health care (Haas et al., 2004). The relationship could be possi bly due to stronger social networks and social support among similar ra cial groups and perhaps less racism and discrimination. In a multilevel study, Scher (2004) examined the influence of the socioeconomic composition of counties on the receipt of contraceptive, family planning, and STI services among sexually active adolescent fema les. When controlling for individual, family, school and other county characteristics, no significant results were found (Scher, 2004). The researcher did not provide any information on building the hierarchical

PAGE 56

43 model, thus the impact of the numerous c ontrol variables as potential mediators is unknown. Furthermore, the study analyzed the impact of the socioeconomic composition at the level of the county and as noted previously this higher level of analysis may be unable to detect the differences in access to health care among those living in disparate neighborhoods. Although the preceding studies add to the l iterature, they exclude the interplay between racial and socioeconomic position, which is a significant limitation throughout the literature (Aceve do-Garcia & Lochner, 2003; Aceve do-Garcia et al., 2003; Kawachi et al., 2005). Researchers often control for one or the other, examine the two independently or they exclude one or the ot her altogether (Acevedo-Garcia & Lochner, 2003; Acevedo-Garcia et al., 2003; Kawachi et al., 2005). However in the United States, social and material reso urces are distributed among the population groups based upon race and social class with those more adva ntaged groups accumulating a greater number of resources (Beeghley, 2005; Thomas, 2003). Furthermore, race is the most powerful characteristic through which residential segr egation takes place, often contri buting to neighborhood poverty and reduced access to resources (Acevedo-Garcia & Lochner, 2003; Acevedo-Garcia et al., 2003; Charles, 2003). Thus excluding the interplay between race and socioeconomic position limits our unders tanding of social disparities in health and health care and impedes our ability to develop more effective policies to address these injustices. One recent ecological analysis examini ng the impact of the neighborhood context on access to supermarkets demonstrates th e potential moderating influence of the neighborhood socioeconomic composition on th e relationship between the neighborhood

PAGE 57

44 racial composition and the neighborhood’s access to supermarkets (Zenk, Schulz, Israel, James, Bao, & Wilson, 2005). Public health prof essionals have been examining the extent to which neighborhood characteristics may in fluence access to other health promoting resources, such as healthy foods (Cradoc k et al., 2005), due to the higher rates of diabetes, heart disease, and stroke among racial and ethnic minorities that may be function of poor diet (Morland, Wing, Diez-Roux, & Poole, 2002). The researchers reported that neighbor hoods characterized by high proportions of Blacks or high levels of poverty were independe ntly associated with fewer supermarkets that those more integrated and wealthy neighborhoods (Zenk et al., 2005). However, when exploring potential in teractions, the researcher s found that neighborhoods characterized by a medium and high concentra tion of Blacks had fewer supermarkets, but only if the neighborhood was also characte rized by high poverty le vels (Zenk et al., 2005). Furthermore, among the least impove rished neighborhoods, the number of supermarkets was found to be similar for low, medium and high concentrations of Blacks (Zenk et al., 2005). In addition to the paucity of research examining the interplay between race and socioeconomic position, the measurement of race and socioeconomic position varies across the studies due to the multidimensional nature of these constructs as well as limitations in the way data are collected. As discussed previously, th e ideal indicator to measure the neighborhood racial and socio economic context would be residential segregation as this measure takes into c onsideration the racial and socioeconomic composition within a neighborhood and how this composition compares to the racial or socioeconomic distribution of the surr ounding metropolitan area (Acevedo-Garcia &

PAGE 58

45 Lochner, 2003; Acevedo-Garcia et al., 2003). However, due to data restrictions many studies are unable to assess the segregati on indices and instead measure racial and socioeconomic composition. Currently none of th e studies examining the influence of the neighborhood context on access to health have employed segregation indices. Unfortunately this study is unable to meas ure racial and socioeconomic residential segregation due to a lack of information regarding conti guous census tracts in the Add Health Contextual database, thus this study is also limited to the examination of the racial and socioeconomic composition of the census tract. Researchers recommend the utilization of several socioeconomic measures in order to capture potential differences on th e outcome (Krieger et al., 1997; Lynch & Kaplan, 2000; Robert, 1999). Theoretic ally, common indicators measuring socioeconomic position, such as education and income, may impact the outcome of interest differently (Krieger et al., 1997; Lynch & Kaplan, 2000; Robert, 1999). The current studies examining the imp act of the socioeconomic structure on individuals’ access to health care employe d a variety of area based socioeconomic indicators and all reported a significant relations hip. The indicators included, (1) per capita income, (2) the Gini coefficient of in come inequality (Andersen et al., 2002), (3) the proportion low income (BrooksGunn et al., 1998; Scher, 2004), (4) the proportion wealthy (Brooks-Gunn et al., 1998; Scher, 2004), (5) the proportion in poverty (Guagliardo et al., 2004; Komaro my et al., 1996), and (6) a composite of socioeconomic disadvantage (percent of residents in the block group with incomes less that 125% of federal poverty level, the percent of reside nts over 16 years old who were unemployed, and the percent of resident s over 18 years old with no hi gh school diploma or GED)

PAGE 59

46 (Kirby & Kaneda, 2005). Although socioeconomic composites can capture more variation in the socioeconomi c structure, they can be di fficult when interpreting and planning interventions as the indicator ha ving the most influen ce is unknown (Robert, 1999). Conceptual and operational issu es also arise in regard to the dependent variable, access to health care. Currently no indicator for access to health care is considered superior over another, but most often the construct is measured either as “realized access” or utilization of health care services, unmet health needs or forgone health care, and the number of preventable hospitalizations (Lur ie, 2004). The studies discussed in this review measured individual access to hea lth care in a variety of ways including, (1) receipt of preventive health care services (Kirby & Kaneda, 2005; Scher, 2004), (2) receipt of health care /seen a physician (Andersen et al., 2002; Brooks-Gunn et al., 1998; Scher, 2004), (3) unmet health needs (Haas et al., 2005; Ki rby & Kaneda, 2005), (4) having a usual source of care (Kirby & Kaneda, 2005), and (5) the utilization of the emergency department (Brooks-Gunn et al., 1998). Conclusions Sexually active adolescent females are at high risk for unintended pregnancy and STI’s, including HIV due to a variety of beha vioral, biological and social factors (Abma et al., 2004; Abma et al., 1998; Grunbaum et al., 2004; Manlove et al., 2001; Mi ller et al., 1995; Raj et al., 2000; Sarigiani et al., 1999; Shafii & Burstein, 2004; USDHHS, 2000). However, adolescents who belong to socially disadvantaged populations are even more vulnerable due to social forces, such as ra cism and discrimination that impede their access to institutional resources at the ne ighborhood and individual level. Current

PAGE 60

47 research has reported that low income and mi nority adolescents are more likely to have decreased access to health care, including a higher incidence of unmet health needs, fewer preventive health care visits, and uninsur ance (Bartman et al., 1997; Elster et al., 2003; Ford et al., 1999; Lieu et al., 1993; Ne wacheck et al., 2003; Shenkman et al., 2003; Simpson et al., 2005; Stevens & Shi, 2003). The lack of health care due to racial and socioeconomic barriers is unjust and a viola tion of human and civi l rights principles (Bambas & Casas, 2003; Braveman & Gr uskin, 2003; Smith, 2005). Furthermore, decreased access to health care services among sexually active adolescent females may increase their likelihood fo r experiencing an unintended pregnancy or STI. Currently, the majority of the research has examined barriers related to the individual, provider, and health care system that may contribut e to social disparities in access to health care. However, social disp arities in access to h ealth care may be a consequence of social stratification if soci al and material res ources are distributed inequitably between neighborhoods and co mmunities based upon their racial and socioeconomic context (Beeghley, 2005; T homas, 2003). Although several studies have reported evidence that the neighborhood r acial and socioeconomic structure may influence access to health care, a paucity of research exists. Furthermore, several limitations in the design and methods of thes e studies and to neighborhood research in general exist. First, several of the studies reviewed for this study employed an ecological design, which limits the inferences of th e findings to neighborhood access only. Several of the other studies employed contextual analyses in which group level data are disaggregated and measured at the individua l level of analysis. Researchers may use

PAGE 61

48 contextual analyses in this wa y, but statistical procedures mu st be employed to correct for the violation of independence since the group an d individual level data may be correlated with one another due to the nesting of the data (Diez-R oux, 2003). Further limitations of contextual analyses are that they only pe rmit the examination of individual variability since the data are examined at the indi vidual level of analys es (Diez-Roux, 2003). Advances in statistical procedures, such as multilevel modeling, enable researchers to examine data across multiple contexts without concerns over violating the independence assumption required of standard statistical procedures. Furthermore, multilevel analyses allows for the exploration of the exte nt to which the neighborhood and individual influence the outcome as well as the variabil ity in the outcome be tween individuals and neighborhoods (Diez-Roux, 2003). Second, many of the studies discussed in the previous review of the literature employ cross-sectional designs, which prohibi t examination of the temporal order of events that is necessary to establish caus ality. Longitudinal studi es of neighborhoods are difficult and expensive, particularly if the study is nationally repres entative. Researchers have conducted randomized community trials, such as the Moving to Opportunity Study, in which individuals and families were randomized to one of three neighborhood conditions and then followed prospectivel y (Leventhal & Brooks-Gunn, 2000). However, these studies are also very expensive and most likely constrained to small geographic areas or communities limiting generalizability of the findings. Third, which area-based indicator (Gini coefficient of income inequality, mean income, poverty levels etc.) should be measur ed and at what level of analysis (block group, census tract, county, state, etc.) has been discussed ex tensively in th e literature.

PAGE 62

49 For example, certain outcomes of interest may not be detected at higher levels of analyses due to the heterogeneity of larger geographi c areas while other outcomes, such as the Gini coefficient of income inequality, ma y demand the heterogeneity (Diez-Roux, 2003; Subramanian et al., 2003). Furthermore, due to the paucity of research examining the extent to which the neighborhood racial and so cioeconomic structure influences access to health care, little is known about which neighborhood measures will most accurately detect an effect. Ideally racial and economi c segregation indices would be utilized, but this is not always possible due to constrai nts of secondary data. The studies reviewed here all examined the neighborhood racial co ntext using the proportion of Blacks within the geographic area and the neighborhood socioeconomic c ontext was measured with variety of indicators. Recent studies examin ing the impact of the neighborhood context on individual health outcomes has reporte d that the most stable area-based socioeconomic measure is the proportion of households below povert y measured at the level of the census tract (Krieg er et al., 2005). Research examin ing the utility of different area-based racial and socioeconomic indicato rs on individual access to health care is needed. Finally, few of the studies discussed in th is literature review and in the wider neighborhood literature have explored the in terplay between the neighborhood racial and socioeconomic structure (Acevedo-Garcia & Lochner, 2003; Acevedo-Garcia et al., 2003; Kawachi et al., 2005). Often researchers will control for one or the other, but few have examined the moderating influences of socioeconomic position on race (AcevedoGarcia & Lochner, 2003; Acevedo-Garcia et al., 2003; Kawachi et al., 2005). As was evident by the Zenk et al., study (2005), both the neighborhood racial and socioeconomic

PAGE 63

50 context demonstrated independent effect s. However, when neighborhood poverty was introduced as a moderator, the relationship between the neighborhood racial structure and access to supermarkets was conditional base d on the level of neighborhood poverty (Zenk et al., 2005). The subtleties in the relationships may only be detected when examining the interplay between the two and the findings may have significant implications for informing the development of policies to help overcome social disparities in health and health care.

PAGE 64

51 Chapter Three: Methodology Purpose of the Study The overall aim of this study was to e xplore the contribution of the neighborhood context to racial and socioeconomic dispar ities in access to health care among sexually experienced adolescent females. Specific objectives of this study included: (1) to examine the extent to which social disparities in acce ss to health care existed and the extent to which access to health care and social dispar ities in access to health care varied across neighborhoods, (2) to examine the relations hips between the neighborhood racial and socioeconomic context and the average odds of access to health care, and (3) to examine the extent to which the neighborhood racial and socioeconomic context contributed to social disparities in access to health care among this sample of sexually experienced adolescent females and the extent to which th is relationship varied across neighborhoods. Figure 1 depicts the conceptual model of th e hypothesized relationships in this study.

PAGE 65

52 Figure 1. Conceptual Model Neighborhood Racial Context Proportion Black (1990 Census) Adolescents’ Access to Health Care Contraceptive Services (yes, no) Yearly Physical (yes, no) Unmet Health Needs (yes, no) (1994-1995 Survey) Neighborhood Socioeconomic Context Proportion Below Poverty Proportion Low Educated (1990 Census) Individual Race White, Black (1994-1995-Survey) Individual Socioeconomic Position Income to Poverty Ratio Parental Education (1994-1995 Survey)

PAGE 66

53 Hypotheses Research Objective I: To examine the extent to which individual race and socioeconomic position influence access to hea lth care services among sexua lly experienced adolescent females and the extent to which th e disparities vary across neighborhoods. IA. Sexually experienced adolescent female s who self-identified racially as Black will be less likely to have received a routine physical exam and contraceptive services and more likely to report unmet health needs than those sexually experienced adolescent females who are White, after adjusting for socioeconomic position and individual control variables. IB. Sexually experienced adolescent fe males who are lower in socioeconomic position will be less likely to have received a routine physical exam and contraceptive services and more likely to report unmet health needs than those sexually experienced adolescent female s who are higher in socioeconomic position, after adjusting for i ndividual race and individua l control variables. IC. The relationship between adolescent race and access to health care will be moderated by adolescent socioeconomic position, so that sexually experienced adolescent females who self -identified racially as Bl ack and who are lower in socioeconomic position will be less likely to have received a routine physical and contraceptive services and more likel y to report unmet health needs than adolescents who are White and higher in socioeconomic posi tion, after adjusting for individual control variables.

PAGE 67

54 ID. There will be variation in the receip t of a routine physical, the receipt of contraceptive services and the report of unmet health needs across neighborhoods. IE. The social disparities in access to health from Hypotheses IA-IC will vary across neighborhoods. Research Objective II: To explore the rela tionships between the neighborhood racial and socioeconomic context and the average odds of access to health care among the sexually experienced adolescent females in this study as the extent to which the neighborhood racial and socioeconomic context contribute s to the variation in access to health care across neighborhoods. IIA. The neighborhood racial context will be a ssociated with access to health care as the average odds of having received a routine physical and having received contraceptive services will be lower wh ile the average odds of reporting unmet health needs will be higher in those neighborhoods characterized by a higher proportion of Black residents than thos e neighborhoods characterized by a lower proportion of Black residents. IIB. The neighborhood socioeconomic context will be associated with access to health care as the average odds of having received a routine physical and having received contraceptive services will be lower while the average odds of reporting unmet health needs will be higher in those neighborhoods characterized by a higher proportion of poor or lower edu cated residents than those neighborhoods characterized by a lower proportion poor or lower educated residents.

PAGE 68

55 IIC. The relationship between the nei ghborhood racial context and the average odds of access to health will be mode rated by the neighborhood socioeconomic context in that the averag e odds of having received a routine physical and having received contraceptive services will be lower while the average odds of reporting unmet health needs will be higher in those neighborhoods that are characterized by a higher proportion of Black and poor or lo wer educated residents than in those neighborhoods that are characterized by a lower proportion of Black and poor or lower educated residents. IID. The neighborhood racial and socioeconom ic context will contribute to the variation in the average odds of acces s to health care across neighborhoods. Research Objective III: To explore the extent to which the neighborhood racial and socioeconomic context moderates the re lationship between individual race, socioeconomic position and access to health care among sexually experienced adolescent females. IIIA1. Adolescents who live in neig hborhoods characterized by a higher proportion of Black residents will be less likely to have received a routine physical or contraceptive se rvices and more likely to report unmet health needs than those adolescents who live in a neighborhood characterized by lower proportions of Black resident s but the negative effect of living in a neighborhood characterized by a higher pr oportion of Black resident s will be greatest among Black adolescents, after adjusting for the individual control va riables, individual

PAGE 69

56 socioeconomic position, neighborhood socioeconomic context and the neighborhood control variables. IIIA2. Adolescents who live in a ne ighborhood characterized by a higher proportion of Black residents will be less likely to have received a routine physical or contraceptive se rvices and more likely to report unmet health needs than those adolescents who live in a neighborhood charac terized by a lower proportion of Black resident s but the negative effect of living in a neighborhood characterized by a higher pr oportion of Black resident s will be greatest among adolescents who are of lower socioec onomic position, after adjusting for the neighborhood racial context and th e neighborhood control variables. IIIB1. Adolescents who live in a neighborhood lower in socioeconomic position will be less likely to have received a routine physical and contraceptive services and more likely to report unmet health need s than those adolescents who live in a neighborhood higher in socioeconomic position but the negative effect of living in a neighborhood lower in socioeconomic position will be greatest among Black adolescents, after adjusting for the neighborhood control variables. IIIB2. Adolescents who live in a neighborhood lower in socioeconomic position will be less likely to have received a routine physical and contraceptive services and more likely to report unmet health need s than those adolescents who live in a neighborhood higher in socioeconomic position but the negative effect of living in a neighborhood lower in socioeconomic position will be greatest among adolescents who are of lower socioeconomic position.

PAGE 70

57 IIIC. The cross-level relationships between IIIA and IIIB will vary across neighborhoods. Study Design This study employed a nonexperimental and multilevel research design via secondary data analyses. The study design was n on-experimental as the participants were not randomized into their neighborhoods, thus alternative explan ations for possible associations may exist. Furthermore, the study was primarily cross-sectional in nature as Level I data, which included the individual sociodemographics, indi vidual covariates, and dependent variables were all measured at th e same time point. There is some temporal sequencing of events for adolescents’ expos ure to their neighbor hoods, as the Level II data were compiled by the Add Health resear chers from the Census of Population and Housing, 1990: Summary Tape File 3A (S TF 3A) while the measurement of the dependent variable occurre d between 1994 to 1995, and only adolescents who lived in their current neighborhood for at least one year were included in this study. The multilevel design of this study enable d the examination of the impact of multiple contexts on access to health car e (Diez-Roux, 2003; Hox, 1995; Subramanian et al., 2003). Specifically, this study examined the extent to which racial and socioeconomic disparities in access to health car e existed at the individual leve l, the extent to which there was variation in access to care between nei ghborhoods and if the neighborhood racial and socioeconomic composition accounted for this variation, and the extent to which the neighborhood racial and socioeconomic com position moderated the relationship between individual race, individual socioeconomic pos ition and access to health care. In addition, multilevel modeling adjusts for the violation of the assumption of independent

PAGE 71

58 observations, which is inherent to nested data (Diez-Roux, 2003; Hox, 1995; Subramanian et al., 2003). Using standard statis tical procedures in th is situation may bias the standard errors downward and potentially lead to false positive results (Diez-Roux, 2003; Hox, 1995; Subramanian et al., 2003). Add Health Study Design and Sampling Procedures This study utilized secondary data from Wave I of the Add Health study. Add Health is a nationally representative long itudinal study in which data were collected across multiple contexts and time points for th e purpose of assessing how these different contexts shape the health and behaviors of adolescents (Harris et al., 2003). Sampling for the Add Health study was based on a clustered design with schools as the primary cluster. The target population included students in 7th through 12th grade. The Add Health sampling frame was a database collected by an outside agency (Quality Education Data, Inc.). From this, researchers randomly sample d 80 high schools that met the criteria of having more than 30 students enrolled and included an 11th grade. The sample was then stratified into 80 clusters based on region, urbanicity, sc hool type, school size, ethnic distribution, grade span and curriculum. More than 70% of the eligible high schools agreed to participate. A replacement school was selected within the same stratum for those schools that refused to participate in the st udy (Harris et al. 2003). The selected high schools assisted the researchers in iden tifying their feeder schools (those schools that included a 7th grade and sent their gra duates to that particular high school). From these potential feeder sc hools, one school was se lected based on the number of students they se nt to the high school. A few of the sampled high schools

PAGE 72

59 contained grades 7th-12th so these schools functioned as the feeder and high school combined. A total of 134 schools were recr uited for the study (Harris et al. 2003). An in-school questionnaire was admini stered to 90,118 students in the 7th-12th grade in 1994-1995. The student questionna ire was self-administered and took approximately 45-60 minutes to complete. No make-up days were provided for the inschool administration. Parents were informed in advance of the survey’s administration and active or passive parent al consent was obtained for participation based upon the school’s requirements (Harris et al. 2003). All students who were listed on the school roster or w ho completed the in-school questionnaire were eligible for selection into the in-home phase of the study. A total of 27,000 students, which consisted of a core sample as well as over-samples, were selected for participation in the in-home interview. Students were chosen for the over-samples based on their responses to the in-school quest ionnaire and consisted primarily of those students from different ethnicities, disabled students, t hose adolescents re siding with one another, and saturated school s (Harris et al. 2003). The Wave I In-Home sample contained 12,105 adolescents from the core sample and an additional 9,783 from the over-sampled groups. Wave I interviews were conducted in 1994-1995 and all data were reco rded on a laptop computer to protect confidentiality. Adolescents were asked about their attitudes, beha viors, relationships with others, school performance, sociodemograp hics, health, health care access, plans for the future among others. The interviewer read less sensitive items to the adolescents and then entered their responses into the comput er. However for more sensitive topics, the adolescent listened to the questions on an audio headset and then independently entered

PAGE 73

60 their responses directly into the computer. Students were also administered the Add Health Picture Vocabulary Test, which is a computerized version of the Peabody Picture Vocabulary Test (Har ris et al. 2003). The Parent Questionnaire was administered during Wave I to a parent or guardian (preferably the mother) of the youth who participated in the in-home interview (n=17,700). The interviewer was available to a ssist the parent if needed. The parent questionnaire contained items related to hous ehold sociodemographics, parental attitudes and behaviors, and parental and adolescent health hist ory (Harris et al. 2003). A School Administrator Qu estionnaire was self-adminis tered during Wave I to administrators of the particip ating schools. The questionnaire contained items related to the school environment, includ ing the provision of health ca re services (Harris et al. 2003). The Contextual Database for Wave I was derived from multiple administrative data sources for the assessment of selected community characteristics. The data were calculated and compiled by the Add Health resear chers into the database and are linked to each adolescent’s participant identification number. Addresses were obtained from the majority of respondents who completed th e In-Home Interview and when possible, residential locations were geocoded to link th em to their state, county, tract and block group Census areas. If the address was unabl e to be geocoded, variables at the block group and census tract were set to missing. No geocodes are available to researchers outside of Add Health. Most of the variables in the contextu al database are available at the level of the county and the only data that is available at the leve l of the block group or census tract are variables that were derived from the Census of Population and Housing,

PAGE 74

61 1990: Summary Tape File 3A (STF 3A). Resear chers are only able to obtain information from the Contextual Database that pertains to the participant. In a ddition, researchers are unable to access data from con tiguous block groups or census tr acts (Harris et al. 2003). The second wave of Add Health data wa s collected in 1996. The sample for the Wave II component was composed of those st udents who completed the Wave I In-Home Interview, excluding those who were in the 12th grade, part of a genetic sample, or disabled at Wave I. The Wave II in-home sample was composed of 14, 738 adolescents who were interviewed in a similar fash ion as Wave I (Harris et al. 2003). In addition, another School Administrator Questionnair e was collected from the administrators of participating schools. The self-administered ques tionnaire was similar to that of Wave I and contained items rela ted to the school environment (Harris et al. 2003). The Contextual Database for Wave II is th e exact same data as that of Wave I. The addresses of participants who complete d the In-Home Interview at Wave II were geocoded when possible to match participants ’ addresses at Wave II to the contextual database. If the addresses were unable to be geocoded, variables at the block group and census tract were set to missing. Thus, th e contextual data differ only for those participants who moved to a different geogr aphic unit between Wave I and II (Harris et al. 2003). Study Sample The sampling frame for this study consis ted of Non-Hispanic White and NonHispanic Black sexually active females between 15 to 19 years of age who participated in Wave I of the Add Health In-Home Interview. In order to conduct reliable analyses, only

PAGE 75

62 those adolescents who had sample weights a nd the neighborhood identifier to link the InHome data to the Contextual data were in cluded in the sampling frame. The following inclusion criteria were imposed on the samp ling frame for this study: (1) never married, (2) lived in current residence greater than one year, (3) only one randomly sampled sibling, and (4) complete data on the variables of interest related to this study. Power Statistical power can be defined as the pr obability of rejecti ng the null hypothesis, which states that there is no statistical relati onship between the variables of interest when a true relationship exists (Cohen, 1992; Tabachnick & Fidell, 2001). Typically researchers desire the power to be between 0.80-0.90 (Cohen, 1992; Raudenbush, Spybrook, Liu, & Congdon, 2005). In hierarchical generalize d linear modeling (HGLM), power is affected by sample sizes across all levels of analysis as well as by the effect size (R audenbush et al., 2005). Thus for this study, power was a function of the number of individuals per neighborhood, the number of neighborhoods, and the strength of the relationship between the individual and neighborhood variables of in terest and access to health care (effect size). There is considerable discussion throughout the research literature regarding the sample size requirement for both the individu al and group samples to establish adequate statistical power. Hierarchical linear m odeling techniques, in cluding HGLM, employ maximum likelihood procedures to pool inform ation from all the groups to estimate the parameters (Ewing, Schmid, Killingswort h, Zlot, & Raudenbush, 2003; Kreft, 1996; Raudenbush, 1997; Raudenbush et al., 2005). Th us, all groups contribute to the estimation of the model, but those groups that have a greater number of individuals will

PAGE 76

63 contribute more to the model than those with smaller numbers of individuals (Ewing et al., 2003; Kreft, 1996; Raudenbush, 1997; Raude nbush et al., 2005). Due to this pooling of information, researchers recommend that even small groups or those with only one individual in the group should remain in the model as the information these small groups provide will contribute to the overall estimati on of the model (Ewing et al., 2003; Kreft, 1996; Raudenbush, 1997; Raudenbush et al., 2005; Snijders & Bosker, 1999) as long as there is variation in th e variables modeled (Snijders & Bosker, 1999). Based on simulation studies, researchers tend to agree th at number of groups has more influence on power, the estimation of the fixed paramete rs, and the size of the standard errors, particularly if estimating cross-level interac tions, than does the numbe r of participants per group (Kreft, 1996; Raudenbush, 1997; Raudenbus h et al., 2005; Snijders & Bosker, 1999). The number of parameters that can be estimat ed at either level is also impacted by the sample size. Consequently, studies that have too few indivi duals per group may be limited in the number of Level I fixed and ra ndom effects that can be examined while those studies with too few groups may be lim ited in the number of Level II effects that can be examined (Raudenbush, 1997; Raudenbush et al., 2005; Snijders & Bosker, 1999). Typically, researchers are c onstrained more by the number of groups they have in the study than by the number of individuals, because data collecti on costs generally are more a function of the number of groups (Raudenbush, 1997; Raudenbush et al., 2005). In contrast, the Add Health data includes a large number of groups with fewer individuals per group, because the sampling for Add Health took place at the level of the school, rather than the level of the neighborhood. A total of 132 schools and 20,745 adolescents

PAGE 77

64 were sampled for Wave I of the In-Hom e portion of the study (Add Health). Since neighborhoods cluster within schools, there ar e more neighborhoods (c ensus tracts) in the Add Health study than there are schools. Sp ecifically, the 20,745 adolescents in Wave I of the In-Home Interview lived across 2,602 census tracts, which results in a large number of groups with fewe r individuals per group. This study examined a sub-sample of the overall Add Health sample since the population of interest consisted of those se xually active adolescent females between 15 years to 19 years of age. The sample for this study had a total of 1,526 adolescents dispersed across 546 census tracts. The numbe r of adolescents per neighborhood ranged from 1 to 40, in which 306 neighborhoods had only one adolescent living in the neighborhood, 153 neighborhoods had 2-4 adoles cents living in th e neighborhood, and 87 neighborhoods had 5 or more adolescents per neighborhood. Power analysis was conducted using the Optimal Design for Multilevel and Longitudinal Research Software, Vers ion 1.55 (Liu, Spybrook, Congdon, & Raudenbush, 2005) in order to examine the impact of samp le size at both levels of analysis on the power of this study. The type of design se lected to run the power analysis was a clustered, randomized design for binary outco mes. The alpha was set to 0.05. Figure 2 depicts the findings form the power anal ysis. Even though this study has 546 groups, power analysis was conducted for only the 240 groups that had 2 or more persons in the group, as multilevel modeling requires at least 2 observations per group (Liu et al., 2005). The analysis indicated that for 240 groups with a range of 2 40 adolescents per group, the power was 0.985 or greater, depending on th e size of the cluster. Thus, there was adequate power to detect fixed eff ects at both levels of the model.

PAGE 78

65 Figure 2. Statistical Power of Sample Cluster size P o w e r 9 16 23 30 37 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 = 0.050 E = 0.600000 C = 0.400000 lower plausible value = 0.30000 0 upper plausible value = 0.50000 0 J = 2 4 0 Measures This study utilizes measures from Wave I of Add Health’s restricted use dataset. The data sources include: the Adolescent In -Home Interview, the Parent Questionnaire, the School Administrator Survey, and the Cont extual Database. The items used for this study are summarized in Appendix A. Dependent Variables Access to Health Care The dependent variables examined in th is study were three measures of access to health care. The items measuring these vari ables were obtained from Wave I of the Adolescent In-Home Interview. Two of the measures pertaine d to the receipt of health care services, including the receipt of a routin e physical and the receipt of contraceptive services. The adolescents were asked during the interview: (1) In the past year have you

PAGE 79

66 had a routine physical examination? and (2) Have you ever received a birth control method from a doctor or clinic? Both variables were dichotomous in nature with response options consisting of “yes” or “no”. Utilization or actual rece ipt of health care services is a commonly used measure of access to health care (USDHHS, 2003). These two specific measures were chosen as they ar e both examples of pr eventive health care services that according to clinical practice guidelines should include safer sex counseling and STI screening of sexually active a dolescent females (AAP, 2000; AMA, 1997). Unfortunately health care provide rs do not always follow clin ical practice guidelines and they may fail to provide these recommended services (Burstein, Lowry, Klein, & Santelli, 2003). However, for the purposes of this study th e receipt of a routin e physical exam and contraceptive services indicates that entr y into the system was achieved, although the quality of the services received could not be ascertained. The third dependent variable measuring access to health care was the subjective report of unmet health needs. The adolescent was asked: Has there been any time over the past year when you thought you shoul d get medical care, but you did not The variable was dichotomous with resp onse options of “yes” or “no”. Independent Variables Adolescents’ Sociodemographic Characteristics The measures for individual sociodemographi cs were derived from Wave I of the Adolescent In-Home Interview and the Parent Questionnaire. The measure for race was composed of three questions derived from the Adolescent In-Home Interview: (1) What is your race? with response options consisting of White, Black or African American, American Indian or Native American, Asian or Pacific Islander, other, or don’t know, (2)

PAGE 80

67 if adolescents selected more than one racial category, they were asked: Which one category best describes your racial category? with response options identical to the first question, and (3) Are you of Hispanic or Latino origin? with yes or no response options. Racial categories were deve loped for Non-Hispanic White (reference) and Non-Hispanic Black race. Individual socioeconomic position was m easured via the household income to poverty ratio and parental educational attain ment. The variable for household income was obtained from the Parent Questionnaire, in which the parent or guardian who completed the questionnaire was asked: About how much total income, before taxes did your family receive in 1994? Include your own income, the income from everyone else in your household, and income from welfare bene fits, dividends, and all other sources. The response option was continuous and reported in thousands. Household income was examined as a c ontinuous variable and as the ratio of household income to the 1994 federal poverty th reshold, adjusted for household size (0100%, 101-200%, 201-300% 301-400%, and >400%). The ratio of income to poverty was based on data related to household si ze obtained from the Adolescent In-Home Interview while household income was obtained from the Parent Questionnaire. Household size was calculated based on adol escents’ yes or no response to 20 items regarding the household structur e. The rates from the U.S. Census Bureau’s 1994 Poverty Threshold, adjusted for family size, were then compared to adolescents’ household income based on household size to create a catego rical variable that in dicates the range of income to poverty ratio (0-100%, 101200%, 201-300% 301-400%, and >400%) for each adolescent.

PAGE 81

68 Parental education was measured as level of parental educational attainment and was derived from both the Parent Questionnai re and the Adolescent In-Home Interview. Parents were asked: How far did you go in school? with response options including: (a) 8th grade or less, (b) more than 8th grade, but did not graduate from high school, (c) went to a business, trade, or vocational school instead of high school, (d) high school graduate, (e) completed a GED, (f) went to a business, trade or vocational sc hool after high school, (g) went to college, but did not graduate, (h) graduated from a college or university, and (i) professional training beyond a 4-year college or universit y. If parents were missing on this response or if they refused to answer, the adolescent’s response to a similar question was imputed. The adolescent was aske d during the In-Home Interview: How far in school did she or he (mom or dad) go? The response options were identical to the parents’ response options, except that th ey also included an option th at the adolescent knew her mom went to school, but unsure of the level attained and an option for “don’t know”. The imputation of the adolescent’s response fo r missing parental data has been used in previous studies examining Add Health data and reported to consistently correlate with parental responses (Cubbin, Santelli, Brindis, & Braveman, 2005). The adolescent’s response to maternal education was imputed first since the majority of parental particip ants in the Add Health study were mothers or mother figures by virtue of the original study design. If the adolescent was missing a response to maternal education, then if available, the adolescent’s response related to paternal education was imputed. The multiple respons es for parental education were then categorized into the following four categor ies: (a) less than a high school degree or

PAGE 82

69 equivalent, (2) high school degr ee or GED, (3) some college or technical training, and (4) college degree or more. Adolescent Control Variables The adolescent control variables for th is study were derived from the Adolescent In-Home Interview, the Parent Questionnaire, and the School Administrator Survey. The control variables were selected based on the research literature a nd several conceptual frameworks related to access to and utilization of health care (Aday, 2001; Aday et al., 2004; Andersen 1995; Davidson, Andersen, Wyn, & Brown, 2004; Shi & Stevens, 2005). Three overall constructs from these framewor ks, health needs, pred isposing factors, and enabling factors guided the selection of the control variables analyzed in this study (Aday, 2001; Aday et al., 2004; Andersen 1995; Davidson et al., 2004; Shi & Stevens, 2005). Acute and chronic health needs are one of the most frequent reasons why individuals seek and utilize heal th care services, predisposi ng factors, such as health beliefs and sociodemographics, may influe nce the propensity to use services, and enabling factors, such as income, insurance, and having a regular provider, facilitate access to and utilization of h ealth care services (Aday, 2001; Aday et al., 2004; Andersen 1995; Davidson et al., 2004; Shi & Stevens, 2005). The predisposing factors that were examin ed with this study included adolescents’ age, three health belief variables – adolescen ts’ beliefs regarding the consequences of pregnancy, their beliefs regarding barriers in accessing and using birth control, and their perceptions related to parental disapproval regarding adolesce nt sexual activity and use of birth control, and the length of time lived in current residence. The first fours variables were derived from items in the Adolescent In -Home Interview while the last variable was

PAGE 83

70 derived from items in both the Adolescent In-Home Interview as well as the Parent Questionnaire. Age was a calculated variable based on an equation provided by the Add Health researchers in which adolescents’ month and ye ar of their date of birth was subtracted from the month and year of the In-Home In terview. Age was measured as a continuous variable. Adolescents’ beliefs regarding the conseque nces of pregnancy were derived from the following 5 items: (1) Getting pregnant at this time in your life is one of the worst things that could happen to you (2) If you got pregnant, it wo uld be embarrassing for your family (3) If you got pregnant, it would be embarrassing for you (4) If you got pregnant, you would be forced to grow up too fast and (5) If you got pregnant, you have had to decide whether or not to have the baby and that would be stressful and difficult The response options for the items included st rongly agree, agree, neither agree nor disagree, disagree, and strongl y disagree. The items were re verse coded so that higher scores indicated that adolescents’ were more likely to agree that pr egnancy could result in these undesirable consequences. Exploratory fa ctor analysis and reliability testing via internal consistency were examined. All five of the items loaded ont o one factor with all factor loadings greater than 0.4. A composite of the items was created based on the mean of the 5 items. The internal consistency reliability of the composite is = 0.76. Two other items related to adolescents’ perceptions of pregnancy consequences were examined for potential inclusion in the composite: (1) If you got pregnant, you would have to quit school and (2) If you got pregnant, you might have to marry the wrong person, just to get married However, these two items were deleted from the

PAGE 84

71 analysis due to low and non-signi ficant factor loadings (< 0. 4). Perhaps these items were less likely to represent a consequence of pregnancy for these adolescents since childbearing outside of marriage was quite hi gh in the early 1990’s (V entura & Bachrach, 2000) and many schools now have programs for pregnant teens to stay in school to complete their education. Adolescents’ beliefs regarding barriers to accessing and using birth control were derived from the following 6 items: (1) In general, birth control is too much of a hassle to use (2) In general, birth control is too expensive to buy (3) It takes too much planning ahead of time to have birth control on hand when you’re going to have sex (4) It is too hard to get a boy to use birth control with you (5) For you, using birth control interferes with your sexual enjoyment and (6) Using birth control is morally wrong The response options for the items included strongly agree, agree, neither agree nor disagree, disagree, and strongly disagree. The items were reverse coded so that higher sc ores indicated that adolescents’ were more likely to agree with the items. Exploratory factor analysis and reliability testing via internal consistency we re examined. All six of the items loaded onto one factor with all factor loadings greater than 0.4. A composite of the items was created based on the mean of the 6 items. The internal consistency reliability of the composite is = 0.79. Two additional predisposing factors included in the analysis were related to adolescents’ perceptions of their mother’s disapproval of adoles cent sexual activity and use of birth control. If th e adolescent reported having a mo ther figure, she was asked by the interviewer: (1) How would she (mom) feel about your having sex at this time in your life? and (2) How would she feel about your using birt h control at this time in your life?

PAGE 85

72 The response options for both items include strongly disapprove, disapprove, neither disapprove or approve, approve and strongly approve. The ite ms were reverse coded so that higher numbers represented higher levels of maternal disapprova l. Correlations were examined to determine if the two items could be measured as one construct, but only a moderate correlation was found (r =0.61), thus th e two items were analyzed individually. The last predisposing factor included in this study was the length of time the adolescent lived in her current neighborhood at the time of Wave I data collection. The length of residence variable was constructed from items available in the Adolescent InHome Interview and the Parent Questionnaire. A series of data steps were required to develop this variable. First, adolescents were asked in the Adolescent In-Home Interview: How old were you when you moved here to your current residence? Response options included “lived here since birth or was under one year old when moved” and age in years ranging from 1 year to 19 years of age or older. Second, the number of years the adolescent reported living in the current residence was then subtracted from the adolescent’s age in years. Third, for those adolescents who were missing this response, the parental data regarding th e length of time the child lived in the current residence was imputed when possible. The parents we re asked on the Parent Questionnaire: Has {child’s name} always lived, since (he/she) was born, in the house or apartment building where (he/she) lives now? Response options for parents incl ude yes, no or don’t know. If the parent reported yes, then the number of years the adolescent lived in the current residence was equal to the adolescent’s age in years so this data was imputed. If the parent reported that the adoles cent did not always live in th e current resident, he or she was asked on the questionnaire: In what month and year did (h e/she) move to the house

PAGE 86

73 or apartment building (he/she) lives in now? From this, a date was calculated, which was then subtracted from the interview date to represent the number of years the adolescent lived in the current residence. The calculation date that wa s implemented to construct the length of residence variable was similar to the procedure provided by the Add Health researchers to calculate the adolescent’s age in years. The length of years lived in current residence variable was c ontinuous in nature and measured in years. Enabling factors analyzed in this study included health insurance, transportation, family structure, the receipt of school hea lth education regardi ng pregnancy, STI’s, and medical attention, and the availa bility of preventive health care services in the school. The variable for health insurance was derive d from the Parent Questionnaire in which the parent or guardian was asked: What kind of health insurance does {Name} have? Response options included Medi care (from Social Security ), Medicaid, individual or group private coverage (such as Blue Cross or Cigna), prepaid health plan (such as an HMO or Cigna), other, none, or don’t know. Respondents were allowed to select more than one answer. Analyses were completed to create mutually exclusive categories based on type on the type of insurance and then grouped into public (Med icare or Medicaid), private (individual, group, or pr epaid), other, mixed (more than one type of insurance that was not exclusively private, public or other) or no health insurance. The respondents who selected don’t know were treated as missi ng on this response. Due to small cell sizes in several of these categories, health insu rance was dichotomized and dummy coded into yes for those adolescents with health insu rance and no for those adolescents without health insurance.

PAGE 87

74 Transportation was operationalized as adoles cents’ possession of a valid driver’s license, as adolescents possessing a valid driver’s license can potentially drive themselves to their appointment. A valid driver’s license may be even more significant for receipt of contraceptive services as many adolescents seek confidential care. The variable was obtained from the Adolescent In Home Inte rview and was measured by a single item: Do you have a valid driver’s licen se (not a driver’s permit)? The item was dichotomous, as the adolescent either responded, yes or no. The item did have a legitimate skip pattern in which only those adolescents who had answer ed that they had driven a car on the previous question were asked a bout their driver’s license stat us. Therefore, adolescents in the legitimate skip pattern were coded for this study as not having a valid driver’s license since they never drove a car. The third enabling factor included family structure. Adolescents may be more likely to receive preventive health care if they live in a two pa rent household, as two parents are then available to take the adol escent to the health care provide. Family structure was derived from 20 items related to the household structur e. Adolescents were asked during the In-Home Interview to list the names of the persons living with them in their household and the nature of their re lationship with each of these individuals. Analyses were conducted to develop mutually exclusive categories describing maternal and paternal relationships (biological parent step-parent, adoptive parent etc), which were then collapsed into a dichotomous vari able representing those adolescents who lived in a two parent household and those who di d not. The dichotomy was created due to small cell sizes in several of the categories.

PAGE 88

75 The fourth enabling factor included in this study was adoles cents’ receipt of school health education on topi cs related to pregnancy, HI V/AIDS, and where to access medical attention. The interviewer asked a dolescents during the In-Home Interview: Please tell me whether you have learned about ea ch of the following things in a class at school… Topics included pregnancy, AIDS, and where to go for help with a health problem along with numerous others. Response options included yes, no, or don’t know. Adolescents who reported don’t know were set to missing. The topics related to pregnancy and AIDS were combined into one dichotomous item with adolescents who reported receiving at least one of the edu cational topics coded as yes and those who reported not receiving educati on on either one of the topics coded as no. The receipt of health education on where to seek medical attention was analyzed as a single dichotomous item with adolescents who recei ved the education coded as yes and those who did not receive the education coded as no. The last enabling factor included in this study was related to th e availability of health care services in the adolescent’s school. Two types of health care services were constructed – the availa bility of non-athletic physicals and the availabi lity of reproductive health care services. The items measuring the availability of these select school health care services were obtained from the School Administrator Survey. The administrator was asked on the survey: For each of the following health-related services, please indicate whether it is provided at your school, is provided by your school district but not at your school, referred to other provid ers, or neither provided nor referred Response options included non-athletic physical, STI tr eatment, and family planning services along with several others. The three items were di chotomized and coded as yes if the school

PAGE 89

76 administrator reported the serv ice was offered on-site at th e school or no if the service was not offered on the site. The two items, availability of STI treatment and family planning services, were combined to repres ent one variable – the availability of reproductive health care services. The items were then disaggre gated from the level of the school to the level of the indi vidual, thus the availability of school reproductive health care services and the availability of non-athl etic physicals represen t whether or not the adolescent attends a school in which th ese health care services are offered. The health needs that were examined in this study include d the adolescent’s subjective report of her genera l health and a history of a STI or pregnancy. The three items were derived from the Adolescent In-Home Interview. The item to represent measuring adolescents’ general health st atus was obtained from the question: In general, how is your health? Would you say…, with response options cons isting of excellent, very good, good, fair, and poor. Analyses were conducte d to reverse code the items so that higher scores represent better health. The second health need to be examined with this study was adolescents’ history of a STI, which was derived from several items related to adolescents’ previous STI diagnoses. Adolescents were asked by the interviewer: Have you ever been told by a doctor or a nurse that you had…chlamydia, syphilis, gonorrhea, HIV or AIDS, genital herpes, genital warts, or trichomoniasis? The response option to each of the infections was either yes or no. A dichotomous variable was created from these items to indicate if the adolescent had a history of any one of these infections. Thus, if the adolescent reported a history of one of these infections, the variable, “history of STI”, was coded as

PAGE 90

77 yes and if the adolescent reported never bei ng diagnosed with a STI, the variable was coded as no. The last health need to be included in the analysis was adolescents’ history of ever being pregnant. Adolescents we re asked by the interviewer: Have you ever been pregnant? Be sure to include if you are cu rrently pregnant and any past pregnancy that ended in an abortion, stillbirth, miscarriage, or a live birth after which the baby died The response options included yes or no. Neighborhood Racial and Socioeconomic Context The neighborhood racial and socioeconomi c composition variables were derived from items in the Add Health Contextual Database, which contains preconstructed variables that the Add Health researchers derived from th e Census of Population and Housing, 1990: Summary Tape File 3A (S TF 3A) (Harris et al., 2003). This study conceptually defined the nei ghborhood as a spatial unit of re sidence and operationalized the neighborhood as the census tr act of residence. Census tr acts contain approximately 2,500 and 8,000 people (U.S. Census Bureau, 1994) The census tract was selected to represent the neighborhood sin ce they are often considered proxies for neighborhoods in the research literature (Lev enthal & Brooks-Gunn, 2000). In a ddition, census tracts are considered “administrative units used by fede ral, state, and local governments, including public health departments, to characterize ju risdictions, determine eligibility for diverse programs, and allocate resources” (Subramania n et al. 2005, p. 260), thus the census tract seemed the most appropriate geographic area fo r considering the potential availability of health care resources. Unfortunately, no data is available in the Add Health Contextual database regarding the number of health care providers at the level of the census tract.

PAGE 91

78 Neighborhood racial composition was measured as the proportion of Black residents living within each census tract. The proportion of Bl ack residents was selected to represent the neighborhood racial compositi on since Blacks are the most residentially segregated and discriminated social gr oup in the United States (Charles, 2003). Furthermore, the proportion of Black residents has been used in ot her studies measuring neighborhood racial composition and access to health promoting resources, such as playgrounds (Cradock et al., 2005), healthy f ood restaurants (Lewis et al., 2005), and supermarkets (Zenk et al., 2005). Neighborhood racial composition was measured as a continuous variable for all an alyses– the proportion of Black residents living within the neighborhood, and as a categorical variable fo r descriptive analyses with the categories similar to those reported in the literature (Rawlin gs, Harris, Turner, & Padilla, 2004). The categories for the neighborhood racial compositi on included: 10% or less Black residents living within the census tract, 11% to 50% Bl ack residents living w ithin the census tract, 51% to 90% Black residents living within the census tract, and 90% or more Black residents living within the census tract. Neighborhood socioeconomic composition was measured with two indicators (1) neighborhood poverty, which was measured as the proportion of families in the census tract with income below th e 1989 poverty level, and (2 ) low education, which is measured as the proportion of adults over the ag e of 25 years in the census tract with less than a high school degree. Neighborhood poverty and low education were analyzed as continuous variables in all anal yses and as categorical variab les for descriptive analyses. Neighborhood poverty is defined in the litera ture as those census tracts in which 20% of the adult residents are below poverty (Krieger et al., n.d.), thus a dichotomous variable

PAGE 92

79 was created to measure thos e census tracts above and below the 20% threshold for families living below the poverty line. An undereducated neighborhood is defined in the literature as those census tracts which 25% of adults who are over 25 years old have less than a high school degree (K rieger et al., n.d.), thus a dichotomous variable was created to measure those census tracts above and below the 25% threshold. Neighborhood Control Variables The neighborhood control measures were derived from the Add Health Contextual Database. The following variables we re selected as control variables: (1) total population, which is measured as the total number of persons per census tract; (2) median age, which is measured as the median ag e of the population in the census tract; (3) urbanicity, which is measured as the proportion in the cens us tract residing within an urban area, and (3) residential stability, whic h is measured as the proportion of persons over 5 years of age living in the same hous e for past 5 years. All variables were continuous in nature. An additional variable measuring neighborhood health needs (child to woman ratio for unmarried women 15-24 years of age in each census tract) was proposed to be included in the analysis, but due to a si gnificant amount of missing data this variable was not incor porated into the study. Data Analysis Multilevel modeling was the primary data analytic technique employed in this study as this statistical method enables the simultaneous examin ation of the two levels of analyses that are of interest to this st udy – neighborhoods and i ndividuals (Diex-Roux, 2003). Consequently, the extent to which individual and neighborhood characteristics independently and interactively influence acce ss to health can be ascertained (Diex Roux,

PAGE 93

80 2003). The ability to model the contributions of more than one level of analysis to the variability in the outcome of interest is the primary reason and advantage for using multilevel modeling techniques over standard statistical procedur es (Diez Roux, 2003; Subramanian et al., 2003). A combination of univariate, bivariat e and multilevel analyses was conducted with this study. Specific software programs to analyze the data included SAS 9.1 (SAS Institute, Inc., Cary, NC) for data manageme nt and univariate and bi variate analyses of neighborhood data1, Sudaan 9.0 (RTI International, Research Triangle Park, NC) for univariate and bivariate analyses of individual data and for bivariate analyses examining individual and neighborhood data together, and HLM 6 (SSI, Inc., Lincolnwood, IL) for multilevel modeling. All analyses that examined individual level data included the grand sample weight that was provided by the Add Health researchers. Data Management Data management consisted of a series of statistical techniques to prepare the data for analysis. First, the Add Health restricted use data set is composed of multiple data sources that required merging prior to analys is. The Parent Questionnaire and Wave I of the Adolescent In-Home Interview were al ready linked together by the Add Health researchers prior to distri bution, thus only the School Ad ministrator Survey and the Contextual Database had to be merged w ith this already combined data set. Second, a series of data management step s were conducted to create the variables of interest for this study, which were outlined in the measurement section of this study. 1 The clustering for the Add Health study is at the level of the school, not the neighborhood. Consequently there are no weights available for the neighborhood data so univariate and bivariate analyses were conducted using standard statistical procedure when neighborhood level data was examined alone.

PAGE 94

81 The overall sample from Add Health was s ubsetted for this study to include only those adolescents who met the following criteria: (a) sexually active female at Wave I, (b) at least 15 years of age at Wave I, (c) Non-Hispanic White and Non-Hispanic Black race, (d) never been married, (e) lived in permanent residence at least one year before Wave I of the Adolescent In-Home Interv iew, (f) had responses to al l variables of interest, and (g) was assigned an individual level sample weight by Add Health researchers. In addition, there were 7 pairs of adolescents ( 14 total) who lived in the same household, thus to avoid violating the assumption of i ndependent observations at the level of the household, one adolescent from the sibling pair was selected via SAS random sampling procedures for inclusion in the study. A variable was created to represent those adolescents who met these inclusion criter ia, which was utilized for univariate and bivariate analyses as a subpopul ation variable. The use of a subpopulation variable, rather than deleting those participants who are not in the sub-sample of interest, prevents violations to the design eff ect and misestimation of the standard errors (Chantala & Tabor, 1999). The HLM software package does not enable the use of a subpopulation variable, so those participants who did not m eet the inclusion criteria were deleted from the sample for multivariate analyses. Third, the proportion of adol escents missing data on all va riables of interest was calculated. Those adolescents who did not have a complete set of data were excluded from the analysis through the use of the subpopulation variable. As described previously, adolescent and parent data were substituted in for one another if the item was missing a response and if there was a para llel question asked to either th e parent or adolescent. This imputation was done for parental education as well as length of residence in the current

PAGE 95

82 neighborhood. Descriptive analyses were conducted to determine how adolescents excluded from the analyses differed from thos e who remained in the study to assess for the representativeness of the final sample. Univariate and Bivariate Analyses Univariate analyses consisted of descript ive statistics to gain an understanding the distribution of the data. Freque ncies, means and standard errors/deviations were analyzed on the variables of interest. In addition, the Level I (adolescent) socioeconomic variables were stratified by Level I (adolescent) race to examine cell sizes for the analysis of the interactions. Bivariate analyses were analyzed to gain a better unders tanding of how the variables of interest are re lated. Odds-ratios were calcula ted between the Level I and Level II predictor variables and the outcome variables measuring access to health care (receipt of routine exam, rece ipt of contraceptive services unmet health needs). Oddsratios were also calculated between a dolescent race and socioeconomic position and adolescent health needs (general health status and history of STI and pregnancy) in order to examine the extent to which social dispar ities in health existe d among the adolescents in this study. Multivariate Analyses Regression diagnostics were conducted in SAS prior to the HGLM analyses to determine if there were any unusual observ ations among the Level I data that could potentially impact model significance. The re gression diagnostics employed for this study included the examination of observations for outliers, leverage and influence. Outliers are defined as those observations in which the value for the dependent variable is

PAGE 96

83 significantly different than e xpected based on the values of the independent variables, leverage measures the distance of an observati on’s value on the independent variable from the mean value for that independen t variable across all the observations, and influence measures the impact on the value of the parameter estimates if an observation is deleted from the analysis (Fox, 1997). For this study, potential outliers were examined via R-studentized residuals – observations with residuals greater than 2 were examined further, leverage was measured via hat scores – observations with scores greater than 2 times the mean value on the independent variable were examined further, and influence was measured via Cook’s D – observations wi th values greater than 1 were examined further. Multicollinearity was also assessed fo r both Level I and Level II independent variables. Multicollinearity occurs when i ndependent variables are so highly correlated with one another they form a near perfect linear relationship (F ox, 1997). This can result in unstable regression coefficients and standard errors, as the unique contribution of each independent variable on the dependent vari able is difficult to ascertain (Fox, 1997). Multicollinearity was examined for this study via the variance inflation factor (VIF), which measures the extent to which the vari ance is inflated due to the high correlation between the independent variables, and tolerance which measures the proportion of unique variance in the independent variab les that is not explained by the other independent variables (Fox, 1997). Tolerance and VIF are interrelated, as tolerance equals 1/VIF (Fox, 1997). For this study, any in dependent variable with a tolerance less than 0.2, which equates to a VIF of 5, was ex amined further to determine if a composite

PAGE 97

84 of the correlated variables w ould be computed or if one of the variables would be excluded from the analysis. Multilevel modeling for nonlin ear distributions (hierarc hical generalized linear modeling-HGLM) was employed for multivariate analyses due to the dichotomous nature of the dependent variables. Binary or dich otomous outcomes are not normally distributed since they can only take on a value of 0 or 1 (Raudenbush & Bryk, 2002). A two level model was examined for this study. Level I of the model contained variables derived from the adolescent and parent data while Le vel II contained variab les derived from the contextual dataset. Since the sample incl uded only one randomly selected adolescent from each household, the adolescent and parent data were analyzed on the same level of analysis and the assumption of independe nt observations was not violated. Hierarchical generalized lin ear modeling is a special t ype of multilevel regression for use with binary outcome variable s (Raudenbush & Bryk, 2002). There are three components to the Level-1 model in HGLM fo r binary outcomes: the sampling model, the link function, and the structural m odel (Raudenbush & Bryk, 2002). The Level2 model in HGLM is similar to the equation used in standard HLM procedures for continuous outcomes (Raudenbush & Bryk, 2002). The sampling model describes the distribu tion of the Level-1 model and is based on the Bernoulli sampling distribution (Raudenbush & Bryk, 2002). The equation for the sampling model is as follows:

PAGE 98

85 Yijij ~ B( ij, ij ) (1) where Yij = the number of successes, ij = the number of trials, ij = the probability of successes. The expected value and variance of Yij are as follows (Raudenbush & Bryk, 2002): E(Yijij ) = ij and Var(Yijij ) = ij (1ij) (2) In order for the Bernoulli sampling model to be used in HGLM, the distribution must be transformed into a linear model, which occurs thr ough the logit link function (Raudenbush & Bryk, 2002). The equation for th e logit link function is as follows: nij = log ij ij 1 (3) where nij =the predicted outcome or the log odds of success. Finally, the Level-1 structural model (Raudenbush & Bryk, 2002): nij = Boj + B1jXi1j + B2jXi2j +…+ BQjXiQj (4) where the result is converted into a predic ted probability in which the value of ij is between zero and one and no error term exists (Raudenbush & Bryk, 2002). The Level-2 model equation is as follows (Raudenbush & Bryk, 2002): oj oju B 00 (5) where ojB = is the group mean in the outcome for group j 00 = intercept or the mean outcome across all groups, and oju is the residual for each group. Note the Level-2 model has no predictors in the model at this time. Centering of continuous variables wa s employed for the multilevel analyses. Centering continuous variables is particularly useful in HLM since we are also interested in interpreting the intercep t (Raudenbush & Bryk, 2002). Researchers can grand mean

PAGE 99

86 center the variable so that the variable is centered based on the mean score of that variable across all the groups in the study or they can group mean center so that the variable is centered based on the mean score within each group (Raudenbush & Bryk, 2002). For the multilevel analyses in this study, the Level I and Level II continuous variables were grand mean centered and dichot omous variables at bot h levels were left uncentered. A series of nested models were analy zed for each hypothesis of this study. After each step, the fit of the model was ex amined via maximum likelihood estimation procedures to determine if th e addition of the pred ictors statistically improved model and explained more of the variance in the out come. Table 10 in Chapter 4 summarizes the model building process that was employed to test the hypotheses of this study. Study Limitations Several limitations exist related to the desi gn of this study. First, due to the crosssectional nature of the study and lack of randomization of adolescents into neighborhoods, causal relationships between the independent and dependent variables cannot be ascertained. Although researcher s have employed quasi-experimental community designs in which families living in impoverished neighborhoods are randomly assigned to live in nei ghborhoods with different levels of social advantage, this type of design was beyond the scope of this study. This study was limited to only those adolescents who lived in their neighborhood for at least one y ear so that exposure to the neighborhood conditions was at least one year or more. Second, the inclusion criteria applied to the study limit the ex ternal validity. For example, the study sample is limited to Non-Hispanic White and Non-Hispanic Black

PAGE 100

87 sexually experienced adolescent females who were between 15 years to 19 years of age, who lived in their census tract for at least one year before participating in the In-Home Interview, who were never married, and w ho had complete data on the variables of interest. However, sexually experienced adol escent females are a particularly vulnerable population at high risk for unintended pr egnancy and STI’s, including HIV, thus investigation of their access to and utiliz ation of health care is essential. Third, the measures used for this study we re restricted to those items available in the Add Health data set. Residential segrega tion indices could not be calculated due to the lack of information on contiguous census tr acts in the Add Health Contextual dataset. Consequently, the racial and socioec onomic composition of the neighborhood was measured instead. Although this measure commonly serves as a proxy for residential segregation throughout the literature, they ar e conceptually and operationally different than measures of segregati on (Acevedo-Garcia & Lochner, 2003; Acevedo-Garcia et al., 2003). In addition, the Add Health data set c ontains no information regarding the health care encounter, which is an important compone nt of access to health care. Finally, the items on the surveys and interviews contain set response options, which may restrict the answers that individuals can give to the questions. Fourth, the items are based upon self-repor t from parents and adolescents. Selfreport measures can be biased due to so cial desirability. In addition, many of the sexuality related questions that were asked to adolescents could be considered sensitive in nature and adolescents may have been fearful to answer correctly due to confidentiality concerns. The researchers tried to minimize th ese concerns during data collection as all

PAGE 101

88 sensitive questions were asked to the adol escent through an audio head set upon which they entered their responses directly into the computer. Finally, communities themselves can under go change due to residential mobility, immigration, and the availabili ty of resources. For example, the census tract data were collected in 1990, but the indivi dual data were collected in 1994. During this timeframe, the racial and socioeconomic composition, the social context, and the resource availability in the census tract could have ch anged. Furthermore, the data were collected over ten years ago and considerable change s in the sociodemographic composition as well as the health care system have occurred in the United States during this time, which could limit the external validity of this st udy. The paucity of nationally representative data that enables examination of neighborhood conditions and adolescent health restricts the use of more timely data for this study. Consequently, numerous researchers continue to conduct cross-sectional analyses of the cont extual and adolescent data from Wave I of Add Health (Cubbin et al., 2005; Scher, 2004; Wickrama & Bryant, 2003).

PAGE 102

89 Chapter Four: Results Study Sample The sampling frame for this study consis ted of Non-Hispanic White and NonHispanic Black sexually active females between 15 to 19 years of age who participated in Wave I of the Add Health In-Home Interv iew. The following inclusion criteria were imposed on the sampling frame for this study : (1) never married, (2) lived in current residence greater than one year, (3) only one randomly sampled sibling, and (4) complete data on the variables of interest related to this study. Table 1 summarizes the impact of the inclusion criteria on the size of the sampling frame, including the number and proportion of adolescents lost due to the inclusion criteria. Table 1 Impact of Inclusion Criteria on Size of Study Sample Inclusion Criteria Beginning Sample Size Ending Sample Size N Lost Proportion Lost Never Married 2531 2493 38 1.5 Lived in Residence 1 Year 2493 2231 262 10.51 Only One Randomly Sampled Sibling 2231 2224 7 0.31 Complete Data 2224 1526 698 31.38 As depicted in Table 1, the size of the sampling frame was N = 2,531 and the final sample size for this study was N = 1,526. The greatest impact of the inclusion criteria on the sample size was the proportion of particip ants who were excluded because they had not lived at their residence for at least one ye ar at the time of Wave I (10.51%) or because

PAGE 103

90 they were missing data on the variables of interest (31.38%). The former was imposed as an inclusion criterion, because adolescents were asked in Wave I if they had received a routine physical or had any unmet health need s during the year pre ceding the interview. Thus, in order for exposure to the nei ghborhood environment to occur prior to the outcome variable, adolescents needed to have lived in the same neighborhood for at least one year. Of the 698 participants who were exclude d due to missing data, the majority of exclusions were due to omissions in pare ntal reports on household income (80.37%) and adolescents’ health insurance status ( 52.72%). Adolescents had considerably fewer missing responses than parents, which may be due to different data collection methods. Specifically, adolescent data were collected via face-to-face interviews for non-sensitive items and self-administered for those more sensitive topics. All parental data were collected via self-administered surveys. Th e majority of missing data for adolescents were on the items regarding maternal disa pproval of adolescent sexual activity (19.77%) and birth control use (20.05%), many of wh ich were missing due to the absence of a mother figure in the household. Although the researchers had intende d to ask parallel questions to the adolescents regarding paternal disapproval of adolescent sexual activity and birth control, errors in data collection prevented this information from being obtained (Harris et al., 2003). Descriptive analyses were examined on the full study sample prior to the exclusion of those participants who were missing data ( N = 2,224) and on the final study sample after those participants who were missing were excluded ( N = 1,526). As summarized in Table 2, the mean proportions of access to health care as well as the

PAGE 104

91 individual and neighborhood sociodemographic ch aracteristics were similar between the two samples. The difference in individual income between the two samples was not assessed due to the high rates of missing data on this variable. Table 2 Descriptive Analyses of St udy Sample Prior to and Afte r Exclusion of Participants Missing Data (N = 2,224 vs. N = 1,526) Study Sample Prior to Exclusion of Participants Missing Data N =2,224 Final Study Sample After Exclusion of Participants Missing Data N =1,526 Mean SEa Range Mean SE Range Dependent Variables Received Routine Physical 0.67 0.02 0-1 0.68 0.01 0-1 Reported Unmet Health Needs 0.23 0.01 0-1 0.26 0.01 0-1 Received Contraceptive Services 0.49 0.02 0-1 0.51 0.01 0-1 Level I Independent Variables Race Black 0.22 0.04 0-1 0.24 0.04 0-1 White 0.78 0.04 0-1 0.76 0.04 0-1 Parental Education Less than High School 0.15 0.02 0-1 0.16 0.02 0-1 High School Degree or GED 0.35 0.02 0-1 0.36 0.02 0-1 Some College/Technical 0.30 0.02 0-1 0.28 0.02 0-1 College Degree or More 0.20 0.02 0-1 0.19 0.02 0-1 Mean Std. Dev.a Range Mean Std. Dev. Range Level II Independent Variables Proportion of Black Residents 10% or less (reference) 0.62 0.49 0-1 0.61 0.49 0-1 11% 50% 0.20 0.40 0-1 0.20 0.49 0-1 51% 90% 0.11 0.32 0-1 0.11 0.40 0-1 91% or more 0.07 0.26 0-1 0.08 0.27 0-1 Total Proportion of Black Reside nts 0.21 0.31 0-1 0.22 0.31 0-1 Proportion of Poor Residents 20% 0.77 0.42 0-1 0.77 0.42 0-1 > 20% 0.23 0.42 0-1 0.23 0.42 0-1 Total Proportion of Poor Residents 0.12 0.11 0-1 0.12 0.31 0-1 Proportion of Low Educated Residents 25% 0.43 0.49 0-1 0.42 0.49 0-1 > 25% 0.57 0.49 0-1 0.58 0.49 0-1 Total Proportion of Low Educated Re sidents 0.28 0.13 0-1 0.28 0.13 0-1 a Standard error is reported for Level I variables rather than standard deviation as the means are weighted. Standard deviation is reported for Level II variables as there is no weight variable available in the Add Health study for the neighborhood level data.

PAGE 105

92 Univariate Analyses Adolescents’ Access to Health Care The study sample was composed of 1,526 sexually experienced adolescent females. The majority of adolescents in th e sample (66.92%) reported they had received a routine physical in the past year and near ly one-half (49.29%) reported having received contraceptives from a health care provider at some point in the past. Approximately 25% of the sample reported unmet health needs ove r the past year, define d as situations in which they thought they should have obtaine d medical assistance, but did not receive care. The most common reasons that the a dolescents in this sample reported for not having received health care was that they thought their problem w ould go away (65.5%), they were afraid of what the doctor w ould say or do (18.28%), they could not pay (15.86%) or they didn’t want their parent s to know (14.36%). Ta ble 3 summarizes the descriptive statistics related to a dolescents’ access to health care. Table 3 Descriptive Statistics of Adolescents’ Access to Health Care (N = 1,526) Dependent Variables N Proportion % SE Received Routine Physical 1044 66.92 2.10 Received Contraceptive Services 734 49.29 1.54 Reported Unmet Health Needs 382 23.50 1.46 Reasons for Unmet Health Needs Didn’t know who to go see 23 6.32 1.30 No transportation 43 10.00 1.74 No one to go along 15 1.43 0.43 Parent would not go 50 12.31 2.18 Didn’t want parents to know 64 14.36 2.39 Hard to make appointment 40 11.38 2.14 Afraid of what doctor would say or do 78 18.28 2.34 Thought problem would go away 230 66.50 2.95 Could not pay 64 15.86 2.42 Other 27 5.75 1.22 a Standard errors are reported for Level I variable s due to the weighting of Level I variables.

PAGE 106

93 Characteristics of Adoles cents in the Study Sample The results revealed that 77.75% of the adolescents in the study sample selfidentified racially as non-Hispanic White and 22.25% self-identif ied racially as nonHispanic Black. The mean household income for the sample in 1994 was $44,660. When examining household income as a functi on of the ratio of income to poverty, approximately 20% of the study sample we re 100% or more below their poverty threshold, 21.65% were 101% to 200% above their poverty threshold, 21.76% were 201% to 300% above their poverty threshold, 12.70% were 301% to 400% above their poverty threshold, and 24.23% we re 401% or above their pover ty threshold. Nearly 20% of the adolescents’ had at l east one parent who had a college degree or more, 30.38% had a parent who had completed some college or technical training, 35.03% had a parent who had a high school degree or GED and 17.71% had a parent who had not completed high school. Descriptive analyses for the Level I contro l variables indicated that the age of the adolescents in the study ranged from 15 year s to 19 years, although the majority were between 15 years to 18 years of age (98.27%). The mean age was 16.60 years. The mean score for self-rated health was 3.70 ( SE = 0.03) on a scale from 1 to 5, which indicated that the average adolescent in this study believed she wa s in good to very good health. Nearly 7% of adolescents reported a hist ory of STI, and 15.89% had a history of a pregnancy. In regard to their attitudes and beli efs about pregnancy and contraception, adolescents’ mean scores s uggested that they believed that a pregnancy would be undesirable at that time of the interview ( M = 3.71, SE = 0.04, range 1-5), and that their

PAGE 107

94 mother would disapprove of their sexual activity ( M = 3.98, SE = 0.03, range 1-5). The mean scores also suggested that adolescents disagreed with the statements that birth control was difficult to access or utilize ( M = 1.87, SE = 0.03, range 1-5), and that their mother would disapprove of their birth control use ( M = 2.35, SE = 0.04, range 1-5). Most adolescents in this sample reporte d living in a household with two parents (61.45%), and approximately half of the samp le had a valid driver’s license (53.82%). A high proportion of adolescents in this samp le had health insura nce (88.87%). Among the adolescents who had health insurance, 82.73% had private health insurance, 11.92% had public insurance, 4.17% had “other” insuranc e, and 1.19% had more than one type of insurance. Most adolescents ha d received some formal health education at school related where to seek medical assist ance (82.57%), but th e availability of school health care services were rather limited, as 17.76% of a dolescents attended a school that offered nonathletic physicals. Table 4 summarizes the desc riptive statistics related to adolescents’ characteristics and their access to health care.

PAGE 108

95 Table 4 Descriptive Statistics of Adoles cents’ Characteristics (N = 1,526) Adolescents’ Characteristics N Proportion% SE Level I Independent Variables Race Black 497 22.25 4.11 White 1,029 77.75 4.11 Ratio of Income to Poverty 0-100% 303 19.65 2.38 101-200% 340 21.65 1.47 201-300% 319 21.76 1.96 301-400% 206 12.70 1.38 400% or More 358 24.23 1.94 Parental Education Less than High School 212 14.71 1.96 High School Degree or GED 508 35.03 1.81 Some College/Technical 479 30.38 1.87 College Degree or More 327 19.89 1.71 STI History 108 6.83 0.81 Pregnancy History 242 15.89 1.60 Health Insurance 1,351 88.87 1.00 Driver’s License 774 53.82 2.99 Two Parent Household 930 61.45 2.27 School Education: Medical Attention 1,261 82.57 1.52 School Health Care: Non-Athl etic Physical 295 17.76 4.27 N M SE Total Household Income in Thousands 1,526 44,660.00 1,980.00 Age 1,526 16.60 0.04 Length of Residence in Years 1,526 8.08 0.25 General Health 1,526 3.70 0.03 Pregnancy Undesirable 1,526 3.71 0.04 Barriers to Birth Control 1,526 1.87 0.03 Mom Disapprove of Sex 1,526 3.98 0.03 Mom Disapprove of Birth Control 1,526 2.35 0.04 Distribution of Adolescents’ Soc ioeconomic Position by Race Descriptive analyses were examined to gain a better understanding of the distribution of adolescents’ socioeconomic position as a function of adolescent race. Among this sample of adolescents, 42.67% of th e adolescents who self -identified racially as Black had a household income to povert y ratio 100% or more below their poverty threshold compared to 13.06% of those adoles cents who self-identifie d racially as White. Furthermore, only 10.67% of the adolescents w ho self-identified racially as Black had a household income to poverty ratio over 400% of their poverty threshold compared to

PAGE 109

96 28.12% percent of those adolesce nts who self-identified racial ly as White. In relation to level of parental educational attainment, 20.21% of the adolescent s who self-identified racially as Black had a parent with less than a high school degree and only 13.13% of adolescents who self-identified racially as White had a parent with less than a high school degree while 13.56% of the adolescents who self-i dentified racially as Black had a parent with a college degree or more compared to 21.70% of adolescents who self-identified racially as White. Table 5 summarizes the findings of these analyses. Table 5 Distribution of Adolescents’ Socioeconomic Position by Adolescent Race (N=1,526) Adolescents’ Socioeconomic Position Adolescents’ Race White n =1029 Adolescents’ Race Black n =497 n % SE n % SE Ratio of Income to Poverty 0-100% 134 13.06 1.74 169 42.67 4.84 101-200% 219 20.49 1.57 121 25.71 2.57 201-300% 235 23.59 2.41 84 15.38 2.23 301-400% 160 14.74 1.56 46 5.57 1.22 400% or More 281 28.12 2.07 77 10.67 2.23 Parental Education Less than High School 127 13.13 2.12 85 20.21 3.25 High School Degree or GED 361 35.19 2.09 147 34.48 3.42 Some College/Technical 321 29.98 2.20 158 31.75 3.38 College Degree or More 220 21.70 2.01 107 13.56 2.52 Characteristics of the Nei ghborhoods in the Study Sample The participants in this study were di stributed across 546 neighborhoods with the number of participants pe r neighborhood ranging from 1 to 40. The characteristics of these neighborhoods were examined to gain a be tter understanding of th e social structural context in which the adolescents in this study lived. The results of these finding are summarized in Table 6.

PAGE 110

97 The mean proportion of Black residents acr oss all the neighborhoods in this study was 21%. When examining the racial dist ribution of the neig hborhoods’ populations, 62% of the adolescents from this study liv ed in neighborhoods with fewer than 11% Black residents, 20% lived in neighborhoods with 11%49% Black residents, 11% lived in neighborhoods with 50% 90% Black residents, and the other 7% lived in neighborhoods with over 90% Bl ack residents. Thus, the majority of the study sample lived in neighborhoods characterized by a low proportion of Black residents. In relation to the neighborhood socioeconomic context, 23% of the adolescents in this study lived in neighborhoods that had 20% or more residents living below the 1989 poverty threshold of $12,674 for a family of four (U.S. Census Bureau, n.d.). The mean proportion of residents living below the 1989 poverty line across all neighborhoods in this study was 12%. Nearly 60% of the adoles cents in this study liv ed in neighborhoods in which 25% or more of the residents who we re 25 years of age or older did not have a high school degree or GED. The mean proportion of lower e ducated residents across all neighborhoods in this study was 28%. The Level II control variables examined in this study included geographic region, the proportion of residents livi ng within urban areas, total po pulation, residential stability, and median age. The geographic distribution of the sample was slightly uneven as nearly 15% of the adolescents lived in neighbor hoods located in the West, 31% lived in neighborhoods located in the Mi dwest, 39% lived in neighborh oods located in the South, and 15% lived in neighborhoods located in th e Northeast. Approximately half of the adolescents lived in neighborhoods that were co nsidered urban (55%) with an average of 5,819 residents per census tract. Across a ll the neighborhoods in the study, the mean

PAGE 111

98 proportion of residentially stable neig hborhoods was 57% and the median age of residents was 32 years old. Table 6 Descriptive Statistics of the Charac teristics of Adolescents’ Neighborhoods (N = 1,526 Adolescents across 546 Neighborhoods) Neighborhood Characteristics a N M SD Range Proportion of Black Residents 10% (reference) 942 0.62 0.49 0-1 11% 50% 299 0.20 0.40 0-1 51% 90% 171 0.11 0.32 0-1 91% or more 114 0.07 0.26 0-1 Total Proportion of Black Residents 1,526 0.21 0.31 0-1 Proportion of Poor Residents 20% 1,175 0.77 0.42 0-1 > 20% 351 0.23 0.42 0-1 Total Proportion of Poor Residents 1,526 0.12 0.11 0-1 Total Median Income 1,526 30,035.74 12,853.83 7,370.00109,242.00 Proportion of Low Educated Residents 25% 651 0.43 0.49 0-1 > 25% 875 0.57 0.49 0-1 Total Proportion of Low Educ ated Residents 1,526 0.28 0.13 0-1 Total Population 1, 526 5,819.70 4,284.54 940.0037,612.00 Proportion of Residents Living in Urban Neighborhood 1,526 0.55 0.48 0-1 Median Age 1,526 31.95 4.07 19.61-56.26 Residential Stabilit y 1,526 0.57 0.12 0-1 West 233 0.15 0.36 0-1 Midwest 467 0.31 0.46 0-1 Northeast 232 0.15 0.36 0-1 South 594 0.39 0.49 0-1 a Unweighted means as no weight variables are available in Add Health for the contextual data. Bivariate Analysis A series of bivariate analyses employi ng logistic regression were conducted to gain a better understanding of the relationships between the variables in this study. First, relationships between the Level I and Level II sociodemographic and control variables and the dependent variables of interest – rece ipt of a routine physical in the past year, report of unmet health needs in the past year and receipt of contraceptives from health care provider at least once in the past – were examined. Finally, so cial disparities in

PAGE 112

99 health were examined to gain a better understanding of disparities in health needs among this sample of sexually expe rienced adolescent females. Adolescents’ Characteristics and Access to Health Care Relationships between adolescent soci odemographic characteristics, Level I control variables, and the dependent vari ables received a rou tine physical, reported unmet health needs, and received contracep tive services, were examined via logistic regression analyses. The purpose of these anal yses were to gain an understanding of the extent to which racial and socioeconomic di sparities in access to health care existed among the sample of adolescents in this st udy as well as to expl ore the relationships between the control variables and access to he alth care. When examining the associations between adolescents’ sociodemographic charact eristics and access to health care, the reference groups for the analyses were the more socially advantaged groups. Thus, adolescents who self-identif ied racially as non-Hispanic White, adolescents whose household income to poverty ratio was 401% or more over the 1989 poverty threshold, and adolescents who had a parent with a colle ge degree or more were considered to be the reference groups for the analyses. Table 7 summarizes the results of these analyses. In relation to racial differences in access to health care, there were no significant differences in the receipt of a routine physic al among those adolescen t who self-identified racially as Black (OR = 1.33, 95% CI = 0.94, 1.87) and those adolescents who selfidentified racially as White. However, adoles cents who self-identifie d racially as Black were more likely to report unmet health need s than those adolescents who self-identified as White (OR = 1.36, 95% CI = 1.01, 1.84). Contra ry to the hypothese s, adolescents who self-identified racially as Black were more lik ely to have received contraceptive services

PAGE 113

100 than those adolescents who se lf-identified racially as White (OR = 1.48, 95% CI = 1.09, 2.03). Socioeconomic disparities in access to h ealth care were also examined via two indicators – ratio of household income to poverty and level of parental educational attainment. There were no statis tically significant differences in the receipt of a routine physical or the receipt of contraceptive servi ces between those adoles cents whose ratio of income to poverty was 400% or lower and t hose adolescents whose ratio of income to poverty was greater than 400%. However, adolescents whose ratio of income to poverty was 100% or below the poverty threshold povert y threshold were more likely to report unmet health needs than those adolescents whose ratio of income to poverty threshold was over 400% of the poverty thres hold (OR = 2.11, 95% CI = 1.46, 2.29). In relation to parental education, there we re no statistically si gnificant differences in the receipt of a r outine physical or unmet health n eeds between those adolescents who had a parent with less than a college degree and those adolescents w ho had a parent with a college degree or more. Furthermore, the only significant finding for access to contraceptive services was that adolescents who had a parent with less than a high school degree were more likely to have received c ontraceptive services than those adolescents who had a parent with a college degree or more (OR = 1.63, 95% CI = 1.05, 2.53). Thus, socioeconomic disparities in access to health ca re were noted in the bivariate analyses for unmet health needs among whose adolescents whose ratio of income to poverty was 100% or below the poverty threshold.

PAGE 114

101 Table 7 Associations between Adolescents’ Sociode mographic Characteristics and Access to Health Care (n = 1,526) Adolescents’ Sociodemographic Characteristics Received Routine Physical Reported Unmet Health Needs Received Contraceptives OR 95% CI OR 95% CI OR 95% CI Race Black 1.33 0.94, 1.88 1.36 1.01, 1.84 1.48 1.09, 2.03 White (reference) Ratio of Income to Poverty 0-100% 1.00 0.67, 1.51 2.11 1.33, 3.35 1.30 0.90, 1.86 101-200% 0.82 0.55, 1.22 1. 18 0.78, 1.80 1.00 0.70, 1.43 201-300% 0.93 0.63, 1.37 1. 49 0.93, 2.39 0.72 0.51, 1.03 301-400% 0.81 0.50, 1.32 1. 39 0.84, 2.30 0.81 0.53, 1.24 400% or More (reference) Parental Education Less than High School 0.73 0.43, 1.24 1.21 0.71, 2.06 1.63 1.05, 2.53 High School Degree GED 0.72 0.51, 1.01 1.23 0.80, 1.91 1.17 0.84, 1.64 Some College/Technical 0.72 0.46, 1.12 1.36 0.84, 2.21 1.22 0.84, 1.78 College Degree or More (reference) Relationships between adolescent contro l variables and receipt of a routine physical, unmet health needs, and receipt of contraceptive services were also examined via logistic regression analysis. The findings from these analyses are summarized in Table 8. Health needs were significantly re lated to access to health care among this study’s sample of adolescents. Specifically, a dolescents who reported better health were more likely to have received a routine physical (OR = 1.20, 95% CI = 1.03, 1.39) and less likely to report unmet health needs (OR = 0.67, 95% CI = 0.57, 0.77). There was no significant relationship betw een self-reported general h ealth and the receipt of contraceptive services (OR = 1.10, 95% CI = 0.92, 1.32). In additi on, adolescents who had a history of a STI were more likely to have received a routine physical (OR = 1.83, 95% CI = 1.12, 2.97) and contraceptive se rvices (OR = 5.21, 95% CI = 2.85, 9.51) as well as more likely to report unmet h ealth needs (OR = 2.17, 95% CI = 1.26, 3.71).

PAGE 115

102 Having experienced a pregnancy was also as sociated with access to health care as adolescents who had been pregnant were more likely to report they had received a routine physical (OR = 1.58, 95% CI = 1.01, 2.49) and contraceptive services (OR = 2.47, 95% CI = 1.70, 3.59). There was no significant relation ship between a history of a pregnancy and unmet health needs (OR = 1.11, 95% CI = 0.74, 1.68). Adolescents’ attitudes and beliefs regarding sexuality and contraceptives were significantly associated with access to health care, particularly in the receipt of contraceptive services. For example, adoles cents who believed that their mothers would disapprove of their sexual activity (OR = 0.86, 95% CI = 0.76, 0.97) or use of birth control (OR = 0.85, 95% CI = 0.75, 0.97) were le ss likely to have received a routine physical. In addition, adolescents who believe d that their mothers would disapprove of their sexual activity (OR = 0.59, 95% CI = 0.51, 0.69) or use of birth control (OR = 0.65, 95% CI = 0.58, 0.73) were less likely to have received cont raceptive services. Adolescents who reported more barriers to bi rth control were more likely to have unmet health needs (OR = 1.29, 95% CI = 1.04, 1.59) and less likely to have received contraceptive services (OR=0.76, 95% CI = 0.63, 0.93). Lastly, adolescents who did not consider a pregnancy as an undesirable event were also less likely to receive contraceptive services (O R = 0.72, 95% CI = 0.63, 0.83). Age was associated with the receipt of c ontraceptive services as older adolescents were more likely to receive contraceptive services than younger adolescents (OR = 1.36, 95% CI = 1.22, 1.51), but no associations were noted with the r eceipt of a routine physical (OR = 1.01, 95% CI = 0.88, 1.15) or unmet health needs (OR = 0.93, 95% CI = 0.81, 1.07). No significant relationships were revealed between the length of time the

PAGE 116

103 adolescent lived in her current residence and any of the indi cators for access to health care. Contrary to the findings in the litera ture, health insurance status was not significantly associated with access to hea lth care among this sample of adolescents. Family structure was significant though, as adolescents from a two parent household were less likely to report unmet health needs (OR = 0.56, 95% CI = 0.41, 0.77) and to receive contraceptive services (OR = 0.65, 95% CI = 0.51, 0.81) than those adolescents who lived in a single parent household. Transpor tation was associated with the receipt of contraceptive services and unmet health need s, as those adolescents who had a driver’s license were more likely to receive contraceptive services (OR = 1.53, 95% CI = 1.22, 1.92) and less likely to report unmet health needs (OR = 0.75, 95% CI = 0.56, 0.77) than those adolescents who did not have a driver’s license. School health education and the provision of routine physicals were significantly related to access to health care among the adolescents in this sample. Specifically, adolescents who received school health educa tion about where to seek medical attention were less likely to report unmet health n eeds (OR = 0.63, 95% CI = 0.42, 0.93) than those adolescents who did not receive this instru ction. Furthermore, adolescents were more likely to have received a routine physical if they attended a school that provide nonathletic physicals than those adolescent who attended a sc hool without these services (OR = 1.65, 95% CI = 1.06, 2.57).

PAGE 117

104 Table 8 Associations between Adoles cents’ Predisposing Factors, Enabling Factors, and Health Needs and Access to Health Care (N = 1,526) Adolescents’ Predisposing Factors, Enabling Factors & Health Needs Received Routine Physical Reported Unmet Health Needs Received Contraceptive Services OR 95% CI OR 95% CI OR 95% CI Age 1.01 0.88, 1.15 0.93 0.81, 1.07 1.36 1.22, 1.51 Pregnancy Undesirable 0.97 0.85, 1.12 0.96 0.83, 1.10 0.72 0.63, 0.83 Barriers to Birth Control 0.86 0.70, 1.04 1.29 1.04, 1.59 0.76 0.63, 0.93 Mother Disapprove: Sex 0.86 0.76, 0.97 1.00 0.84, 1.18 0.59 0.51, 0.69 Mother Disapprove: Birth Control 0.85 0.75, 0.97 1.09 0.97, 1.21 0.65 0.58, 0.73 Health Insurance 1.31 0.83, 2. 08 0.79 0.51, 1.22 0.97 0.65, 1.44 Valid Driver’s License 1.06 0.82, 1.38 0.75 0.56, 0.99 1.53 1.22, 1.92 Two Parent Household 1.01 0.78, 1.32 0.56 0.41, 0.77 0.65 0.51, 0.81 School Education: Medical Attention 1.36 0.93, 2.01 0.63 0.42, 0.93 1.17 0.88, 1.55 School Health Care: Physical 1.65 1.06, 2.57 1.09 0.81, 1.46 0.87 0.67, 1.13 General Health 1.20 1.03, 1.39 0.67 0.57, 0.77 1.10 0.92, 1.32 STI History 1.83 1.12, 2.97 2.17 1.26, 3.71 5.21 2.85, 9.51 Pregnancy History 1.58 1.01, 2.49 1.11 0.74, 1.68 2.47 1.70, 3.59 Neighborhood Characteristics and Access to Health Care The relationships between neighborhood sociodemographic characteristics, neighborhood control variables, and adoles cents’ access to health care were also examined via logistic regression analyses Table 9 summarizes the results of these analyses. Similar to the analyses examining adolescent sociodemographic characteristics and access to health care, the reference groups for these analyses were those more socially advantaged neighborhoods. Thus, th e reference group when examining the neighborhood racial context were those neighborhoods in which 10% or fewer the residents were Black while the refere nce groups when examining the neighborhood socioeconomic context were those neighborhoods in which less than 20% of the residents were living below the 1989 poverty threshold and those neighborhoods in which less than 25% of the residents aged 25 years and ol der did not have a high school degree.

PAGE 118

105 In relation to the neighborhood racial context, adolescen ts who lived in neighborhoods where over 90% of the resident s were Black were more likely to report unmet health needs than those adolescents w ho lived in neighborhoods in which less than 10% of the residents were Black (OR = 1.66, 95% CI = 1.27, 2.17). However, this relationship was only significant when the ne ighborhood racial context was examined as categorically based on their cut-points rath er than continuously (OR = 1.42, 95% CI = 0.97, 2.09). In addition, contrary to the hypotheses, adolescents who lived in neighborhoods characterized by a higher propor tion of Black residents were more likely to have received a routine physical than those adolescents who lived in neighborhoods where the proportion of adol escents was 10% or less (O R = 1.91, 95% CI = 1.04, 3.52). This relationship was significant when th e neighborhood racial context was measured continuously and categorically. Similar findings were noted for the neighborhood socioeconomic context as adolescents who lived in neighborhoods in wh ich more than 20% of the residents were below the 1989 poverty threshold were more likely to receive a physical than those adolescents who lived in neighborhoods in wh ich less than 20% of the residents were living below the 1989 poverty threshold (OR = 1.44, 95% CI = 1.08, 1.92). There were no significant relationships be tween the educational level of the residents within the neighborhoods and access to health care. In addition, only one control variable was significant in which adolescents who lived in the Northeast were more likely to have a routine physical than those adolescents who lived in th e Midwest (OR = 2.10, 95% CI = 1.14, 3.88).

PAGE 119

106 Table 9 Associations between the Characteristic s of Adolescents’ Neighborhoods and Access to Health Care (N = 1,526 adoles cents across 546 neighborhoods) Neighborhood Sociodemographic Characteristics Received Routine Physical Reported Unmet Health Needs Received Contraceptive Services OR 95% CI OR 95% CI OR 95% CI % of Black Residents 10% (reference) 11% 50% 1.39 0.88, 2.21 1.05 0.68, 1.61 0.98 0.70, 1.38 51% 90% 1.53 0.98, 2.41 1.11 0.64, 1.93 1.23 0.78, 1.92 91% or more 1.91 1.04, 3.52 1.66 1.27, 2.17 1.64 0.91, 2.94 Continuous 2.05 1.19, 3.53 1.42 0.97, 2.09 1.46 0.92, 2.33 % of Poor Residents 20% (reference) > 20% 1.44 1.08, 1.92 1.13 0.76, 1.66 1.02 0.75, 1.39 Continuous 4.33 1.31, 14.34 1.76 0.59, 5.27 0.96 0.30, 3.05 % of Low Educated Residents 25% (reference) > 25% 1.07 0.75, 1.52 1.02 0.73, 1.42 1.16 0.91, 1.47 Continuous 1.96 0.65, 5. 91 1.31 0.51, 3.32 1.59 0.56, 4.51 Region Midwest (reference) West 0.72 0.41, 1.29 1. 31 0.75, 2.26 1.34 0.85, 2.13 South 0.97 0.62, 1.51 1. 14 0.75, 1.73 1.00 0.75, 1.34 Northeast 2.10 1.14, 3.88 1.13 0.69, 1.86 0.99 0.71, 1.38 Residential Stability 2.40 0.72, 7.97 1.05 0.37, 2.95 1.21 0.35, 4.23 % of Residents in Urban Area 0.99 0. 68, 1.44 0.98 0.70, 1.38 1.04 0.82, 1.32 Median Age of Resident s 0.99 0.96, 1.03 0.99 0. 95, 1.03 0.99 0.95, 1.02 Total Population 1.00 1.00, 1. 00 1.00 1.00, 1.00 1.00 1.00, 1.00 Social Disparities in Adolescents’ Health Bivariate analyses were conducted to determine the extent to which the adolescents in this sample experienced raci al and socioeconomic disparities in health. Three health outcomes were examined-pregnanc y history, STI history, and general health status. In these analyses, ge neral health status was cons idered a dependent variable instead of an independent variable, as noted in the preceding and subsequent analyses. Thus, general health status was dichotom ized as poor/fair health and good/very good/excellent health to facilita te these particular analyses but in all other analyses,

PAGE 120

107 general health status is measured as a con tinuous variable. The findi ngs of these analyses are summarized in Table 10. The findings from these analyses revealed racial and socioeconomic disparities in health among this sample of sexually expe rienced adolescent females. For example, adolescents who self-identified racially as Black were more likely to have experienced a pregnancy (OR = 2.27, 95% CI = 1.45, 3.55) and a STI (OR = 3.70, 95% CI = 2.32, 5.91) than those adolescents who self-identified racially as White. However, no racial disparities were noted in general health status. In relation to socioeconomic dispari ties in health, adolescents who had a household income to poverty ratio below 100% of their poverty threshold (OR = 3.63, 95% CI = 2.18, 6.05), between 101% -200% of their poverty threshold (OR = 3.13, 95% CI = 1.95, 5.04), and between 201% to 300% of their poverty threshold (OR = 1.77, 95% CI = 1.07, 2.95) were more likely to have expe rienced a pregnancy than those adolescents who had a household income to poverty ratio over 400% of their poverty threshold. In addition, those adolescents who had a parent with less than a high school degree (OR = 3.82, 95% CI = 2.15, 6.79) or those who had a pa rent with some college or technical training (OR = 1.81, 95% CI = 1.03, 3.22) were more likely to have experienced an unintended pregnancy than thos e adolescents who had a parent with a college degree or more. Few socioeconomic disparities were not ed in relation to having experienced a STI or in general health status. For example, the findings revealed that adolescents who had a parent with less than a high school degree were more likely to have experienced a STI (OR = 2.58, 95% CI = 1.25, 5.32) and they were le ss likely to rate th eir health as good,

PAGE 121

108 very good or excellent (OR = 0.31, 95% CI = 0.14, 0.72) than those adolescents who had a parent with a college degree or more. Table 10 Associations between Adoles cents’ Sociodemographic Charac teristics and Adolescents’ Health (N = 1,526) Adolescent Sociodemographic Characteristics Pregnancy History STI History General Health Status OR 95% CI OR 95% CI OR 95% CI Race Black 2.27 1.45, 3.55 3.70 2.32, 5.91 0.95 0.61, 1.48 White (reference) Ratio of Income to Poverty 0-100% 3.63 2.18, 6.05 1.84 0.90, 3.76 0.85 0.43, 1.69 101-200% 3.13 1.95, 5.04 1.22 0.55, 2.70 1.00 0.59, 1.69 201-300% 1.77 1.07, 2.95 0.60 0.24, 1.49 0.84 0.49, 1.44 301-400% 1.57 0.74, 3.36 1. 23 0.52, 2.93 0.72 0.35, 1.46 401% or More (reference) Parental Education Less than High School 3.82 2.15, 6.79 2.58 1.25, 5.32 0.31 0.14 ,0.72 High School Degree GED 1.57 0.84, 2.93 1.60 0.78, 3.30 0.48 0.23, 1.01 Some College/Technical 1.82 1.03, 3.22 1.16 0.54, 2.49 0.47 0.22, 1.03 College Degree or More (reference) Multivariate Analyses The hypotheses for this dissertation were tested using a series of nested hierarchical generalized linear models (HGLM) for each of the three dependent variables – receipt of a routine physical reported unmet health needs, and receipt of contraceptive services. The full and reduced models were compared via likelihood ratio testing (ChiSquare) to determine if the addition of vari ables into the model improved the model fit. The size and significance of the random intercept (T00) were examined between each step in the model building to determine if the para meters entered into the model could account for any of the variance in the dependent va riable across neighborhoods. The results for the HGLM are reported as odds ratios and a p-value of less than 0.05 was the criterion

PAGE 122

109 employed to determine the level of significan ce for rejecting the null hypothesis. Due to the limited sample size of the cells when examining individual race stratified by socioeconomic position, Level I interactions between individual race and socioeconomic position could not be reliably examined with this study. Thus, Hypothe sis IC could not be tested. Table B1 in Appendix A summari zes the model building process and the contribution of each model in testing the three Hypotheses for this study. Hypothesis I The aims of Hypothesis I were: (1) to test whether individual race and socioeconomic position influenced access to health care among sexually experienced adolescent females after adjusting for the Leve l I control variables, and (2) to determine the extent to which access to health care varied across the neighborhoods in this study. A series of four nested models were used to test Hypothesis I for each of the three dependent variables – receipt of a routine physical, unmet h ealth needs, and receipt of contraceptive services. Model I: The null model. There were no independent variables in the model. The Level I intercept varied ra ndomly across the neighborhoods so the extent to which the dependent variable varied across th e neighborhoods could be ascertained.2 Model II: The Level I race variable was entered into the model. The slope of the race variable3 and the Level I intercept varied randomly across the neighborhoods. 2 In a two-level multilevel model, the Level I intercept is set by the researcher to vary randomly across the Level II groups so the extent to which the dependent variable varies across the Level II groups (e.g. neighborhoods) can be ascertained. 3 In a two-level multilevel model, the slope of an independent variable is set by the researcher to vary randomly across the Level II group so the extent to which the magnitude of the relationship between the independent variable and the dependent variable varies across neighborhoods can be ascertained.

PAGE 123

110 Model III: The Level I socioeconomic va riablesratio of household income to poverty4 and level of parental educational attain ment, were entered into the model. The ratio of income to poverty was measured via five categorical dummy variables and parental educational attainment was measur ed via four categorical dummy variables. Direct and indirect effects of socioeconomic position on access to health care were examined. The variables were all uncentere d and their slopes va ried randomly across neighborhoods. The Level I intercept vari ed randomly across the neighborhoods. Model IV: The Level I control variables we re entered into the model. Continuous variables were grand mean centered and cat egorical variables were uncentered. The slopes of the Level I control variables were fi xed. The Level I intercep t and the slopes of the variables for race and socioeconomic position varied randomly across the neighborhoods. Receipt of a Routine Physical Table 11 summarizes the results of Models I-IV that tested Hypothesis I for the dependent variable, receipt of a routine physic al. The results of Model I, the null model, indicated that in the typi cal neighborhood the probability of having received a routine physical among the sexually experienced adoles cents in this sample was 69%. The results of the null model also revealed that the odds of having received a routine physical varied across neighborhoods (T00 = 0.09, p < 0.05). Model II estimated the effects of i ndividual race on access to health care. The random effect of race was non-significant, wh ich indicated that the relationship between 4 Household income measured as a continuous variable was examined initially in all the HGLM models, but was not a significant predictor. The ratio of household income to poverty takes into account household size and in several of the models, this was a significant predictor.

PAGE 124

111 race and the odds of having received a r outine physical did not vary across neighborhoods. The slope was then fixed5 and the model was reanal yzed with the results presented in Model II of Table 11. The findings revealed that in th e typical neighborhood, there were no significant differences in th e odds of having receive d a routine physical between those adolescents who were categori zed racially as Black and those who were categorized racially as White (OR = 1.21, 95% CI = 0.93, 1.58). Since independent variables were entered into the regression equation, the fixed intercept became a regression coefficient in this model and repr esented the mean of the dependent variable when all the independent variables were equa l to zero. In comparing Model I to Model II, the size and significance of the random intercept remained relatively unchanged, which indicated that there was still significant variation in the odds of having received a routine physical across neighborhoods after accounting for individual race. Model III examined the extent to which socioeconomic position influenced access to health care above and beyond adolescents’ individual race as well as the extent to which socioeconomic position mediated the relationship between individual race and having received a routine physical The random effects for the ratio of income to poverty and parental educational attainment were non-significant, which indicated that the relationship between socioeconomic position and the odds of having received a routine physical did not vary across neighborhoods6. The slopes were fixed and the model was reanalyzed with the results presen ted in Model III of Table 11. 5 When the slope of an independent variable is fixed, the magnitude of the relationship between the independent and dependent variable is constant across all neighborhoods (Raudenbush & Bryk, 2002). 6 Initially, the slopes of the variables for ratio of income to poverty and parental educational attainment varied randomly in the model simultaneously, but the model failed to converge. The model was then reanalyzed with the slopes for the ratio of income to poverty randomly varying and parental educational

PAGE 125

112 Direct effects of socioeconomic positi on on access to health care were noted. Specifically, after adjusting for individual ra ce and the ratio of in come to poverty, the results indicated that in a typical neighbor hood, adolescents who had a parent with some college education but no college degree (OR = 0.68, 95% CI = 0.49, 0.94) had significantly lower odds than thos e adolescents who had a parent with a college degree or more to have received a routine physical. However, in the typical neighborhood there were no significant differences in the odds of having received a routine physical between those adolescents who had a parent with a college degree or more and those adolescents who had a parent with less than a high school degree (OR = 0.77, 95% CI = 0.49, 1.01) or those adolescents who had a parent with a high school degree (OR = 0.73, 95% CI = 0.52, 1.04). In a typical neighborhood, there were no significant differences in the odds of having received a routine physical and thos e adolescents whose household income to poverty ratio was 400% or less and those adol escents whose household income to poverty ratio was 401% or greater, after adjusting for individual race and parental educational attainment. After adjusting for socioeconomic positi on, the findings suggested that in a typical neighborhood there were no significant differences in the odds of having received a routine physical between those adolescents who self-identified racially as Black and those adolescents who self -identified as White (OR = 1.27, 95% CI = 0.96, 1.67). In comparing Model II to Model III, the estima te and significance of the random intercept remained relatively unchanged, which indicated that there was still significant variation fixed, and then again with the slopes for parental educational attainment randomly varying and the ratio of income to poverty fixed. Multilevel models can be constrained by the number of random effects that can be estimated at one time (Raudenbush & Bryk, 2002).

PAGE 126

113 in the odds of having received a routine physical across neighborhoods after accounting for individual race and socioeconomic position. In Model IV, the control va riables that represented a dolescents’ health needs, predisposing factors, and enabling factors were entered into the model, which impacted the significance of several individual sociode mographic variables. Specifically, in the typical neighborhood, after adjusting for individual race, pa rental educational attainment, and the control variables, no significant di fference were noted in the odds of having received a routine physical between those adolescents whose household income to poverty ratio was 400% or lower and those whose household income to poverty ratio was 400% or greater. However, in the typical ne ighborhood, after adjusting for the income to poverty ratio, race, and the control variables, those adolescents who had a parent with some college or technical training had si gnificantly lower odds of having received a routine physical than those a dolescents whose parents had a college education or more (OR = 0.71, 95% CI = 0.50, 0.99). In comparing M odel III to Model IV the size of the estimate for the random intercept decreased from a T00 of 0.10 to a T00 of 0.03, which suggested that the variance in the receipt of a routine physical across neighborhoods was in part a function of the individual co mposition of the neighborhood. However, the random intercept was statistic ally significant, which indi cated that there was still variation in the odds of having received a routine physical across neighborhoods after adjusting for individual race, socioecono mic position, and the individual control variables.

PAGE 127

114 Table 11 HGLM Analyses: Social Disparities in the Receipt of a Routine Physical (N=1,526) Model I Model II Model III Model IV Level I Fixed Effects OR 95% CI OR 95% CI OR 95% CI OR 95% CI Intercept 2.25 2.00, 2.52 2.11 1.83, 2.43 3.06 2.28, 4.12 1.09 0.59, 2.03 Race Black 1.21 0.93, 1.58 1.27 0.96, 1.67 1.32 0.97, 1.81 White (reference) Parental Education Less than High School 0.77 0.50, 1.19 0.84 0.54, 1.31 High School Degree or GED 0.73 0.52, 1.04 0.77 0.54, 1.11 Some College/Technical 0.68 0.49, 0.94 0.71 0.50, 0.99 College Degree or More (reference) Ratio of Income to Poverty 0-100% 0.81 0.57, 1.17 0.95 0.63, 1.43 101-200% 0.76 0.53, 1.08 0.78 0.53, 1.15 201-300% 0.98 0.70, 1.39 1.05 0.73, 1.52 301-400% 0.91 0.62, 1.33 0.92 0.62, 1.37 401% or More (reference) General Health 1.19 1.04, 1.35 STI History 1.76 1.03, 3.02 Pregnancy History 1.64 1.17, 2.29 Age 0.92 0.80, 1.05 Pregnancy Undesirable 1.02 0.88, 1.18 Barriers to Birth Control 0.91 0.77, 1.08 Mom Disapprove of Sex 0.95 0.80, 1.13 Mom Disapprove of Birth Control 0.88 0.79, 0.97 Length of Residence in Years 0.99 0.97, 1.01 Health Insurance 1.56 1.05, 2.31 Driver’s License 1.32 0.95, 1.82 Two Parent Household 0.99 0.75, 1.33 School Education: Medical Atten tion 1.21 0.88, 1.66 School Health Care Non-Athl etic Physical 2.11 1.51, 2.94 Random Intercept 0.09 0.09 0.09 0.03 -2LL 4705.34 4703.00 4690.97 4623.33 2 2.33(1) 12.03(7) 67.64(14)

PAGE 128

115 Unmet Health Needs Table 12 summarizes the results of the models testing Hypothesis I for the dependent variable, unmet health needs. The re sults of Model I, the null model, revealed that in the typical neighborhood, the probabi lity of reporting unmet health needs among the sexually experienced adolescent females in this sample was 24.8%. In addition, the results of the null model indicated that there was variance across neighborhoods in the odds of unmet health needs (T00 = 0.02, p < 0.05). Model II estimated the effects of indi vidual race on unmet health needs. The random effect for race was non-significant, wh ich indicated that the relationship between race and the odds of unmet health needs di d not vary across neighborhoods. The slope was fixed and the model was reanalyzed with the results presented in Model II of Table 12. The findings revealed that in a typi cal neighborhood, there were no significant differences in the odds of unmet health needs between those adolescents who selfidentified racially as Black and those adoles cents who self-identifie d racially as White (OR = 1.26, 95% CI = 0.99, 1.59). Since independ ent variables were entered into the regression equation, the fixed in tercept became a regression coefficient in this model and represented the mean of the de pendent variable when all th e independent variables were equal to zero. In comparing Model I to Mode l II, the size and signi ficance of the estimate for the random intercept was unchanged, which s uggested that there was still variation in unmet health needs across neighborhoods af ter accounting for individual race. Model III examined the extent to whic h socioeconomic position influenced the report of unmet health needs above and beyond adolescents’ individual race as well as the extent to which socioeconomic position medi ated the relationship between individual

PAGE 129

116 race and unmet health needs. The random effect s for the ratio of in come to poverty ratio and parental educational attainment were non-significant, which indicated that the relationship between socioeconomic position an d the odds of unmet health needs did not vary across neighborhoods.5 The slopes were fixed and the model was reanalyzed with the results presented in Model III of Table 12. The direct effects for socioeconomic pos ition and unmet health needs revealed that in a typical neighborhood, adolescents had significantly greater odds of unmet health needs if their household income was 100% or below their poverty threshold (OR = 1.80, 95% CI = 1.23, 2.64), 201% to 300% over thei r poverty threshold (OR = 1.54, 95% CI = 1.05, 2.28), and 301% to 400% over their povert y threshold (OR = 1.50, 95% CI = 1.02, 2.19) than those adolescents whose household income wa s over 400% of their poverty threshold, after adjusting for adolescents’ race and parental educational attainment. However, in a typical neighborhood, after adjust ing for individual race and the income to poverty ratio, the findings suggested that there were no significant differences in the odds of unmet health needs between those adolescents who had a parent with less than a high school degree, a high degree or GED, or some college education but no college degree and those adolescents who had a parent with a college degree or more. In relation to race, in a typical neig hborhood, after adjusting for socioeconomic position, there were no significant differences in the odds of unmet health needs between those adolescents who self-ide ntified as Black and those ad olescents who self-identified 5 Initially, the slopes of the variables for ratio of income to poverty and parental educational attainment varied randomly in the model simultaneously, but the model failed to converge. The model was then reanalyzed with the slopes for the ratio of income to poverty randomly varying and parental educational fixed, and then again with the slopes for parental educational attainment randomly varying and the ratio of income to poverty fixed. Multilevel models can be constrained by the number of random effects that can be estimated at one time (Raudenbush & Bryk, 2002).

PAGE 130

117 as White (OR = 1.17, 95% CI = 0.91, 1.50). In comparing Model II to Model III, the estimate and significance of the random inte rcept remained relatively unchanged, which indicated that there was still significant variation in the odds of unmet health needs across neighborhoods after accounti ng for individual race, household income, and level of parental educational attainment. In Model IV, the control va riables that represented a dolescents’ health needs, predisposing factors, and enabling factors we re entered into the model. In a typical neighborhood, after adjusting for socioeconom ic position and control variables, there were no significant differences in the odds of unmet health needs between those adolescents who self-identifie d as Black and those adolescen ts who self-identified as White (OR = 1.08, 95% CI = 0.82, 1.42). In addi tion, in the typical neighborhood, after adjusting for adolescent race and the control variables, there were no significant differences in the odds of unmet health needs between those adolescents whose household income to poverty ratio was 400% or less and those adolescents whose household income to poverty ratio was 401% or greater. These findings suggest that the relationship between household income to poverty ratio and the odds of unmet health needs was mediated by the control variables. In comparing Model III to Model IV, the estimate and significance of the random inte rcept were relatively unchanged, which indicated that there was still significant variation in the odds of unmet health needs across neighborhoods after accounti ng for individual race, soci oeconomic position, and the individual control variables.

PAGE 131

118 Table 12 HGLM Analyses: Social Disparities in Unmet Health Needs (N=1,526) Model I Model II Model III Model IV Level I Fixed Effects OR 95% CI OR 95% CI OR 95% CI OR 95% CI Intercept 0.33 0.30, 0.37 0.31 0.27, 0.36 0.23 0.16, 0.32 0.50 0.26, 0.96 Race Black 1.26 0.99, 1.59 1.17 0.91, 1.50 1.08 0.82, 1.42 White (reference) Parental Education Less than High School 0.90 0.57, 1.43 0.77 0.47, 1.27 High School Degree or GED 0.99 0.70, 1.39 0.95 0.66, 1.36 Some College/Technical 1.09 0.77, 1.54 1.04 0.73, 1.48 College Degree or More (reference) Ratio of Income to Poverty 0-100% 1.80 1.23, 2.64 1.39 0.90, 2.16 101-200% 1.20 0.81, 1.78 1.04 0.67, 1.62 201-300% 1.54 1.05, 2.28 1.39 0.92, 2.10 301-400% 1.50 1.02, 2.19 1.43 0.94, 2.17 401% or More (reference) General Health 0.74 0.64, 0.86 STI History 1.43 0.90, 2.25 Pregnancy History 1.03 0.71, 1.51 Age 0.94 0.81, 1.09 Pregnancy Undesirable 1.08 0.91, 1.27 Barriers to Birth Control 1.23 1.03, 1.47 Mom Disapprove of Sex 1.05 0.89, 1.24 Mom Disapprove of Birth Control 0.99 0.89., 1.11 Length of Residence in Years 0.99 0.97, 1.02 Health Insurance 0.88 0.59, 1.32 Driver’s License 0.96 0.71, 1.30 Two Parent Household 0.67 0.51, 0.89 School Education: Medical Attention 0.69 0.50, 0.94 School Health Care NonAthletic Physical 1.06 0.76, 1.50 Random Intercept 0.02 0.02 0.02 0.01 -2LL 4521.96 4518.61 4506.21 4451.46 2 3.34(1) 12.39(7) 54.76(14)

PAGE 132

119 Receipt of Contraceptive Services Table 13 summarizes the results of the models testing Hypothesis I for the dependent variable, receipt of contraceptive services. The re sults of Model I, the null model, revealed that in the typical ne ighborhood, the probability of having received contraceptive services among the sexually experi enced adolescent females in this sample was 48.2%. In addition, the result s of the null model indicated that there was variance across neighborhoods in the odds of havi ng received contraceptive services (T00 = 0.05, p < 0.05). Model II estimated the effects of indivi dual race on the receipt of contraceptive services. The random effect for race was nonsignificant, which indicated that the relationship between race and the odds of ha ving received contraceptive services did not vary across neighborhoods. The slope was fixe d and the model was reanalyzed with the results presented in Model II of Table 13. The results indicated that in a typical neighborhood, there were no signi ficant differences in the odds of having received contraceptive services between those adolesce nts who self-identifie d racially as Black and those adolescents who se lf-identified racially as White (OR = 1.20, 95% CI = 0.97, 1.50). Since independent variables were ente red into the regression equation, the fixed intercept became a regression coefficient in th is model and represented the mean of the dependent variable when all the independent variables were equal to zero. In comparing Model I to Model II, the estimate and si gnificance of the random intercept were unchanged, which indicated that there was stil l significant variance in the odds of having

PAGE 133

120 received contraceptive serv ices across neighborhoods afte r accounting for individual race. Model III examined the extent to whic h socioeconomic position influenced the receipt of contraceptive services above and be yond adolescents’ individual race as well as the extent to which socioeconomic position me diated the relationship between individual race and the receipt of contraceptive servic es. The random effects for household income and parental educational attainment were non-significant, which indicated that the relationship between socioeconomic pos ition and the odds of having received contraceptive services did not vary across neighborhoods.5 The slopes were fixed and the model was reanalyzed with the results presented in Model III of Table 13. The direct effects for socioeconomic pos ition and having received contraceptive services revealed that in a typica l neighborhood the odds of having received contraceptive services were lower for t hose adolescents whose household income was 201-300% above their poverty threshold than for those adolescents whose household income was 401% above their poverty thre shold (OR = 0.67, 95% CI = 0.48, 0.94). In a typical neighborhood, the odds of having received contraceptive services were also lower for those adolescents whose household income was 301-400% above their poverty threshold than for those adolescents whos e household income was over 400% of their poverty threshold, after adjusting for adoles cents’ race and parental educational attainment (OR = 0.64, 95% CI = 0.44, 0.93). In relation to pare ntal educational 5 Initially, the slopes of the variables for ratio of income to poverty and parental educational attainment varied randomly in the model simultaneously, but the model failed to converge. The model was then reanalyzed with the slopes for the ratio of income to poverty randomly varying and parental educational fixed, and then again with the slopes for parental educational attainment randomly varying and the ratio of income to poverty fixed. Multilevel models can be constrained by the number of random effects that can be estimated at one time (Raudenbush & Bryk, 2002).

PAGE 134

121 attainment, the results revealed that in a typical neighborhood ther e were no significant differences in the odds of having receiv ed contraceptive serv ices between those adolescents whose parents had a high de gree or GED (OR = 1.20, 95% CI = 0.89, 1.63) or those adolescents whose parents had some college education or technical training (OR = 1.16, 95% CI = 0.86, 1.63) and those adolescents whose parents had a college degree or more, after adjusting for individual race a nd the income to poverty ratio. However, adolescents whose parents had less than a hi gh school degree had significantly greater odds of having received contraceptive services than those adolescents whose parents had a college degree or more, after adjusting fo r individual race and the income to poverty ratio (OR = 1.87, 95% CI = 1.24, 2.82). The results revealed that in a typi cal neighborhood, after controlling for socioeconomic position, there were no signifi cant differences in the odds of having received contraceptive services between those adolescents who self-identified racially as Black and those adolescents w ho self-identified racially as White (OR = 1.15, 95% CI = 0.90, 1.47). In comparing Model II to Model I II, the estimate and significance of the random intercept slightly increased, which indicated that there was still significant variance in the odds of having received cont raceptive services across neighborhoods after accounting for individual race and socioeconomic position. In Model IV, the control va riables that represented a dolescents’ health needs, predisposing factors, and enabling factors were entered into the model. After the inclusion of these control variables into the model, there were no significant differences in the odds of having received contraceptive se rvices between those adolescents who selfidentified racially as Black and those adoles cents who self-identifie d racially as White

PAGE 135

122 (OR = 1.05, 95% CI = 0.79, 1.38). In addition, in a typical neighborhood the differences in the odds of having received contraceptive services between those adolescents whose household income was 301-400% above their p overty threshold and those adolescents whose household income was over 400% of their poverty threshold were no longer significant (OR = 0.68, 95% CI = 0.44, 1.05). However, in the typical neighborhood, after adjusting for individual race, parental educational attainment and the control variables, adolescents w hose household income was 201-300% above their poverty threshold had significantly lower odds of ha ving received contraceptive services than those adolescents whose household income was over 400% of their poverty threshold, (OR = 0.68, 95% CI = 0.47, 0.99). In addition, th ose adolescents who had a parent with less than a high school educati on continued to have signific antly greater odds of having received contraceptive services than those adolescents who had a parent with a college education or more, after adjusting for indivi dual race, household income, and the control variables (OR = 1.76, 95% CI = 1.12, 2.77). In comparing Model III to Model IV, the size and significance of the random intercept slightly increased, wh ich indicated that there was still significant va riance in the odds of having re ceived contraceptive services across neighborhoods after accounting for indi vidual race, socioeconomic position, and the individual control variables.

PAGE 136

123 Table 13 HGLM Analyses: Social Disparities in the R eceipt of Contraceptiv e Services (N=1,526) Model I Model II Model III Model IV Level I Fixed Effects OR 95% CI OR 95% CI OR 95% CI OR 95% CI Intercept 0.93 0.84, 1.04 0. 88 0.77, 1.01 0.89 0.69, 1.16 0.52 0.28, 0.97 Race Black 1.20 0.97, 1.50 1.15 0.90, 1.47 1.05 0.79, 1.38 White (reference) Parental Education Less than High School 1.87 1.24, 2.82 1.76 1.12, 2.77 High School Degree or GED 1.20 0.89, 1.63 1.13 0.79, 1.62 Some College/Technical 1.16 0.86, 1.56 1.10 0.78, 1.56 College Degree or More (reference) Ratio of Income to Poverty 0-100% 0.95 0.67, 1.33 0.91 0.61, 1.36 101-200% 0.84 0.60, 1.17 0.77 0.51, 1.15 201-300% 0.67 0.48, 0.94 0.68 0.47, 0.99 301-400% 0.64 0.44, 0.93 0.68 0.44, 1.05 401% or More (reference) General Health 1.03 0.91, 1.18 STI History 3.91 2.25, 6.80 Pregnancy History 2.32 1.63, 3.30 Age 1.12 0.98, 1.27 Pregnancy Undesirable 0.84 0.72, 0.99 Barriers to Birth Control 0.77 0.64, 0.93 Mom Disapprove of Sex 0.81 0.69, 0.94 Mom Disapprove of Birth Control 0.75 0.67, 0.83 Length of Residence in Years 0.99 0.98, 1.02 Health Insurance 1.40 0.90, 2.17 Driver’s License 1.53 1.15, 2.05 Two Parent Household 0.70 0.53, 0.93 School Education: Medical Atten tion 1.13 0.81, 1.58 School Health Care NonAthletic Physical 1.01 0.72, 1.42 Random Intercepts 0.05 0.05 0.08 0.11 Intercepts -2LL 4917.12 4914.42 4890.47 4661.85 2 2.70(1) 23.95(7) 228.62(14)

PAGE 137

124 Hypothesis II The aims of Hypothesis II were: (1) to examine the association between the average odds in access to health care among se xually experienced adolescent females and the neighborhood racial and soci oeconomic context, and (2) to determine the extent to which the neighborhood racial and socioeconomic context explained the variation in the average odds of access to health care acro ss neighborhoods. A series of five nested models were employed to test Hypothesis II for each of the three dependent variables – receipt of a routine physical, unm et health needs, and receipt of contraceptive services. Model I: The null model. There were no independent variables in the model. The Level I intercept varied ra ndomly across the neighborhoods so the extent to which the dependent variable varied across th e neighborhoods could be ascertained. Model II: The Level II neighborhood racial context variable – the proportion of Black residents within the census tract, was entered into the mode l and measured as a continuous variable. The variable was grand mean centered. The slopes of the Level II independent variables were fixed by default.6 The Level I intercept varied randomly across the neighborhoods. Model III: The Level II socioeconomic context variables – the proportion of residents within the census tract that are below the 1989 poverty threshold and the proportion of residents within the census trac t who are 25 years of age and older without a high school degree, were entered into the model. Both variables were continuous in 6 Since this study examined a two-level HGLM – adolescents nested within neighborhoods, the slopes of the Level II independent variables were fixed by default as there were no higher level of analyses over which the Level II slopes could vary (Raundenbush & Bryk, 2002). Thus, the parameter estimate for a Level II independent variable represented the mean or average value for that independent variable across all neighborhoods, adjusting for the other independent variables in the model.

PAGE 138

125 nature and grand mean centered. The Leve l I intercept varied randomly across the neighborhoods. The direct and indirect e ffects of the neighborhood socioeconomic context on the average odds of access to health care were examined. Model IV: The Level II control variables were entered into the model. All Level II control variables were measured as c ontinuous variables and grand mean centered, except geographic region, which was measured as three categorical variables that were dummy coded and uncentered. The Level I intercept varied randomly across the neighborhoods. Model V: The Level II interaction va riables between the neighborhood racial context and the neighborhood socioeconomic context – proportion of the neighborhood 25 years of and older without a high school degree-was entered into the model. The interaction variable was continuous in na ture and grand mean centered. The Level I intercept varied randomly across the neighborhoods. Model VI: The Level II interaction va riables between the neighborhood racial context and the neighborhood socioeconomic context – proportion of the neighborhood below the 1989 poverty threshold-was entered in to the model. The in teraction variable was continuous in nature and grand mean cen tered. The Level I intercept varied randomly across the neighborhoods. Receipt of a Routine Physical Table 14 summarizes the results of the models testing Hypothesis II for the dependent variable, receipt of a routine phys ical. The null model, which was analyzed with the testing of Hypothesis I, was the first model to be examined in the testing of

PAGE 139

126 Hypothesis II. As reported earlier, the averag e odds of having receive d a routine physical versus not having received a routine physical were 2.25 (95% CI = 2.00, 2.52) across all the neighborhoods. Thus, in the typical nei ghborhood, the probability of having received a routine physical among the adolescents in this study was 69%. The results of the null model revealed that the average odds of ha ving received a routine physical varied across neighborhoods (T00 = 0.09, p < 0.05). In Model II, the variable that repres ented the neighborhood racial context – the proportion of Black residents within the neighborhood was entered into the model. Model II examined the association between the proportion of Black residents in the neighborhood and the average odds of havi ng received a routine physical across neighborhoods, but the results revealed ther e was no significant relationship (OR = 1.41, 95% CI = 0.94, 2.11). Furthermore, when comp aring Model II to Model I the size and significance of the variation in the average odds of having received a routine physical across neighborhoods were unchanged (T00 = 0.08, p < 0.05), which indicated that the proportion of Black residents in the neighborhood did not account for the variation in the average odds of access to health care across neighborhoods. In Model III, the Level II variab les that represented the neighborhood socioeconomic context – the proportion of residents below the 1989 poverty threshold and the proportion of residents who were 25 ye ars of age or older without a high school degree were entered into the model to exam ine the relationship between the average odds of having a routine physical and the neighbor hood socioeconomic context. Only direct effects of the neighborhood socioeconomic cont ext were assessed in the model since the relationship between the neighbor hood racial context and the average odds of receiving a

PAGE 140

127 routine physical was insignifi cant. After adjusting for th e neighborhood racial context and the proportion of lower educated reside nts in the neighborhood, the results of Model III revealed that the proportion of residents within the neighborhood that lived below the 1989 poverty threshold was not associated with the average odds of having received a routine physical (OR = 2.24, 95% CI = 0.35, 16.97) In addition, after adjusting for the neighborhood racial context a nd the proportion of the poor re sidents in the neighborhood, there was no relationship between the averag e odds of having receive d a routine physical and the proportion of residents within th e neighborhood who were 25 years of age or older without a high school degr ee (OR = 0.42, 95% CI = 0.09, 1.97). When comparing Model III to Model II, th e size and significance of the variation in the average odds of having received a routine physical across neighborhoods was slightly lower, but significant (T00 = 0.08, p < 0.05). These findings indicated that after the addition of the neighborhood socioeconomic variables, the r eceipt of a routine physical varied across neighborhoods, t hus further modeling was indicated. Model IV included the additi on of the neighborhood level control variables – total population of residents within the neighborhood, median age of the re sidents within the neighborhood, residential st ability of the neighbor hood, the proportion of the neighborhood that was considered an urba n area, and geographi c area. The results revealed no significant relations hip in the odds of having rece ived a routine physical and the neighborhood racial and socioeconomic c ontext, after the control variables were entered into the model. Furthermore, the onl y significant control variable was that the odds of having received a routine physical we re greater for those adolescents who lived in the Northeast than those adolescents who lived in th e Midwest (OR = 2.63, 95% CI =

PAGE 141

128 1.67, 4.13). When comparing Model V to Model IV, the size and significance of the variation in the average odds of having received a rou tine physical acr oss neighborhoods was substantially lower, yet significant (T00 = 0.001, p < 0.05). These findings suggest that the only significant contro l variablegeographic regi on, explained a considerable proportion of the variation in the average odds of having received a routine physical across neighborhoods, but there was fu rther variance to be explained. In Model V and Model VI, interactions were examined to determine the extent to which the relationship between the neighborhood racial context and the average odds of having received a routine physical vari ed as a function of the neighborhood socioeconomic context. In Model V, the in teraction variablethe proportion of Black residents in the neighborhood by the proportion of residents within the neighborhood that were lower educatedwas entered into th e model. The findings revealed that the relationship between the proportion of Black residents within the neighborhood and the average odds of having received a routine phys ical did not vary based on the proportion of the residents within the neighbor hood that were lower educated ( 2 = 0.19(1), p > 0.05). The results of the random effect for the Level I intercept indicated that the average odds of having received a ro utine physical varied across neighborhoods (T00 = 0.001, p < 0.05). However, the size of the parameter estimate for the Level I intercept was unchanged from Model IV to Model V, which i ndicated that the inte raction variable did not account for the variation in the averag e odds of having receive d a routine physical across neighborhoods. In Model VI, the interaction variablethe proportion of Black residents in the neighborhood by the proportion of residents wi thin the neighborhood that were below the

PAGE 142

129 poverty thresholdwas entered into the mode l. The interaction va riable from Model V was removed from the regression equation due to non-significant findings. The results of Model VI revealed that the relationship betw een the proportion of Black residents within the neighborhood and the average odds of having received a r outine physical did not vary based on the proportion of the residents with in the neighborhood that were below the poverty threshold ( 2 = 0.001(1), p > 0.05). The results of the random effect for the Level I intercept indicated that the average odds of having receiv ed a routine physical varied across neighborhoods (T00 = 0.001, p < 0.05). However, the size of the parameter estimate for the Level I intercept was uncha nged from Model IV to Model VI, which indicated that the intera ction variable did not account for the variation in the average odds of having received a routine physical across neighborhoods.

PAGE 143

130 Table 14 HGLM Analyses: Contribution of the Neighborhood Context to the Receipt of a Routine Physical (N=1,526) Model I Model II Model III Model IV Model V Model VI OR 95% CI OR 95% CI OR 95% CI OR 95% CI OR 95% CI OR 95% CI Fixed Effects Level I Intercept 2.25 2.00-2.52 2.28 2.02-2.56 2.29 2.04-2.58 2.50 1.99-3.16 2.50 1.993.16 2.50 1.98-3.16 Level II % Black Residents 1.41 0.94-2.11 1.35 0. 84-2.17 1.46 0.84-2.52 1.18 0.34-4.09 1.47 0.68-3.21 % Poor Residents 2.44 0.35-16. 97 4.91 0.54-45.06 4.51 0. 46-44.13 5.06 0.29-87.81 % Low Educated Residents 0.42 0. 09-1.97 0.34 0.06-1.86 0.30 0.05-1.76 0.33 0.06-1.96 % Black Residents X % Poor Residents 0.94 0.03-33.47 % Black Residents X % Low Educated Residents 1.91 0.06-66.27 Total Population 1.00 1.00-1.00 1.00 1.001.00 1.00 1.00-1.00 Median Age of Residents 1.01 0.97-1.05 1.01 0. 97-1.05 1.01 0.97-1.05 Residential Stability 0. 68 0.16-2.93 0.71 0.163.07 0.68 0.16-2.94 Proportion Residents in Urban Area 1.13 0.84-2.52 1.14 0.84-1.55 1.13 0.83-1.54 West 0.69 0.47-1. 01 0.70 0.47-1.02 0.69 0.47-1.01 Northeast 2.63 1.67-4.13 2.62 1. 66-4.12 2.63 1.66-4.15 South 0.79 0.57-1. 08 0.79 0.57-1.08 0.79 0.57-1.08 Midwest (reference) Random Effects Intercepts 0.09 0.08 0.08 0.001 0.001 0.001 Model Fit -2LL 4705.34 4702.18 4700.64 4641.31 4641.12 4641.31 2 3.15(1) 1.54(2) 59.33(7) 0.19(1) 0.001(1)

PAGE 144

131 Unmet Health Needs Table 15 summarizes the results of the models testing Hypothesis II for the dependent variable, reported unmet health needs. The null model, which was analyzed with the testing of Hypothesis I, was the first model to be examined in the testing of Hypothesis II. As reported earlier, the pr obability of unmet h ealth needs among the adolescents in this sample in the typical neighborhood was 24.8%. The results of the null model revealed that the average odds of re porting unmet health n eeds varied slightly across neighborhoods (T00 = 0.02, p < 0.05). In Model II, the variable that repres ented the neighborhood racial context – the proportion of Black residents within the ne ighborhood was entered into the model. Model II examined the association between the proportion of Black residents in the neighborhood and the average odds of unmet hea lth needs, but the re sults revealed there was no significant relationship (OR = 1.42, 95% CI = 0.99, 2.03). Furthermore, when comparing Model II to Model I, the size and significance of the variation in the average odds of unmet health needs acro ss neighborhoods were unchanged (T00 = 0.01, p < 0.05). In Model III, the Level II variab les that represented the neighborhood socioeconomic context – the proportion of residents below the 1989 poverty threshold and the proportion of residents who were 25 ye ars of age or older without a high school degreewere entered into the model to ex amine the relationships between the average odds of unmet health needs and the neighbor hood socioeconomic context. Only direct effects of the neighborhood socioeconomic cont ext were assessed in the model since the relationship between the nei ghborhood racial context and the average odds of unmet

PAGE 145

132 health needs was insignificant. After adjust ing for the neighborhood racial context and the proportion of the lower educated resident s in the neighborhood, the results of Model III revealed that the proportion of residents within the neighborhood that lived below the 1989 poverty threshold was not significantly as sociated with the average odds of unmet health needs across neighborhoods (OR = 3.16, 95% CI = 0.46, 21.61). Likewise, after adjusting for the neighborhood racial context and the proportion of poor residents within the neighborhood, the results also suggested that the proportion of residents in the neighborhood 25 years of age and older without a high school degree was not significantly associated w ith the average odds of un met health needs across neighborhoods (OR = 0.27, 95% CI = 0.06, 1.31). When comparing Model III to Model II, the size and significance of the variation in the average odds of unmet health needs were unchanged and significant (T00 = 0.01, p < 0.05). These findings indicated that the neighborhood socioeconomic context did not account for the varia tion in the average odds of unmet health needs across neighborhoods. Model IV included the additi on of the neighborhood level control variables – total population of residents within the neighborhood, median age of the re sidents within the neighborhood, residential st ability of the neighbor hood, the proportion of the neighborhood that was considered an urba n area, and geographic region. The results revealed no significant relations hips between the average odds of unmet health needs and the proportion of Black residents in th e neighborhood (OR = 1.29, 95% CI = 0.76, 2.19), the proportion of poor residents in th e neighborhood (OR = 3.87, 95% CI = 0.47, 31.82), or the proportion of lower educated resi dents in the neighborhood (OR = 0.20, 95% CI = 0.03, 1.24), after adjusting for one another and the control variables. Furthermore, none

PAGE 146

133 of the control variables were significantly related to the average odds of unmet health needs. When comparing Model IV to Model V, the size and significan ce of the variation in the average odds of unmet health need s across neighborhoods were unchanged and significant (T00 = 0.01, p < 0.05). The overall findings of these models suggest that the neighborhood racial and socioeconomic cont ext as well as the neighborhood control variables did not explain the variation in th e average odds of unmet health needs across neighborhoods for this sample of adolescents. In Model V and Model VI, interactions were examined to determine the extent to which the relationship between the neighborhood racial context and the average odds of unmet health needs varied as a function of the neighborhood socioeconomic context. In Model V, the interaction vari ablethe proportion of Black residents in the neighborhood by the proportion of resident s within the neighborhood that were lower educatedwas entered into the model. The findings revealed that the relationshi p between the proportion of Black residents within th e neighborhood and the average od ds of unmet health needs did not vary based on the proportion of the re sidents within the ne ighborhood that were lower educated ( 2 = 0.03(1), p > 0.05). The results of the ra ndom effect for the Level I intercept indicated that the average odds of unmet health needs varied across neighborhoods (T00 = 0.01, p < 0.05). However, the size of the parameter estimate for the Level I intercept was relatively unchanged from Model IV to Model V, which indicated that the interaction variable did not account for the variati on in the average odds of unmet health needs across neighborhoods. In Model VI, the interaction variablethe proportion of Black residents in the neighborhood by the proportion of residents wi thin the neighborhood that were below the

PAGE 147

134 poverty thresholdwas entered into the mode l. The interaction va riable from Model V was removed from the regression equation due to non-significant findings. The results of Model VI revealed that the relationship betw een the proportion of Black residents within the neighborhood and the average odds of unmet health need s did not vary based on the proportion of the residents within the ne ighborhood that were below the poverty threshold ( 2 = 0.41(1), p > 0.05). The results of the random effect for the Level I intercept indicated that the average odds of unmet health needs varied across neighborhoods (T00 = 0.01, p < 0.05). However, the size of the paramete r estimate for the Level I intercept was unchanged from Model IV to Model VI, which in dicated that the inte raction variable did not account for the variation in the aver age odds of unmet health needs across neighborhoods.

PAGE 148

135 Table 15 HGLM Analyses: Contribution of the Neighborhood Context to Unmet Health Needs (N=1,526) Model I Model II Model III Model IV Model V Model VI OR 95% CI OR 95% CI OR 95% CI OR 95% CI OR 95% CI OR 95% CI Fixed Effects Level I Intercept 0.33 0.30, 0.37 0.34 0.30, 0.38 0.34 0.30, 0.38 0.32 0.26, 0.39 0.32 0.26, 0.39 0.32 0.26, 0.39 Level II % Black Residents 1.42 0.99, 2.03 1.38 0.85, 2.24 1.29 0.76, 2.19 1.19 0.29, 4.95 1.08 0.44, 2.65 % Poor Residents 3.16 0.46, 21.61 3.87 0.47, 31.82 3.73 0.43, 32.09 2.28 0.16, 32.68 % Low Educated Residents 0.27 0.06, 1.31 0.20 0.03, 1.24 0.19 0. 03, 1.28 0.22 0.04, 1.39 % Black Residents X % Poor Residents 2.78 0.06, 132.34 % Black Residents X % Low Educated Residents 1.27 0.03, 58.54 Total Population 1.00 1.00, 1.00 1.00 1.00, 1.00 1.00 1.00, 1.00 Median Age of Residents 0. 98 0.94, 1.02 0.98 0.94, 1.02 0.98 0.94, 1.02 Residential Stability 2.28 0.57, 9.12 2.31 0.58, 9.23 2.38 0.60, 9.55 Proportion Residents in Urban Area 0.98 0.74, 1.29 0. 98 0.74, 1.30 0.99 0.74, 1.32 West 1.24 0.83, 1. 84 1.24 0.83, 1.84 1.25 0.77, 1.41 Northeast 1.05 0.70, 1.58 1.04 0.69, 1.58 1.03 0.68, 1.56 South 1.03 0.76, 1. 39 1.03 0.76, 1.39 1.04 0.77, 1.41 Midwest (reference) Random Effects Intercepts 0.02 0.01 0.01 0.01 0.01 0.01 Model Fit -2LL 4521.96 4518.46 4515.12 4512.44 4512.41 4512.03 2 3.49(1) 3.34(2) 2.69(7) 0.03(1) 0.41(1)

PAGE 149

136 Receipt of Contraceptive Services Table 16 summarizes the results of the models testing Hypothesis II for the dependent variable, received contraceptive services. The nu ll model, which was analyzed with the testing of Hypothesis I, was the first model to be examined in the testing of Hypothesis II. As reported previously, the pr obability of having received contraceptive services among this sample of adolescen ts in a typical neighborhood was 48.2%. The results of the null model revealed that the average odds of having received contraceptive services varied across neighborhoods (T00 = 0.05, p < 0.05). In Model II, the variable that repres ented the neighborhood racial context – the proportion of Black residents within the neighborhoodwas entered into the model. Model II examined the association between the proportion of Black residents in the neighborhood and the average odds of having received contraceptiv e services, but the results revealed no significant rela tionship (OR = 1.25, 95% CI = 0.90, 1.74). Furthermore, when comparing Model II to Mo del I, the size and significance of the variation in the average o dds of having received cont raceptive services across neighborhoods was unchanged (T00 = 0.05, p < 0.05). In Model III, the Level II variab les that represented the neighborhood socioeconomic context – the proportion of residents below the 1989 poverty threshold and the proportion of residents who were 25 ye ars of age or older without a high school degree were entered into the model to exam ine the relationships between the average odds of having received contraceptive se rvices and the neighborhood socioeconomic context. Only direct effects of the neighbor hood socioeconomic context were assessed in

PAGE 150

137 the model since the relationship between th e neighborhood racial context and the average odds of having received contraceptive serv ices across neighborhoods was insignificant. After adjusting for the neighborhood racial co ntext and the proportion of lower educated residents in the neighborhood, the results of Model III revealed that the proportion of residents within the neighborhood that live d below the 1989 poverty threshold was not associated with the average odds of havi ng received contraceptive services (OR = 0.58, 95% CI = 0.12, 2.91). Furthermore, after ad justing for the neighborhood racial context and the proportion of the poor residents in the neighborhood, the result s indicated that the proportion of residents in th e neighborhood 25 years of age and older without a high school degree was not associated with the av erage odds of having received contraceptive services (OR = 2.57, 95% CI = 0.73, 9.07). When comparing Model III to Model II, the size and significance of the variation in the average odds of having received contraceptive services acro ss neighborhoods was unchanged (T00 = 0.09, p < 0.05). These findings indicated that the aver ages odds of having received contraceptive services varied across neighborhoods after th e addition of the neighborhood socioeconomic variables, thus further modeling was indicated. In Model IV, the neighborhood level c ontrol variables – total population of residents within the neighborhood, median age of the residents within the neighborhood, residential stability of the neighborhood, th e proportion of the neighborhood that was considered an urban area, and geographic region-were entered into the regression equation. The results revealed no relations hips between the average odds of having received contraceptive servi ces and the proportion of Black residents in the neighborhood (OR = 1.19, 95% CI = 0.77, 1.85), the proportion of poor residents in the neighborhood

PAGE 151

138 (OR = 0.48, 95% CI = 0.09, 2.71), or the proporti on of lower educated residents in the neighborhood (OR = 3.43, 95% CI = 0.86, 13.65), afte r adjusting for one another and the control variables. Furthermore, none of the c ontrol variables were si gnificantly related to the average odds of having received contracep tive services. When comparing Model IV to Model III, the size and significance of the variation in the average odds of received contraceptive services across nei ghborhoods decreased and significant (T00 = 0.01, p < 0.05), which suggested that the average odds of having received contraceptive services varied across neighborhoods. The overall findings of these m odels suggest that despite the non-significant findings, the relationships between the neighborhood racial context, the neighborhood socioeconomic context, and the neighborhood control variables explained some of the variation in the average odds of having received contraceptive services for this samp le of adolescents. In Model V and Model VI, interactions were examined to determine the extent to which the relationship between the neighborhood racial context and the average odds of having received contraceptive services va ried as a function of the neighborhood socioeconomic context. In Model V, the in teraction variablethe proportion of Black residents in the neighborhood by the proportion of residents within the neighborhood that were lower educatedwas entered into th e model. The findings revealed that the relationship between the proportion of Black residents within the neighborhood and the average odds of having received contraceptiv e services did not vary based on the proportion of the residents within the ne ighborhood that were lower educated ( 2 = 0.40(1), p > 0.05). The results of the random effect for the Level I inter cept indicated that the average odds of having received contra ceptive services varied across neighborhoods

PAGE 152

139 (T00 = 0.01, p < 0.05). However, the size of the parameter estimate for the Level I intercept was unchanged from Model IV to Mode l V, which indicated that the interaction variable did not account for the variation in the aver age odds of having received contraceptive services across neighborhoods. In Model VI, the interaction variablethe proportion of Black residents in the neighborhood by the proportion of residents wi thin the neighborhood that were below the poverty thresholdwas entered into the mode l. The interaction va riable from Model V was removed from the regression equation due to non-significant findings. The results of Model VI revealed that the relationship betw een the proportion of Black residents within the neighborhood and the average odds of having received cont raceptive services did not vary based on the proportion of the reside nts within the neighbor hood that were below the poverty threshold ( 2 = 1.19(1), p > 0.05). The results of the random effect for the Level I intercept indi cated that the average odds of un met health needs varied across neighborhoods (T00 = 0.01, p < 0.05). However, the size of the parameter estimate for the Level I intercept was unchange d from Model IV to Model VI which indicated that the interaction variable did not account for th e variation in the average odds of having received contraceptive serv ices across neighborhoods.

PAGE 153

140 Table 16 HGLM Analyses: Contribution of the Neighborhood Context to the Receipt of Contraceptive Services (N=1,526) Model I Model II Model III Model IV Model V Model VI OR 95% CI OR 95% CI OR 95% CI OR 95% CI OR 95% CI OR 95% CI Fixed Effects Level I Intercept 0.93 0.84, 1.04 0.93 0. 85, 1.04 0.93 0.84, 1.04 0.95 0.79, 1.14 0.95 0.79, 1.14 0.94 0.78, 1.13 Level II % Black Residents 1.25 0.90, 1.74 1.20 0.79, 1.82 1.19 0.77, 1.85 0.90 0.36, 2.28 0.91 0.47, 1.76 % Poor Residents 0.58 0.12, 2.91 0.48 0.09, 2.71 0.43 0.07, 2.55 0.23 0.02, 2.13 % Low Educated Residents 2.57 0.73, 9.07 3.43 0.86, 13.65 2.94 0. 71, 12.15 3.97 0.95, 16.56 % Black Residents X % Poor Residents 4.85 0.33, 70.71 % Black Residents X % Low Educated Residents 2.37 0.20, 28.26 Total Population 1.00 1.00, 1.00 1.00 1.00, 1.00 1.00 1.00, 1.00 Median Age of Residents 0. 99 0.96, 1.02 0.99 0.96, 1.02 0.99 0.96, 1.02 Residential Stability 0.82 0.25, 2.71 0.86 0.26, 2.83 0.88 0.27, 2.89 Proportion Residents in Urban Area 1.05 0.81, 1.37 1. 06 0.81, 1.38 1.07 0.82, 1.41 West 1.21 0.87, 1. 68 1.21 0.87, 1.70 1.22 0.88, 1.72 Northeast 1.03 0.73, 1.45 1.02 0.73, 1.44 1.00 0.71, 1.42 South 0.92 0.70, 1. 21 0.92 0.70, 1.22 0.94 0.71, 1.23 Midwest (reference) Random Effects Intercepts 0.05 0.05 0.05 0.01 0.01 0.01 Model Fit -2LL 4917.12 4915.50 4912.99 4907.73 4907.33 4906.54 2 1.62(1) 2.51(2) 5.26(7) 0.40(1) 1.19(1)

PAGE 154

141 Hypothesis III The aim of Hypothesis III was to examine the extent to which racial and socioeconomic disparities in access to h ealth care varied as a function of the neighborhood racial and socioeconomic context, which tested a cross-level interaction between Level I and Level II variables. Th e models for Hypothesis III build on the models from Hypothesis I, which examined th e extent to which social disparities in access to health care existed among the sexuall y experienced adolescent females in this study. A series of three models were analyzed to test Hypothesis III for each of the dependent variables representing access to health carereceived a routine physical, reported unmet health needs, and received contraceptive services. Model I: All of the Level I variables ar e entered into the model. The continuous control variables were grand mean centered and all categorical variables, including the Level I variables in the crosslevel interaction – race, leve l of parental education, and ratio of household income to povertywere uncentered due to their dichotomous nature. The Level II variable representing the neighbor hood racial contextproportion of Black residents within the neighborhoodwas entere d into the model and grand mean centered. The Level I intercept and the slopes of th e parameters for race, level of parental educational attainment and ratio of househol d income to poverty va ried randomly across the neighborhoods. Model I was compared to Model 4 from Hypothesis I via likelihood ratio testing (Chi-Square) to de termine if the cross-level interactions were significant. Model II: All of the Level I variables are entered into the model. The continuous control variables were grand mean centered and all categorical variables, including the Level I variables in the crosslevel interaction – race, leve l of parental education, and

PAGE 155

142 ratio of household income to povertywere uncentered due to their dichotomous nature. The Level II variable representing the nei ghborhood socioeconomic contextproportion of residents in the neighborhood who were 25 years of age and older without a high school degreewas entered into the mode l and grand mean centered. The Level I intercept and the slopes of the parameters for race, level of parental educational attainment and ratio of household income to poverty varied randomly across the neighborhoods. Model II was compared to M odel X from Hypothes is I via likelihood ratio testing (Chi-Square) to de termine if the cross-level interactions were significant. Model III: All of the Level I variables are entered into the model. The continuous control variables were grand mean centered and all categorical variables, including the Level I variables in the crosslevel interaction – race, leve l of parental education, and ratio of household income to povertywere uncentered due to their dichotomous nature. The Level II variable representing the nei ghborhood socioeconomic contextproportion of residents within the ne ighborhood who were living below the 1989 poverty thresholdwas entered into the model and grand mean cen tered. The Level I intercept and the slopes of the parameters for race, level of parental educational attainment and ratio of household income to poverty varied randomly across th e neighborhoods. Model III was compared to Model IV from Hypothesis I via likelihood ratio testing (Chi-Square) to determine if the cross-level interactions were significant. Receipt of a Routine Physical In Model I from Hypothesis III, the random effects for the cross-level interactions between the neighborhood racial context and individual race and socioeconomic position were non-significant. Thus, the slopes were fi xed and the model was reanalyzed with the

PAGE 156

143 results presented in Model I of Table 17. The results of the analysis revealed that the fixed effects for the cross-le vel interactions between the neighborhood racial context and individual race and socioeconomic position we re non-significant, as the overall model was not significant ( 2 =9.92(10), p > 0.05) when comparing Model I to Model IV from Hypothesis I. Thus, in the typical neighborhood, social disparities in the odds of having received a routine physical were not depende nt on the neighborhood racial context among this sample of sexually expe rienced adolescent females.

PAGE 157

144 Table 17 HGLM Analyses: Contribution of the Proporti on of Black Residents in the Neighborhood to Social Disparities in the Receipt of a Routine Physical (N=1,526) Model I OR 95% CI Fixed Effects Level I Variables Intercept 1.35 0.66, 2.76 Race Black 1.01 0.63, 1.61 White (reference) Parental Education Less than High School 0.80 0.50, 1.27 High School Degree or GED 0.74 0.50, 1.27 Some College/Technical 0.72 0.50, 1.05 College Degree or More (reference) Ratio of Income to Poverty 0-100% 0.97 0.62, 1.51 101-200% 0.82 0.54, 1.24 201-300% 1.16 0.77, 1.75 301-400% 0.86 0.56, 1.34 401% or More (reference) General Health 1.19 1.04, 1.36 STI History 1.76 1.01, 3.08 Pregnancy History 1.66 1.17, 2.37 Age 0.91 0.79, 1.05 Pregnancy Undesirable 1.02 0.88, 1.18 Barriers to Birth Control 0.91 0.77, 1.08 Mom Disapprove of Sex 0.95 0.80, 1.13 Mom Disapprove of Birth Control 0.87 0.78, 0.97 Length of Residence in Years 0.99 0.97, 1.02 Health Insurance 1.53 1.03, 2.28 Driver’s License 1.33 0.96, 1.85 Two Parent Household 1.02 0.76, 1.37 School Education: Medical Attention 1.22 0.88, 1.69 School Health Care Non-Athlet ic Physical 2.27 1.61, 3.19 Cross Level Interactions Intercept X % Black Residents 3.00 0.48, 18.58 Individual Race X % Black Residents Black X % Black Residents 0.23 0.05, 1.11 White X % Black Residents (reference) Parental Education X % Black Residents Less than High School X % Black Residents 0.91 0.20, 4.07 High School Degree/ GED X % Black Residents 0.66 0.20, 2.22 Some College/Technical X % Black Residents 1.10 0.30, 3.96 College Degree or More X % Black Residents (reference) Ratio of Income to Poverty X % Black Residents 0-100% X % Black Residents 2.36 0.60, 9.30 101-200% X % Black Residents 1.82 0.46, 7.14 201-300% X % Black Residents 2.43 0.56, 10.68 301-400% X % Black Residents 0.64 0.14, 3.03 401% or More X % Black Residents (reference) Random Effects Intercept 0.02 -2LL 4613.41 2 9.92(9)

PAGE 158

145 In Model II from Hypothesis III, th e random effects for the cross-level interactions between the neighborhood socioeconomic c ontext (proportion of low educated residents) and individual race a nd socioeconomic position were non-significant. Thus, the slopes were fixed and the model was reanalyzed with the results presented in Model II of Table 18. The results of the analysis revealed that the fixed effects for the cross-level interactions betw een the neighborhood socioecono mic context (proportion of low educated residents) a nd individual race and soci oeconomic position were nonsignificant, as the overall model was not significant ( 2 =7.36(9), p > 0.05) when comparing Model II to Model IV from Hypot hesis I. Thus, in the typical neighborhood, social disparities in the odds of having rece ived a routine physical were not dependent on the proportion of low educated residents w ithin the neighborhood among this sample of sexually experienced adolescent females.

PAGE 159

146 Table 18 HGLM Analyses: Contribution of the Proporti on of Low Educated Residents in the Neighborhood to Social Disparities in the R eceipt of a Routine Physical (N=1,526) Model II OR 95% CI Fixed Effects Level I Variables Intercept 1.04 0.54, 1.99 Race Black 1.29 0.93, 1.80 White (reference) Parental Education Less than High School 0.77 0.47, 1.26 High School Degree or GED 0.78 0.53, 1.14 Some College/Technical 0.71 0.49, 1.03 College Degree or More (reference) Ratio of Income to Poverty 0-100% 0.91 0.59, 1.42 101-200% 0.82 0.54, 1.23 201-300% 1.07 0.73, 1.57 301-400% 0.91 0.59, 1.40 401% or More (reference) General Health 1.19 1.04, 1.36 STI History 1.76 1.02, 3.04 Pregnancy History 1.66 1.17, 2.35 Age 0.92 0.80, 1.06 Pregnancy Undesirable 1.02 0.88, 1.18 Barriers to Birth Control 0.92 0.77, 1.09 Mom Disapprove of Sex 0.94 0.79, 1.12 Mom Disapprove of Birth Control 0.87 0.79, 0.98 Length of Residence in Years 0.99 0.97, 1.01 Health Insurance 1.60 1.07, 2.39 Driver’s License 1.33 0.96, 1.84 Two Parent Household 1.02 0.76, 1.38 School Education: Medical Attention 1.21 0.87, 1.68 School Health Care Non-Athlet ic Physical 2.15 1.54, 3.01 Cross Level Interactions Intercept X % Low Educated Residents 0.68 0.04, 10.42 Individual Race X % Low Educated Residents Black X % Low Educat ed Residents 1.43 0.13, 15.60 White X % Low Educated Residents (reference) Parental Education X % Low Educated Residents Less than High School X % Low Educated Residents 2.77 0.08, 95.63 High School Degree/ GED X % Lo w Educated Residents 0.57 0.03, 12.01 Some College/Technical X % Low Educated Residents 2.77 0.14, 55.16 College Degree or More X % Low Educated Residents (reference) Ratio of Income to Poverty X % Low Educated Residents 0-100% X % Low Educated Residents 2.26 0.12, 42.32 101-200% X % Low Educated Residents 0.52 0.02, 12.72 201-300% X % Low Educated Residents 7.20 0.28, 186.41 301-400% X % Low Educated Residents 0.47 0.01, 18.75 401% or More X % Low Educated Residents (reference) Random Effects Intercept 0.02 -2LL 4615.97 2 7.36(9)

PAGE 160

147 In Model III from Hypothesis III, th e random effects for the cross-level interactions between the neighborhood socioeconomic c ontext (proportion of poor residents) and individual race and socioec onomic position were non-significant. Thus, the slopes were fixed and the model was reanalyzed with the results presented in Model III of Table 19. The results of the analysis revealed that the fixed effects for the cross-level interactions between the neighborhood socioeconomic c ontext (proportion of poor residents) and individual race and socioec onomic position were non-significant, as the overall model was not significant ( 2 =9.05(9), p > 0.05) when comparing Model III to Model IV from Hypothesis I. Thus, in the ty pical neighborhood, social disparities in the odds of having received a routine physical we re not dependent on th e proportion of poor residents within the neighborhood among this sa mple of sexually experienced adolescent females.

PAGE 161

148 Table 19 HGLM Analyses: Contribution of the Proporti on of Poor Residents in the Neighborhood to Social Disparities in the Receipt of a Routine Physical (N=1,526) Model III OR 95% CI Fixed Effects Level I Variables Intercept 1.13 0.58, 2.20 Race Black 1.22 0.87, 1.72 White (reference) Parental Education Less than High School 0.79 0.50, 1.27 High School Degree or GED 0.77 0.52, 1.13 Some College/Technical 0.72 0.50, 1.05 College Degree or More (reference) Ratio of Income to Poverty 0-100% 0.83 0.53, 1.31 101-200% 0.76 0.49, 1.16 201-300% 1.06 0.71, 1.60 301-400% 0.80 0.50, 1.28 401% or More (reference) General Health 1.20 1.05, 1.37 STI History 1.78 1.04, 3.06 Pregnancy History 1.65 1.16, 2.34 Age 0.91 0.80, 1.05 Pregnancy Undesirable 1.02 0.88, 1.19 Barriers to Birth Control 0.91 0.77, 1.08 Mom Disapprove of Sex 0.94 0.79, 1.13 Mom Disapprove of Birth Control 0.87 0.79, 0.97 Length of Residence in Years 0.99 0.97, 1.01 Health Insurance 1.57 1.06, 2.34 Driver’s License 1.33 0.96, 1.85 Two Parent Household 1.00 0.75, 1.35 School Education: Medical Attention 1.22 0.88, 1.69 School Health Care Non-Athlet ic Physical 2.15 1.55, 3.00 Cross Level Interactions Intercept X % Poor Residents 2.83 0.05, 154.78 Individual Race X % Poor Residents Black X % Poor Residents 0.85 0.05, 14.33 White X % Poor Residents (reference) Parental Education X % Poor Residents Less than High School X % Poor Residents 0.83 0.02, 37.77 High School Degree/ GED X % Poor Residents 0.57 0.02, 17.93 Some College/Technical X % Poor Residents 1.32 0.04, 43.52 College Degree or More X % Poor Residents (reference) Ratio of Income to Poverty X % Poor Residents 0-100% X % Poor Residents 3.09 0.06, 150.81 101-200% X % Poor Residents 1.03 0.02, 71.25 201-300% X % Poor Residents 4.03 0.05, 312.31 301-400% X % Poor Residents 0.04 0.00, 4.52 401% or More X % Poor Residents (reference) Random Effects Intercept 0.02 -2LL 4614.28 2 9.05(9)

PAGE 162

149 Unmet Health Needs In Model I from Hypothesis III, the random effects for the cross-level interactions between the neighborhood racial context and individual race and socioeconomic position were non-significant. Thus, the slopes were fi xed and the model was reanalyzed with the results presented in Model I of Table 20. The results of the analysis revealed that the fixed effects for the cross-le vel interactions between the neighborhood racial context and individual race and socioeconomic position we re non-significant, as the overall model was not significant ( 2 =10.74(9), p > 0.05) when comparing Model I to Model IV from Hypothesis I. Thus, in the typical neighborhood, social disparities in the odds of unmet health needs were not dependent on the pr oportion of Black residents within the neighborhood among this sample of sexuall y experienced adolescent females.

PAGE 163

150 Table 20 HGLM Analyses: Contribution of the Proporti on of Black Residents in the Neighborhood to Social Disparities in Unmet Health Needs (N=1,526) Model I OR 95% CI Fixed Effects Level I Variables Intercept 0.53 0.26, 1.09 Race Black 1.00 0.67, 1.50 White (reference) Parental Education Less than High School 0.77 0.46, 1.30 High School Degree or GED 0.97 0.67, 1.41 Some College/Technical 1.03 0.72, 1.48 College Degree or More (reference) Ratio of Income to Poverty 0-100% 1.30 0.83, 2.05 101-200% 0.99 0.63, 1.58 201-300% 1.27 0.83, 1.97 301-400% 1.35 0.84, 2.20 401% or More (reference) General Health 0.74 0.63, 0.86 STI History 1.46 0.90, 2.37 Pregnancy History 1.02 0.69, 1.50 Age 0.94 0.81, 1.10 Pregnancy Undesirable 1.08 0.91, 1.28 Barriers to Birth Control 1.22 1.02, 1.47 Mom Disapprove of Sex 1.06 0.89, 1.25 Mom Disapprove of Birth Control 1.00 0.89, 1.12 Length of Residence in Years 0.99 0.97, 1.02 Health Insurance 0.87 0.58, 1.32 Driver’s License 0.96 0.71, 1.31 Two Parent Household 0.66 0.50, 0.88 School Education: Medical Attention 0.67 0.49, 0.92 School Health Care Non-Athlet ic Physical 1.01 0.71, 1.44 Cross Level Interactions Intercept X % Black Residents 1.69 0.32, 8.89 Individual Race X % Black Residents Black X % Black Residents 1.97 0.45, 8.72 White X % Black Residents (reference) Parental Education X % Black Residents Less than High School X % Black Residents 1.18 0.29, 4.79 High School Degree/ GED X % Black Residents 1.87 0.64, 5.44 Some College/Technical X % Black Residents 0.68 0.22, 2.09 College Degree or More X % Black Residents (reference) Ratio of Income to Poverty X % Black Residents 0-100% X % Black Residents 0.29 0.08, 1.04 101-200% X % Black Residents 0.47 0.11, 2.06 201-300% X % Black Residents 0.27 0.07, 1.07 301-400% X % Black Residents 0.34 0.06, 1.86 401% or More X % Black Residents (reference) Random Effects Intercept 0.01 -2LL 4440.73 2 10.74(9)

PAGE 164

151 In Model II from Hypothesis III, th e random effects for the cross-level interactions between the neighborhood socioeconomic c ontext (proportion of low educated residents) and individual race a nd socioeconomic position were non-significant. Thus, the slopes were fixed and the model was reanalyzed with the results presented in Model II of Table 21. The results of the analysis revealed that the fixed effects for the cross-level interactions betw een the neighborhood socioecono mic context (proportion of low educated residents) a nd individual race and soci oeconomic position were nonsignificant, as the overall model was not significant ( 2 =7.49(9), p > 0.05) when comparing Model II to Model IV from Hypot hesis I. Thus, in the typical neighborhood, social disparities in the odds of unmet h ealth needs were not dependent on the proportion of low educated residents within the neighborhood among this sample of sexually experienced adolescent females.

PAGE 165

152 Table 21 HGLM Analyses: Contribution of the Proporti on of Low Educated Residents in the Neighborhood to Social Disparities in Unmet Health Needs (N=1,526) Model II OR 95% CI Fixed Effects Level I Variables Intercept 0.48 0.25, 0.95 Race Black 1.15 0.86, 1.55 White (reference) Parental Education Less than High School 0.75 0.43, 1.31 High School Degree or GED 0.96 0.66-1.40 Some College/Technical 1.01 0.71, 1.46 College Degree or More (reference) Ratio of Income to Poverty 0-100% 1.60 0.97, 2.64 101-200% 1.07 0.67, 1.71 201-300% 1.44 0.92, 2.26 301-400% 1.49 0.91, 2.44 401% or More (reference) General Health 0.74 0.63, 0.86 STI History 1.43 0.89, 2.31 Pregnancy History 1.02 0.69, 1.50 Age 0.94 0.81, 1.09 Pregnancy Undesirable 1.06 0.89, 1.26 Barriers to Birth Control 1.22 1.02, 1.46 Mom Disapprove of Sex 1.06 0.89, 1.26 Mom Disapprove of Birth Control 0.99 0.89, 1.12 Length of Residence in Years 0.99 0.97, 1.03 Health Insurance 0.87 0.57, 1.32 Driver’s License 0.96 0.70, 1.31 Two Parent Household 0.66 0.49, 0.88 School Education: Medical Attention 0.69 0.50, 0.95 School Health Care Non-Athlet ic Physical 1.07 0.75, 1.51 Cross Level Interactions Intercept X % Low Educated Residents 0.81 0.04, 17.00 Individual Race X % Low Educated Residents Black X % Low Educat ed Residents 1.54 0.17, 13.70 White X % Low Educated Residents (reference) Parental Education X % Low Educated Residents Less than High School X % Low Educated Residents 1.09 0.02, 50.40 High School Degree/ GED X % Lo w Educated Residents 0.24 0.02, 3.54 Some College/Technical X % Low Educated Residents 0.22 0.01, 3.68 College Degree or More X % Low Educated Residents (reference) Ratio of Income to Poverty X % Low Educated Residents 0-100% X % Low Educated Residents 0.41 0.01, 14.24 101-200% X % Low Educated Residents 1.77 0.05, 67.56 201-300% X % Low Educated Residents 0.59 0.02, 21.71 301-400% X % Low Educated Residents 1.82 0.03, 111.73 401% or More X % Low Educated Residents (reference) Random Effects Intercept 0.01 -2LL 4443.99 2 7.49(9)

PAGE 166

153 In Model III from Hypothesis III, th e random effects for the cross-level interactions between the neighborhood socioeconomic c ontext (proportion of poor residents) and individual race and socioec onomic position were non-significant. Thus, the slopes were fixed and the model was reanalyzed with the results presented in Model III of Table 22. The results of the analysis revealed that the fixed effects for the cross-level interactions between the neighborhood socioeconomic c ontext (proportion of poor residents) and individual race and socioec onomic position were non-significant, as the overall model was not significant ( 2 =6.84(9), p > 0.05) when comparing Model III to Model IV from Hypothesis I. Thus, in the ty pical neighborhood, social disparities in the odds of unmet health needs were not dependent on the proportion of poor residents within the neighborhood among this sample of sexua lly experienced adolescent females.

PAGE 167

154 Table 22 HGLM Analyses: Contribution of the Proporti on of Poor Residents in the Neighborhood to Social Disparities in Unmet Health Needs (N=1,526) Model III OR 95% CI Fixed Effects Level I Variables Intercept 0.56 0.28, 1.09 Race Black 1.10 0.82, 1.47 White (reference) Parental Education Less than High School 0.75 0.44, 1.29 High School Degree or GED 0.89 0.61, 1.31 Some College/Technical 0.97 0.67, 1.41 College Degree or More (reference) Ratio of Income to Poverty 0-100% 1.38 0.85, 2.27 101-200% 0.97 0.61, 1.55 201-300% 1.27 0.81, 1.98 301-400% 1.46 0.89, 2.40 401% or More (reference) General Health 0.74 0.63, 0.86 STI History 1.39 0.86, 2.25 Pregnancy History 1.04 0.71, 1.53 Age 0.94 0.81, 1.10 Pregnancy Undesirable 1.06 0.89, 1.27 Barriers to Birth Control 1.2 1.02, 1.47 Mom Disapprove of Sex 1.06 0.89, 1.26 Mom Disapprove of Birth Control 1.00 0.90, 1.12 Length of Residence in Years 0.99 0.97, 1.02 Health Insurance 0.88 0.58, 1.34 Driver’s License 0.96 0.70, 1.30 Two Parent Household 0.66 0.49, 0.88 School Education: Medical Attention 0.68 0.50, 0.94 School Health Care Non-Athlet ic Physical 1.06 0.75, 1.50 Cross Level Interactions Intercept X % Poor Residents 10.76 0.23, 496.67 Individual Race X % Poor Residents Black X % Poor Residents 2.74 0.22, 8.69 White X % Poor Residents (reference) Parental Education X % Poor Residents Less than High School X % Poor Residents 0.17 0.00, 8.69 High School Degree/ GED X % Poor Residents 0.11 0.00, 3.01 Some College/Technical X % Poor Residents 0.07 0.00, 1.71 College Degree or More X % Poor Residents (reference) Ratio of Income to Poverty X % Poor Residents 0-100% X % Poor Residents 0.14 0.00, 11.21 101-200% X % Poor Residents 0.29 0.00, 26.49 201-300% X % Poor Residents 0.09 0.00, 6.98 301-400% X % Poor Residents 1.59 0.01, 212.31 401% or More X % Poor Residents (reference) Random Effects Intercept 0.01 -2LL 4444.64 2 6.84(9)

PAGE 168

155 Receipt of Contraceptive Services In Model I from Hypothesis III, the random effects for the cross-level interactions between the neighborhood racial context and individual race and socioeconomic position were non-significant. Thus, the slopes were fi xed and the model was reanalyzed with the results presented in Model I of Table 23. The results of the analysis revealed that the fixed effects for the cross-le vel interactions between the neighborhood racial context and individual race and socioeconomic position we re non-significant, as the overall model was not significant ( 2 =6.58(9), p > 0.05) when comparing Model I to Model IV from Hypothesis I. Thus, in the typical neighborhood, social disparities in the odds of having received contraceptive services were not de pendent on the proportion of Black residents within the neighborhood among this sample of sexually experienced adolescent females.

PAGE 169

156 Table 23 HGLM Analyses: Contribution of the Proporti on of Black Residents in the Neighborhood to Social Disparities in the Receipt of Contraceptive Services (N=1,526) Model I OR 95% CI Fixed Effects Level I Variables Intercept 0.53 0.26, 1.08 Race Black 1.12 0.74, 1.70 White (reference) Parental Education Less than High School 1.73 1.08, 2.79 High School Degree or GED 1.11 0.77, 1.60 Some College/Technical 1.04 0.72, 1.50 College Degree or More (reference) Ratio of Income to Poverty 0-100% 0.85 0.56, 1.30 101-200% 0.74 0.48, 1.13 201-300% 0.66 0.44, 0.99 301-400% 0.66 0.40, 1.09 401% or More (reference) General Health 1..03 0.90, 1.17 STI History 4.00 2.26, 7.06 Pregnancy History 2.29 1.60, 3.29 Age 1.12 0.98, 1.28 Pregnancy Undesirable 0.84 0.72, 0.99 Barriers to Birth Control 0.77 0.64, 0.93 Mom Disapprove of Sex 0.80 0.69, 0.94 Mom Disapprove of Birth Control 0.75 0.67, 0.84 Length of Residence in Years 0.99 0.98, 1.02 Health Insurance 1.40 0.89, 2.18 Driver’s License 1.51 1.12, 2.03 Two Parent Household 0.68 0.51, 0.91 School Education: Medical Attention 1.13 0.80, 1.59 School Health Care Non-Athlet ic Physical 0.97 0.68, 1.39 Cross Level Interactions Intercept X % Black Residents 1.23 0.24, 6.45 Individual Race X % Black Residents Black X % Black Residents 1.38 0.31, 6.09 White X % Black Residents (reference) Parental Education X % Black Residents Less than High School X % Black Residents 0.52 0.11, 2.48 High School Degree/ GED X % Black Residents 0.88 0.28, 2.75 Some College/Technical X % Black Residents 0.40 0.11, 1.43 College Degree or More X % Black Residents (reference) Ratio of Income to Poverty X % Black Residents 0-100% X % Black Residents 1.24 0.37, 4.15 101-200% X % Black Residents 0.57 0.14, 2.32 201-300% X % Black Residents 0.75 0.19, 2.94 301-400% X % Black Residents 0.72 0.13, 4.07 401% or More X % Black Residents (reference) Random Effects Intercept 0.12 -2LL 4655.27 2 6.58(9)

PAGE 170

157 In Model II from Hypothesis III, th e random effects for the cross-level interactions between the neighborhood socioeconomic c ontext (proportion of low educated residents) and individual race a nd socioeconomic position were non-significant. Thus, the slopes were fixed and the model was reanalyzed with the results presented in Model II of Table 24. The results of the analysis revealed that the fixed effects for the cross-level interactions betw een the neighborhood socioecono mic context (proportion of low educated residents) a nd individual race and soci oeconomic position were nonsignificant, as the overall model was not significant ( 2 =15.23(9), p > 0.05) when comparing Model II to Model IV from Hypot hesis I. Thus, in the typical neighborhood, social disparities in the odds of having received contraceptiv e services were not dependent on the proportion of low educated residents within the neighborhood among this sample of sexually expe rienced adolescent females.

PAGE 171

158 Table 24 HGLM Analyses: Contribution of the Proporti on of Low Educated Residents in the Neighborhood to Social Disparities in the Rece ipt of Contraceptiv e Services (N=1,526) Model II OR 95% CI Fixed Effects Level I Variables Intercept 0.64 0.34, 1.24 Race Black 0.99 0.73, 1.36 White (reference) Parental Education Less than High School 1.67 1.02, 2.72 High School Degree or GED 1.02 0.70, 1.47 Some College/Technical 0.99 0.70, 1.43 College Degree or More (reference) Ratio of Income to Poverty 0-100% 0.77 0.49, 1.22 101-200% 0.69 0.45, 1.06 201-300% 0.60 0.40, 0.90 301-400% 0.58 0.37, 0.92 401% or More (reference) General Health 1.03 0.91, 1.18 STI History 3.91 2.23, 6.86 Pregnancy History 2.31 1.61, 3.33 Age 1.11 0.98, 1.27 Pregnancy Undesirable 0.84 0.71, 0.99 Barriers to Birth Control 0.76 0.63, 0.92 Mom Disapprove of Sex 0.80 0.69, 0.94 Mom Disapprove of Birth Control 0.74 0.66, 0.82 Length of Residence in Years (0.99 0.98, 1.02 Health Insurance 1.39 0.89, 2.16 Driver’s License 1.54 1.14, 2.08 Two Parent Household 0.70 0.53, 0.93 School Education: Medical Attention 1.14 0.81, 1.61 School Health Care Non-Athlet ic Physical 0.99 0.70, 1.41 Cross Level Interactions Intercept X % Low Educated Residents 47.23 3.40, 656.11 Individual Race X % Low Educated Residents Black X % Low Educat ed Residents 0.97 0.08, 11.22 White X % Low Educated Residents (reference) Parental Education X % Low Educated Residents Less than High School X % Low Educated Residents 0.08 0.00, 5.28 High School Degree/ GED X % Lo w Educated Residents 0.26 0.01, 4.44 Some College/Technical X % Low Educated Residents 0.06 0.00, 0.99 College Degree or More X % Low Educated Residents (reference) Ratio of Income to Poverty X % Low Educated Residents 0-100% X % Low Educated Residents 0.35 0.01, 11.26 101-200% X % Low Educated Residents 0.07 0.00, 3.23 201-300% X % Low Educated Residents 0.46 0.02, 12.87 301-400% X % Low Educated Residents 0.02 0.00, 1.54 401% or More X % Low Educated Residents (reference) Random Effects Intercept 0.11 -2LL 4646.61 2 15.23(9)

PAGE 172

159 In Model III from Hypothesis III, th e random effects for the cross-level interactions between the neighborhood socioeconomic c ontext (proportion of poor residents) and individual race and socioec onomic position were non-significant. Thus, the slopes were fixed and the model was reanalyzed with the results presented in Model III of Table 25. The results of the analysis revealed that the fixed effects for the cross-level interactions between the neighborhood socioeconomic c ontext (proportion of poor residents) and individual race and socioec onomic position were non-significant, as the overall model was not significant ( 2 =14.90(9), p > 0.05) when comparing Model III to Model IV from Hypothesis I. Thus, in the ty pical neighborhood, social disparities in the odds of having received contraceptive servi ces were not dependent on the proportion of poor residents within the neighborhood among this sample of sexually experienced adolescent females.

PAGE 173

160 Table 25 HGLM Analyses: Contribution of the Proporti on of Poor Residents in the Neighborhood to Social Disparities in the Receipt of Contraceptive Services (N=1,526) Model III OR 95% CI Fixed Effects Level I Variables Intercept 0.68 0.34, 1.34 Race Black 0.99 0.73, 1.36 White (reference) Parental Education Less than High School 1.71 1.06, 2.76 High School Degree or GED 1.02 0.70, 1.49 Some College/Technical 0.98 0.67, 1.42 College Degree or More (reference) Ratio of Income to Poverty 0-100% 0.78 0.49, 1.25 101-200% 0.66 0.42, 1.03 201-300% 0.59 0.38, 0.91 301-400% 0.58 0.35, 0.98 401% or More (reference) General Health 1.03 0.91, 1.18 STI History 3.85 2.17, 6.81 Pregnancy History 2.33 1.64, 3.33 Age 1.12 0.98, 1.27 Pregnancy Undesirable 0.83 0.71, 0.98 Barriers to Birth Control 0.76 0.63, 0.82 Mom Disapprove of Sex 0.81 0.69, 0.94 Mom Disapprove of Birth Control 0.75 0.67, 0.83 Length of Residence in Years 0.99 0.98, 1.02 Health Insurance 1.39 0.89, 2.17 Driver’s License 1.51 1.12, 2.04 Two Parent Household 0.68 0.51, 0.91 School Education: Medical Attention 1.12 0.80, 1.58 School Health Care Non-Athlet ic Physical 1.01 0.71, 1.43 Cross Level Interactions Intercept X % Poor Residents 124.37 2.07, 7459.23 Individual Race X % Poor Residents Black X % Poor Residents 3.24 0.22, 49.04 White X % Poor Residents (reference) Parental Education X % Poor Residents Less than High School X % Poor Residents 0.01 0.00, 49.04 High School Degree/ GED X % Poor Residents 0.07 0.00, 1.97 Some College/Technical X % Poor Residents 0.01 0.00, 0.61 College Degree or More X % Poor Residents (reference) Ratio of Income to Poverty X % Poor Residents 0-100% X % Poor Residents 0.12 0.00, 9.17 101-200% X % Poor Residents 0.05 0.00, 4.41 201-300% X % Poor Residents 0.12 0.00, 0.98 301-400% X % Poor Residents 0.05 0.00, 13.08 401% or More X % Poor Residents (reference) Random Effects Intercept 0.10 -2LL 4646.95 2 14.90(9)

PAGE 174

161 Summary of Overall Findings The primary aims of this study were: (1) to examine the extent to which social disparities in access to health care existed a nd the extent to which access to health care and social disparities in a ccess to health care varied acr oss neighborhoods, (2) to examine the relationships between th e neighborhood racial and soci oeconomic context and the average odds of access to health care, and (3) to examine the extent to which the neighborhood racial and socioeconomic context c ontributed to social disparities in access to health care among this sample of sexua lly experienced adolescent females and the extent to which this relations hip varied across neighborhoods. Social Disparities in Access to Health Care The findings from this study revealed few social disparities in access to health care among this sample of sexually experien ced adolescent females. Contrary to the hypotheses, after adjusting for adolescents ’ socioeconomic position and the control variables, there were no racial disparities in the receipt of a routine physical, in the presence of unmet health needs, or in the receipt of contraceptive services among this sample of sexually experienced adolescent females. Socioeconomic disparities were noted among this sample of sexually experien ce adolescent females in their receipt of a routine physical and contraceptive services. Spec ifically, after adjusting for adolescents’ race, household income, and the control variables, the findings revealed that adolescents who had a parent with some co llege or technical training were less likely to have received a routine physical than those adolescents who had a parent w ith a college degree or more. In relation to contraceptive services, afte r adjusting for adoles cents’ race, parental educational attainment, and the control vari ables, adolescents whose household income

PAGE 175

162 was 201% to 300% above their poverty thresh old had significantly lower odds of having received contraceptive servi ces than those adolescents w hose household income was over 400% of their poverty threshold. The findings revealed that adoles cents whose household income was 301% to 400% above their poverty threshold had significa ntly lower odds of having received contraceptive services than those adolescents whose household income was over 400% of their poverty threshold, but these socioeconomic disparities were fully mediated by the control variables. Contrary to the hypotheses, socioeconomic disparities were also noted in the receipt of contracep tive services for those adolescents who had a college educated parent. Sp ecifically, after adjusting for adolescent race, household income, and the control variables, adolescents who had a parent with less than a high degree were more likely than those adolescents who had a parent with a college degree or more to have received contraceptive servic es. Lastly, socioecono mic disparities were noted in unmet health needs among adoles cents whose household income was 100% or below their poverty threshold, 201% to 300% ab ove their poverty th reshold, and 301% to 400% above their poverty threshold, but these disparities were fully mediated by the control variables. These findings indicated th at the socioeconomic disparities in unmet health needs were a function of adolescents ’ predisposing factors, enabling factors, and/or health needs. Thus, Hypothesis IA was not supported and Hypothesis IB was partially supported. Interactions between race and socioec onomic position were not analyzed due to small cell sizes when race and socioec onomic position were cross-multiplied for examination with the dependent variables. Thus, Hypothesis IC c ould not be reliably tested.

PAGE 176

163 In testing Hypothesis I, th e results of the null model indicated that there was significant variation in the av erage odds of having receiv ed a routine physical, unmet health needs, and of having received cont raceptive services across neighborhoods among this sample of sexually experienced a dolescent females, which fully supported Hypothesis ID. The findings suggested that ad olescents’ race and so cioeconomic position did not account for this varia tion in access to health care fo r any of the three dependent variables. However, predisposing factors, en abling factors, and hea lth needs did account for some of the variation across neighborhoods in access to a routine physical, as the random intercept decreased in size after the addition of the control variables into the model. The random intercept was still sign ificant after the addition of the control variables, which suggested that there was stil l variation across neighborhoods in access to a routine physical that could not be explained by the Level I variables examined in this study. The size and significance of the random intercept remained unchanged throughout the modeling for unmet health needs and actually increased for the receipt of contraceptive services, which s uggested that the Leve l I variables examined in this study did not account for any of the variation acr oss neighborhoods in unmet health needs or the receipt of contraceptive services. The extent to which the relationships betw een race and access to health care and socioeconomic position and access to health care varied across neighborhoods was also examined. However, the findings revealed that no significan t variation across neighborhoods between race, socioeconomic pos ition and access to health care. Thus, Hypothesis IE was not supported.

PAGE 177

164 The Neighborhood Context and Access to Health Care One of the aims of this study was to examine the relationship between the neighborhood racial and socioeconomic contex t and the average odds of access to health care among the sexually experienced adolescent females in this sample. However, no significant relationships between access to health care and th e proportion of Black residents within the neighborhood, the propor tion of poor resi dents within the neighborhood, or the proportion of lower e ducated residents within the neighborhood were noted among this sample of sexually ex perienced adolescent females. Thus, IIA and IIB were not supported. In addi tion, no significant interacti ons were noted between the neighborhood racial and socioeconomic context and access to health care. Thus for this sample, the relationship between the neighbor hood racial context a nd access to health care was not a significant function of th e neighborhood socioeconomic context and Hypothesis IIC was not supported. As noted in the preceding discussion the random effects of the null model revealed that there was significant vari ation in access to health care across neighborhoods. However, the neighborhood raci al and socioeconomic context did not account for any of the variation in the receipt of a r outine physical, in unmet health needs, or in the receipt of contraceptive servi ces across neighborhoods. Thus, Hypothesis IID was not supported. For the dependent variablesthe receipt of a routine physical and the receipt of contraceptive serv icesthe neighborhood control variables did impact the size of the random intercept when entered into th e model. These findings suggested that the neighborhood control variables accounted for some of the variation in access to a routine physical and to contraceptive services across neighborhoods. The random intercept

PAGE 178

165 remained significant in the model, which indicat ed that there was still variation in access to health care that was not explained by th e Level II variables examined in this study. The Neighborhood Context and Social Dispari ties in Access to Health Care The purpose of Hypothesis III was to examine the extent to which social disparities in access to health care were a function of the neighborhood racial and socioeconomic context. The findings indicated th at for this sample of adolescents, social disparities in the receipt of a routine physical, unmet health needs, and the receipt of contraceptive services were not de pendent upon the neighborhood racial or socioeconomic context. Furthermore, the relationships between the neighborhood racial and socioeconomic context, adolescents’ race and socioeconomic position, and access to health care did not vary across neighborhoods nor did they account for any of the variation in average odds of access to health care across neighborhoods. Thus, Hypotheses IIIA-Hypotheses IIIC were not supported. Post-Hoc Analyses: Associations between Type of Health Care Place and Service Receipt Post-hoc univariate and bi variate analyses were conducted to explore, (1) associations between adolescents’ soci odemographic characteris tics and the type of health care place adolescents’ received health care services, and (2) associations between type of health care place where adolescents received their routine physical and their receipt of contraceptive services. The reason fo r these additional analyses is that the type of health care place where adolescents receiv e health care services may have important implications for the prevention of STI’s, HI V, and unintended pre gnancies, as evidence suggests that adolescents who receive care at a clinic may be more likely to receive reproductive health counseling and services than adolescents who receive care at a

PAGE 179

166 private office (Ashton et al., 2002; C ook, Wiesenfeld, Ashton, Krohn, Zamborsky, & Scholle; 2001; Porter & Ku, 2000). Furthermore, research has also indicated that minority and poor adolescents are more likely to rece ive health care from a community health clinic than White and higher-income adoles cents (Frost, 2001), thus community clinics play an important role in ameliorating social disparities in access to health care, and potentially social dispari ties in adolescents’ health. Univariate Analyses Type of Health Care Place for Adolescents’ Receipt of Routine Physical The Add Health data provides information regarding the type of health care place for the receipt of a routine physical, but not for the receipt of contraceptive services. The type of health care place included a private doctor’s office, community health clinic, school, hospital, or some other place. Table 26 summarizes the descriptive statistics related to the type of health care place that the adolescents in the study sample received a routine physical. Accord ing to the findings, a total of 1,044 adolescent females in the study reported that they had r eceived a routine physical. Near ly half of the adolescents received their routine physical at a privat e doctor’s office (53.39%), approximately 22% received their routine physical at a community health clin ic (21.83%), and less than 10% received their routine physical at school (5.75%), at a hosp ital (6.85%), or at some other place (2.76%). Nearly 10% of th e adolescents reported that th ey had received a routine physical at more than on health care place.

PAGE 180

167 Table 26 Descriptive Statistics of th e Type of Health Care Place Where Adolescents Received Their Routine Physical (n = 1,044) Type of Health Care Place Received Routine Physical n Proportion % SE Private Doctor’s Office 555 53.39 2.72 Community Health Clinic 225 21.83 2.07 School 58 5.75 1.35 Hospital 97 6.85 1.41 Other 24 2.76 0.73 Received Physical at More th an One of the Above 85 9.42 1.53 Bivariate Analyses Associations between Adoles cents’ Characteristics and th e Type of Health Care Place for their Receipt of Routine Physical Relationships between adolescent sociode mographic characteristics and the type of health care place where adolescents received their routine physical were examined via logistic regression analyses. The purpose of th ese analyses was to gain an understanding of the extent to which racial and socioeconomic differences existed in the type of health care place where adolescents received their r outine physical. The analyses only included those adolescents who recei ved a routine physical at a private doctor’s office ( n = 555), a community health clinic ( n = 225), a school ( n = 58), or a hospital ( n = 97) due to the ambiguity of receiving a physic al at “some other place” ( n = 24) and also of receiving a routine physical at more than one health care place ( n = 85). Thus, the sample size for the following bivariate analyses is 935 adolescents In addition, Bonferroni corrections were utilized with all post-hoc analyses to adjust for multiple comparisons, thus the p-value for significance in these analyses was < 0 .004. Table 27 summarizes the findings.

PAGE 181

168 With the Bonferroni corrections, racial di fferences were noted only in the receipt of a routine physical at a private doctor’s o ffice, as adolescents who self-identified as Non-Hispanic Black were less likely than adolescents who self-identified as NonHispanic White to have received a routine physical at a private physician’s office (OR = 0.48, 95% CI = 0.33, 0.70). Socioeconomic differ ences were also note d, as adolescents whose ratio of income to poverty was 100% or below the poverty threshold poverty threshold (OR = 0.23, 95% CI = 0.14, 0.36) or 101%-200% above the poverty threshold (OR = 0.21, 95% CI = 0.13, 0.34) were less likely to have received a routine physical at a private doctor’s office than those adolescents whose ratio of income to poverty was over 400% of the poverty threshold. In addition, adol escents whose ratio of income to poverty was 100% or below the poverty threshol d (OR = 4.48, 95% CI = 2.47, 8.12) and 101%200% above the poverty threshold (OR = 3.63, 95% CI = 1.82, 7.24) were more likely to have received a routine physical at a comm unity health clinic than those adolescents whose ratio of income to poverty was over 400% of the poverty threshold. In relation to level of parental educational attainment, a dolescents who had a parent with less than a high school degree (OR = 0.27, 95% CI = 0.14, 0.51) were less likely to have received a routine physical at a private doctor’s office and more likely to have a physical at a community health clinic (OR = 2.78, 95% CI = 1.50, 5.15) than those adolescents who had a parent with a college degree or more.

PAGE 182

169 Table 27 Associations between Adol escents’ Sociodemographic Characterist ics and the Type of Health Care Place for their Receipt of a Routine Physical (n = 935) Adolescents’ Sociodemographic Characteristics Private Doctor Office Community Health Clinic School Hospital OR 95% CI a OR 95% CI a OR 95% CI a OR 95% CI a Race Black 0.48 0.33, 0.70 1.95 1.19, 3.21 0.70 0.26, 1.89 2.18 1.04, 4.57 White (reference) Ratio of Income to Poverty 0-100% 0.23 0.14, 0.36 4.48 2.47, 8.12 2.00 0.67, 5.97 1.66 0.64, 4.33 101-200% 0.21 0.13, 0.34 3.63 1.82, 7.24 2.59 0.90, 7.47 2.80 1.23, 6.38 201-300% 0.55 0.33, 0.93 1.78 0.87, 3.65 1.59 0.51, 4.95 1.34 0.52, 3.51 301-400% 0.60 0.32, 1.14 1.00 0.50, 1.99 3.69 1.47, 9.28 1.36 0.44, 4.20 400% or More (reference) Parental Education Less than High School 0.27 0.14, 0.51 2.78 1.50, 5.15 1.22 0.46, 3.26 3.46 1.32, 9.07 High School Degree GED 0.56 0.33, 0.96 2.25 1.23, 4.10 0.82 0.37, 1.79 0.98 0.40, 2.36 Some College/Technical 0.60 0.37, 0.95 1.50 0.87, 2.58 1.63 0.81, 3.28 1.35 0.60, 3.04 College Degree or More (reference) a Significance adjusted for post-hoc analyses via Boneferroni due to multiple comparisons, thus p-value significant at < 0.004

PAGE 183

170 Associations between Adolescents’ Receipt of Contraceptive Services and the Type of Health Care Place for thei r Receipt of a Routine Physical Associations between the r eceipt of contraceptive servic es and the type of health care place where adolescents received their rou tine physical were examined via logistic regression analyses. The purpose of these analys es was to gain a bett er understanding of the extent to which the receipt of contracep tive services may vary based on the type of health care place where sexually experienced adolescent females receive their routine physical. Receipt of a routine physical at a private doctor’ s office was considered the reference group for these analyses Bonferroni corrections were utilized with all post-hoc analyses to adjust for multiple comparisons, thus the p-value for significance in these analyses was < 0.004. Table 28 summarizes the findings of these analyses. Of the 935 adolescent females in the preced ing bivariate analyses who received a routine physical, 504 (54.31%) al so received contraceptive services. With the Bonferroni corrections, no significant associations were revealed between the t ype of health care place where the adolescents received thei r routine physical and their receipt of contraceptive services. Table 28 Associations between Adolescents’ Receipt of Contraceptive Services and the Type of Health Care Place for their Receip t of a Routine Physical (n = 935) Type of Health Care Place Received Contraceptive Services OR 95% CI Private Doctor Office (reference) Community Health Clinic 1.51 1.01, 2.24 School 0.54 0.20, 1.42 Hospital 0.97 0.53, 1.75 a Significance adjusted for post-hoc analys es via Boneferroni due to multiple comparisons, thus p-value significant at < 0.004

PAGE 184

171 Chapter Five: Summary of Findings Social Disparities in Access to Health Care The results from this study revealed few so cial disparities in access to health care among this sample of sexually experienced adolescent females. However, this study was unable to determine if adolescents received care because they had a health need, such as a pregnancy or STI, or if the visit was preventa tive in nature. This distinction is important to note, as the findings from this study reve aled that the adolescents in this sample experienced racial and socioeconomic disp arities in reproductive and overall health, which may have contributed to their receipt of services. The findings from this study revealed no r acial disparities in the receipt of a routine physical, unmet health needs, or in the receipt of contraceptive services between those adolescents in this sample who self-ide ntified racially as Black and those who selfidentified racially as White, before and af ter adjusting for adolescents’ socioeconomic position and the Level I control variables. Althou gh the lack of racial disparities in access to health care among these adolescents is appl auded, particularly since racial disparities in their health were noted, the findings from this study are somewhat in contrast with previous evidence. Specifically, studies have revealed that Blac k adolescents are less likely than White adolescents to receive any h ealth care (Bartman et al., 1997; Lieu et al., 1993) as well as a preventive health care vi sit (Shenkman et al., 2003) and they are more likely to report unmet health needs (Ford et al., 1999) and to utilize the emergency

PAGE 185

172 department as their usual source of care (Wilson & Klein, 2000). However, one recent comprehensive review examining racial dispar ities in access to adol escent health care, noted that Black adolescents were less lik ely to have received primary heath care services, such as a routine physical, or to ha ve a routine health care provider than White adolescents in only 3 of the 6 studies that met the criteria for their review (Elster et al., 2003). Two of the three studies that did not reveal any di fferences between Black and White adolescents in having received a routin e physical utilized the full samples of the restricted and public use data set from Add Health, which is the same national data set employed in this study (Ford et al., 1999; Yu et al., 2001). While the study by Ford and colleagues (1999) did not reveal any racial disparities in the receipt of a routine physical, they did note racial disparities in forgone health care (unmet health needs), as Non-Hi spanic Black female adolescents were more likely to forgo health care than Non-Hispanic White female adolescents, after adjusting for household income, insurance status, health needs, and risk behavi ors. In the findings from this study, bivariate analyses revealed that sexually experienced adolescent females who self-identified racially as Non-Hispanic Black were more likely to report unmet health needs than sexually experienced adolesce nt females who self-identified racially as Non-Hispanic White, although the effect size wa s small. In multivariate analyses, these findings were not significant, even before adjusting for socioeconomic position and the Level I control variables. The differences in the findings between the study by Ford and colleagues (1999) and this study may be due di fferences in the size of the study samples. For example, Ford and colleagues (1999) anal yzed the core sample of Wave I from the Add Health restricted data set (N=12,102), while this study examined only those sexually

PAGE 186

173 experienced Non-Hispanic Black and Non-Hi spanic White adolescent females who met the inclusion criteria for this study (N=1,526) Consequently, the smaller sample size in this study may have limited the ability to dete ct significant relationships with small effect sizes (Tabachnick & Fidell, 2001). Differences in the characteristics of the study samples could have also accounted for the disparat e findings, as this study only included only those adolescent females who were sexually experienced while the study by Ford and colleagues (1999) included those adolescent females who were sexually experienced and sexually non-experienced. Interestingly, the Fo rd study (1999) revealed that sexually active adolescent females were more likely to forgo health care th an those adolescent females who were not sexually active, but the moderating effects of race and ethnicity on the relationship between sexually activity and forgone health care were not examined. Few studies have examined racial dispari ties in access to repr oductive health care services. However, the findings that do exist suggest that minority adolescents may have similar or higher rates of receiving STI a nd contraceptive services than non-minority adolescents. For example, Porter and Ku ( 2000) examined racial disparities among male adolescents in the receipt of reproductive health care services using the National Survey of Adolescent Males and noted that African Am erican males were more likely than White males to have discussed reproduc tive health issues with thei r health care provider and to have received HIV and STI testing. Among a dolescent females, a study utilizing data from Wave I of the Add Health data indicated that Black adolescent females were more likely to have received screening or treatment for STI’s than any other race or ethnicity (Fiscus, Ford, & Miller, 2004) and in this study no significant diffe rences were noted between Non-Hispanic Black and Non-Hispanic White adolescent females in the receipt

PAGE 187

174 of contraceptive services. Further studie s are necessary though, as once recent study (Scher, 2004) that utilized a larger sample of sexually experienced adolescent females (N=2,929) from the Add Health data revealed that Black adolescents were less likely to have received contraceptive services th an White adolescents, after adjusting for covariates in the model. However, the study by Scher (2004) did not stat e if the racial and ethnic categories were mutually exclusive, t hus racial and ethnic differences must be interpreted with caution. The most likely re asons for the difference in the sample size between the Scher study (2004) and this study, wa s that in contrast to the Scher study, this study included household income and insura nce status as covariates, both of which had a significant proportion of missing data and Scher ( 2004) also included Latina adolescents. In relation to socioeconomic disparities in access to health care, this study revealed disparities in the receipt of a routin e physical and in the receipt of contraceptive services. Although socioeconomic disparities in unmet hea lth needs were also noted, these disparities were fully mediated by the c ontrol variables. The findings of this study revealed that adolescents who had a parent wi th some college or te chnical training were less likely to have received a routine physical than those adolescents who had a parent with a college education or more, after ad justing for adolescent race, the adolescent household income to poverty ratio, and the Leve l I control variables. The analyses related to the receipt of contraceptive services re vealed that adolescen ts whose household income was 201% to 300% above their povert y threshold were less likely to have received contraceptive servi ces than those adolescents w hose household income was over 400% of their poverty threshold, after adjusting for adolescent race and level of parental

PAGE 188

175 education. However, contrary to the hypothese s, adolescents who had a parent with less than a high school degree were more likely th an those adolescents who had a parent with a college degree or more to have received contraceptive services. The findings from this study related to socioeconomic dispar ities in access to health care are generally in accordance with the findings in the lite rature. For example, previous research revealed that adolescents from lower income families are less likely to receive any health care or to have a preventi ve health care visit (Newacheck et al., 2003; Simpson et al., 2005) and they are more likely to forgo health care (unmet health needs) due to costs (Newacheck et al., 2003) and to utilize the emergency department as their usual source of care (Wilson & Klein, 2000). Ho wever, the socioeconomic disparities in access to health care noted in this study we re not apparent among those adolescents who came from the poorest households or those who had a parent with least amount of education, and in relation to contraceptive services, those adolescen ts who had a parent with the least amount of educat ion were more likely to rece ive services. Se veral studies have noted similar findings (Yu et al., 2001; Porter & Ku, 2000). Yu and colleagues (2001) hypothesized that eligibility for social welfare programs, such as Medicaid, may help to reduce the financia l barriers in accessing health care among those lowest in socioeconomic position (Yu et al., 2001). In addition, the availability of school health care services may also facilitate access to health care for socioeconomically disadvantaged adolescents, as evidence sugge sts that schools with a greater number of students on Medicaid may be more likely to provide mental health services for their students (Slade, 2003). Furthermore, Scher (200 4) noted that schools were more likely to offer contraceptive services in their schools if school admi nistrators reported teenage

PAGE 189

176 pregnancy to be a problem in their sc hools. Thus, providers and schools may be responding to increased health and socioec onomic needs among socially disadvantaged adolescents. The mixed findings related to racial and socioeconomic differences in the receipt of STI and contraceptive services may also be due to differences in screening practices by health care providers. Numerous studies have reported that certain provider characteristics are associated with increased screening for STI’s, including female gender (Cook et al., 2001; Millstei n, Ingra & Gans, 1996; Torkko, Gerhman, Vrane, Hamman, & Baron, 2000), practicing in a clinic or HMO practice versus a priv ate practice (Cook et al., 2001; Millstein et al., 1996; Porter & Ku, 2000), African American race (Ray et al., 2005), patient population equal to or great than 20% African American (Cook et al., 2001), practicing in a lower-income nei ghborhood (Ray et al., 2005), provider comfortable with discussing re productive health is sues and knowledgeable about STI’s and contraceptives (Cook et al., 2001), and mo re recent graduate of medical education (Millstein et al., 1996). Although the findings fr om this study revealed some racial and socioeconomic differences in the type of h ealth care place where adolescents receive their routine physical, the rece ipt of contraceptive serv ices did not vary based on the type of setting in which they receive d their routine physi cal (with Bonferr oni corrections). However, unmeasured differences in provide r characteristics coul d have potentially accounted for some of the racial and socio economic differences in services rendered. The mixed findings may also be due widespread educational campaigns increasing awareness about ra cial and ethnic disparities in reproductive health and reproductive health screening (Mosher, Martinez, Chandra, Abma, & Wilson, 2004;

PAGE 190

177 USDHHS, 2000) as well as efforts to incr ease STI testing among adolescents to reduce the negative health outcomes of untreated infections (USDHHS, 2000). While increased reproductive health counseling and STI te sting among adolescents from socially disadvantaged racial and socioeconomic groups may be due to providers’ greater awareness that these groups are at hi gher risk for STI’s, HIV, and unintended pregnancies, this also may lead to soci al profiling. Clinical practice guidelines recommend that all adolescents receive yearly repr oductive health counseling and if sexually active, adolescents should be tested at least yearly fo r STI’s, regardless of adolescents’ race, level of educati on, or income (AAP, 2000; AMA, 1997). The Neighborhood Context and Access to Health Care The findings from this study indicated th at access to health care among sexually experienced adolescent females varied acr oss neighborhoods, even after the addition of individual and neighborhood le vel covariates. However, the relationships between adolescent race and access to health care and adolescent socioeconomic position and access to health care did not vary across ne ighborhoods. The findings also revealed that the neighborhood racial and soci oeconomic context was not asso ciated with adolescents’ access to health care nor did it account for a ny of the variation in access to health care across neighborhoods. The control variables, particularly geographic region, accounted for some of the variation in the receipt of a routine physical, but not for any of the variation in unmet health needs or contra ceptive services. Lastly, the relationships between adolescent race and access to health care and adolescent socioeconomic position and access to health care were not m oderated by the neighborhood racial and socioeconomic context in this study.

PAGE 191

178 The paucity of findings in this study related to the rela tionship between the neighborhood racial and socioeconomic context and access to health ca re are contrary to those found in the literature, although few studies have examined these relationships. For example, in one contextual analysis, Kirby and Kaneda (2005) not ed that individuals living in neighborhoods characterized by lower incomes were less likely to have received blood pressure screening and to have a usual source of care and they were more likely to report unmet health needs th an those individuals living in neighborhoods characterized by higher incomes. Brooks-Gunn and colleague s (1998) reported similar findings in a contextual analysis examining pediatric health care, as formerly premature children who lived in a socioeconomically disadvantaged neighborhood at the time of birth were more likely to utilize the emergency department dur ing the first three years of their life than those children who lived in a socioeconomica lly advantaged neighborhood at the time of birth (Brooks-Gunn et al., 1998). Studies have reported significant relationships between the racial and/or socioeconomic context and in dividual access to health care at the level of the MSA (Andersen et al., 2002) and count y (Haas et al., 2003). Furthermore, Haas and colleagues examined cross-level interac tions between individua l race and ethnicity and the neighborhood racial and ethnic context and reporte d that Blacks experienced fewer problems in obtaining health care and fewer financial barriers in obtaining care when they lived in a county that had a highe r percentage of Black residents than those who lived in counties with a lower percentage of Black residents. However, since none of the preceding studies employed multilevel analyses the variation in access to health care across neighborhoods could not be examined (Diez Roux, 2003).

PAGE 192

179 Although the neighborhood vari ables of interest -the neighborhood racial and socioeconomic context-were not significantly as sociated with access to health care in this study, the findings indicated that access to health care varied across neighborhoods among this vulnerable population of adolesce nt females. Regional differences in the receipt of a routine physical were noted and also accounted for some of the variation in access to health care across neighborhoods. Res earchers examining geographic variations in access to health care hypothesize that di fferences in state policies and market influences may impact the availability of se rvices and contribute to differences in access to health care (Sturm, Ringel, & Andreyev a, 2003). Further studies are needed to examine the relationships between these macrolevel forces, the availa bility of services, and individual access to health care and their potential contribu tion to social disparities in access to health care. Limitations of the Study The most significant limitation to this study was the combination of the small sample size and the limited variability in th e sociodemographic composition of the neighborhoods. Since this study did not em ploy primary data collection, sampling procedures to control the size and compos ition of the sample were not possible. Furthermore, the sampling for the Add Health study was done at the level of the school, rather than the neighborhood (Harris et al., 2003 ). Consequently, the researchers had no control over the racial and socioeconomic diversity of the neighborhoods that were represented in this study, pa rticularly since this study re quired subsetting of the full sample to only include those sexually expe rienced adolescent females in the study.

PAGE 193

180 The power analyses revealed that there was adequate power to conduct multilevel analyses with this sample, thus, the fixed e ffects and variability of the random intercept should be relatively stable for this st udy (Snijders & Bosker, 1999). Unfortunately though, the stability of some of the Level II fi xed effects, the Level II fixed interactions, and the cross-level interactions were impacted by the lack of variability in the sociodemographic composition of the neighborhood, as the majority of the adolescents in this study lived in neighborhoods in which th ere were a small proportion of Black or poor residents. This can be noted in the large confidence intervals for some of the Level II effects as well as the cross-level interactions Furthermore, the ability to reliably examine the random slopesthe variation in the relatio nships between adolesce nt race, adolescent socioeconomic position, and access to health care across neighborhoodswas most likely compromised by the limited number of adoles cents per neighborhood, and potentially the limited dispersion of Black and White adoles cents across the neighborhoods (Snijders & Bosker, 1999). In other words, if there were high levels of segr egation in which few White adolescents lived in Black neighborhoods or vice versa, then there would be a limited number of cases to examine the varia tion in the relationships between race and access to health care across neighborhoods. Vari ation in the data would be more likely with a larger sample size. The small sample sizes also restricted the ability to reliably examine the Level I interactions between race and socioeconomic position, as the cell sizes became quite small (see Table 4). Second, the study design was non-experimental thus causality in the relationships between the independent and dependent vari ables could not be established. The study included only those adolescents who lived in their neighborhood for at least a year to

PAGE 194

181 ensure exposure to the neighborhood occurred pr ior to adolescents’ access to health care. However, the Level I control variables-predis posing factors, enabling factors, and health needswere measured at the same time as the dependent variables. As a result, the preventive nature of the visit is confounded, as adolescents may have utilized access to care because of a health need. Add Health is a longitudinal study, thus Wave I and Wave II could have been examined with this study to control for this issu e. However, attrition between samples, particularly among older adolescents due to graduation impacts the sample size between waves. Since older a dolescents are more likely to be sexually experienced, and the sample size was already smaller by virtue of only examining those sexually experienced adolescent females, the Wave I sample was selected for this study due to the larger sample size. Longitudinal st udies will need to be conducted in the future to reduce the potential fo r confounding relationships. Causality could also not be established as adolescents were not randomized into the neighborhoods in which they lived. Randomization of a dolescents into specific neighborhoods is not ethnically justifiable or economically feasible to conduct a study of this scale. Consequently, the lack of random ization can result in se lection or endogeneity bias, as families often choose to live in a certain neighborhood for a specific reason (e.g. to be closer to hospitals, doc tors office, schools, work, family etc.). When these reasons are not accounted for, the neighborhood effects can be overestimated or underestimated (Leventhal & Brooks-Gunn, 2000). The Add Hea lth study does include data on reasons why families chose to live in a particular neighborhood, but proximity to health care resources was not an option for parents to choos e. Therefore, if the parents in this study moved to a particular neighborhood to be closer to their health care provider, the true

PAGE 195

182 relationship between the neighborhood context and access to health care could have been obscured. Finally, this study included c ontrol variables at both leve ls of analyses. Although this is considered standard practice, some neighbor hood researchers believe that this can lead to “statistical overadjustment”, as the variab les that are considered confounders are often on the causal pathway between the focal independ ent variable and the outcome of interest (Macintyre & Ellaway, 2003). This overadjustmen t also pertains to the common practice of partitioning the random effect s, which is conducted in an attempt to determine if the variation in the outcome is due to contex tual or compositional factors (Macintyre & Ellaway, 2003). The context and the individuals that compose the context are interrelated, and overlooking these reciproc al relationships can have serious policy implications (Macintyre & Ellaway, 2003). In this study, the control variables were entered into the model after the independent variables of inte rest, thus the potential mediating effects of the control variables were noted. For exampl e, the relationships between the household income to poverty ratio and unmet health need s as well as the rece ipt of contraceptive services were fully mediated by the contro l variables. Further analyses are needed though, to gain a better unders tanding of which control variables mediated the socioeconomic disparities in access to health care, as this can have important policy implications for reduc ing disparities. Implications for the Field Social disparities in access to health care are widespread throughout the U.S., occurring across the lifespan a nd among a variety of health care services. However, few studies have examined the extent to which social disparities in access to health care

PAGE 196

183 among adolescents, particularly those who are sexually active. This study revealed no racial disparities in access to health care and few socioeconomic disparities in access to health care in this sample of adolescent fe males, which suggested that these adolescents were obtaining health care services. Howeve r, this study was unable to determine if adolescents received care because they had a he alth need, such as a pregnancy or STI, or if the visit was preventative in nature. This di stinction is important to note, as the findings from this study revealed that the adolescen ts in this sample experienced racial and socioeconomic disparities in reproductive and overall health, which may have contributed to their receipt of services. The findings from this study also revealed that adolescents’ acce ss to health care varied across neighborhoods. A lthough this study suggested the neighborhood racial and socioeconomic context did not contribute to th is variation in access to health care, the findings indicate that future research is necessary to explore the neighborhood characteristics that may contribute to this disparity across neighborhoods. Furthermore, once potential pathways between neighbor hoods and access to health care are more clearly elucidated, policies can be developed to target those factors th at contribute to the disparity in access to health care across neighborhoods. Methodologically, this study uncovered seve ral important considerations in the design of future multilevel studies. First, the secondary data employed for this study were derived from a school-based study, thus the sampling from the original study (Add Health) was conducted at the level of th e school, rather than the neighborhood. Consequently, the original study was not de signed to ensure a diverse range of neighborhoods, in terms of sociodemographic composition, were included in the sample.

PAGE 197

184 Future studies will need to explore the pot ential implications of this limitation when designing multilevel studies that utilize secondary data. Second, the Add Health data are nationally representative, which may have pros and cons in assessing neighborhood effects, pa rticularly on access to health promoting resources, such as access to health care. Although nationally repr esentative data can improve the external validity of the study, the effects of specific area-based factors associated with access to health care may go unrecognized in a national study (Sturm et al., 2003). For example, one community may have implemented a campaign to improve access to health care in resource-poor nei ghborhoods with a high proportion of Black and poor residents. Thus, the members in this community may not have difficulty accessing health care, while members in another communi ty without these services may not be able to access health care. In a national study, these two findings may cancel one another and give the impression the neighborhood or comm unity had no impact on access to health care. This is called the omitted variable bias, which researchers must consider when designing any study. Neighborhood studies may not be as generalizable to other communities, but the dynamics in one’s community may play an important role in the feasibility of certain strategies targeting social disparities in access to health care. Future methodological discussions and comparisons between nationally representative and neighborhood based studies are needed to explore these issues further. Directions for Future Research This study revealed few social disparities in access to health care in this sample of sexually experienced adolescent females. However, the primary reason why this sample of adolescents received health care, whethe r it was the need for primary, secondary, or

PAGE 198

185 tertiary health care services, was unable to be determined by this study. Future, longitudinal studies are necessa ry to explore potential social disparities in access to primary health care services among sexually experienced adolescents to better understand the predisposing factors, enabli ng factors, and health needs that may impact their access to and utilization of primary health care se rvices. In addition, future studies should include other racial and ethnic groups to explore potential soci al disparities in access to health care among other vul nerable populations. Multilevel studies are also needed examin ing the interactions between health care providers, characteristics of th e health care setting, and indivi dual characteristics to gain a better understanding of how patient -provider relationships in the context of the health care setting may impact access t o, utilization of, and the quality of health care among this vulnerable population of adolescents. These multilevel studies should also incorporate information regarding the neighborhood and co mmunity context, as individual patients and providers are nested with in the health care setting as well as the broader social structure and social context of the local community. However, to increase the likelihood of having an adequate sample si ze and diversity at all levels of analysis, researchers need to carefully consider the type of data, whether it is primary or seconda ry, that they utilize for analyses. Unfortunately, few existing data sets are available to examine the interactions between individuals, providers, health care settings, and communities, but data from the Community Tracking Survey or neighborhood-based stud ies, such as the Project on Human Development in Chicago Neighborhoods or the Los Angeles Family and Neighborhood Survey may be applicable depending on the research question.

PAGE 199

186 Further study is also needed on crosslevel interactions between neighborhoods and individuals to explore th e extent to which the neighbor hood racial and socioeconomic context may contribute to social disparitie s in access to health care. In addition, researchers should also consider examini ng potential neighborhood so cial processes and material resources that may impact access to health care, such as neighborhood social networks, engagement in risky health behavi ors, and the availability of health care providers (Prentice, 2006). These social processes and resources may be potential mediating and/or moderating factors on the relationships between the neighborhood racial and socioeconomic context and access to hea lth care. Finally, multilevel studies should also include higher levels of an alyses, as neighborhoods are nest ed within cities and cities within states. State policies that focus on the distribution of health care resources as well as social welfare services may have a mode rating influence on the relationships between the neighborhood racial and soci oeconomic context and access to health care. Elucidating these relationships will help to facilitate the development of state, neighborhood and individual level interventions that may more effectively prevent social disparities in access to health care. Conclusions Recent research has suggested that access to and utilization of preventive health care services may play an important role in mitigating social disparities in health (Starfield, Shi, & Macinko, 2005) Although this study was unable to differentiate if the health care services that the adolescents rece ived were primary, secondary or tertiary in nature, the safety net for access to health care services among the most socially disadvantaged adolescents in this study appeared to be intact due to the lack of racial and

PAGE 200

187 socioeconomic disparities in the receipt of a routine physical, unmet h ealth needs, and the receipt of contraceptive services. Unfortunately th ough, significant racial and socioeconomic disparities in STI’s, pregnancy, and general health st atus were revealed, which indicates the adolescents in this study ha d a greater need for pr eventive health care services. Racial and socioeconomic disparities in h ealth and health care are a violation of human and civil rights principles (Bamba s & Casas, 2003; Braveman & Gruskin, 2003; Smith, 2005; United Nations, 1948). As public hea lth professionals we must continue to research the extent to which social disparitie s in health and health care exist and their potential relationships with racial and socioeconomic residential segregation. By understanding the factors that c ontribute to social disparities in health and health care, more effective interventions can be develope d to ameliorate these social injustices.

PAGE 201

188 References Abma, J. C., Martinez, G.M., Mosher, W.D ., & Dawson, B.S. (2004). Teenagers in the United States: Sexual activity, cont raceptive use, and childbearing. Vital Health Statistics No. 23 (24): National Center for Health Statistics. Abma J., Driscoll, A., & Moore, K. (1998). Young women's degree of control over first intercourse: An exploratory analysis. Family Planning Perspectives, 30 12-18. Acevedo-Garcia, D., Lochner, K.A.,Osypuk, T.L., & Subramanian, S.V. (2003). Future directions in residential segregation and health res earch: A multilevel approach. American Journal of Public Health 93, 215-221. Acevedo-Garcia, D. & Lochner, K.A. (2003). Re sidential segregation and health. In L. F. Berkman & I. Kawachi (Eds.), Neighborhoods and Health (pp. 265-287). New York: Oxford University Press. Aday, L.A. (2001). At risk in America: the health and health care needs of vulnerable populations in the United States. San Francisco: Jossey-Bass. Aday, L.A., Begley, C.E., Lairs on, D.R., & Balkrishnan, R. (2004). Evaluating the health care system: Effectivenes s, efficiency, and equity Washington, DC: AcademyHealth. Alan Guttmacher Institute. (1999). Facts in brief : Teen sex and pregnancy. Washington, DC: Author.

PAGE 202

189 American Academy of Pediat rics. (2000). Recommendations for preventive pediatric healthcare. Pediatrics, 105 645-646. American Academy of Pedi atrics. (2001). Sexuality e ducation for children and adolescents. Pediatrics, 108 498-502. American Medical Association. (1997). Guidelines for adolescent preventive services. Chicago: Author. Andersen, R.M. (1995). Revisiting the behavior al model and access to health care: Does it matter? Journal of Health and Social Behavior, 36 1-10. Andersen, R.M., Yu, H., Wyn, R., Davidson, P.L., Brown, E.R., & Teleki, S. (2002). Access to medical care for low-income persons: How do communities make a difference? Medical Care Research and Review, 59 384-411. Anderson, G.F., Reinhardt, U.E., Hussey, P.S ., & Petrosyan V. (2003). It’s the prices, stupid: Why the United States is so different from other countries. Health Affairs, 22 89-105. Andrulis, D.P., & Duchon, L.M. (2005). Hospital care in the 1 00 largest cities and their suburbs, 1996-2002: Implicati ons for the future of th e hospital safety net in metropolitan America New York: SUNY Downstate Medical Center. Ashton, M.R., Cook, R.L., Wiesemfeld, H. C., Krohn, M.A., Zamborsky, T., Scholle, S.H., & Switzer, G.E. (2002). Primary car e physician attitudes regarding sexually transmitted diseases. Sexually Transmitted Diseases, 29 246-252. Bambas, A. & Casas, J.A. (2003). Assessing equ ity in health: Conceptual criteria. In R. Hofrichter (Ed.), Health and social justice: Politi cs, ideology, and inequity in the distribution of disease (pp. 321-334). San Francisco: Jossey-Bass.

PAGE 203

190 Bartman, B. A., Moy, E., & D'Angelo, L.J. (1997). Access to ambulatory care for adolescents: The role of a usual source of care. Journal of Health Care for the Poor and Underserved, 8, 214-226. Beauchamp, D. E. (2003). Public health as social justice. In R. Hofrichter (Ed.), Health and social justice: Politics, ideology, and inequity in the distribution of disease (pp. 267-284). San Francisco: Jossey-Bass. Beeghley, L. (2005). The structure of social stra tification in the United States Boston: Pearson Education. Boekeloo, B.O., Schamus, L.A., Simmens, S. J., Cheng, T.L., O’Connor, K., & D’Angelo, L.J. (1999). A STD/HIV prevention tria l among adolescents in managed care. Pediatrics, 103 107-115. Boonstra, H. (2002). Teen pregnanc y: Trends and lessons learned. The Guttmacher Report on Public Policy, 5 7-10. Braveman, P., & Gruskin, S. (2003) Defining equity in health. Journal of Epidemiology and Community Health, 57, 254-258. Brindis, C. (2002). Advancing the adolescent reproductive he alth policy agenda: Issues for the coming decade. Journal of Adolescent Health, 31 296-309. Brindis, C., Park, M.J., Ozer, E.M., & Irw in, C.E. Jr. (2002). Adolescents' access to health services and clinical preventive health care: crossing the great divide. Pediatric Annals, 31 575-581. Brooks-Gunn, J., Duncan, G.J., Klebanov, P.K ., & Sealand, N. (1993). Do neighborhoods influence child and adolescent development? The American Journal of Sociology, 99 353-395.

PAGE 204

191 Brooks-Gunn, J., McCormick, M.C., Klebanov, P.K., & McCarton, C. (1998). Health care use of 3-year-old low bi rth weight premature children: Effects of family and neighborhood poverty. Journal of Pediatrics, 132 971-975. Centers for Disease Control & Prevention. (2004). Sexually Transmitted Disease Surveillance: 2003 Atlanta, GA: U.S. Depart ment of Health and Human Services. Centers for Disease Control & Prevention. (2005). Fact Sheet: HIV/AIDS among youth Atlanta, GA: U.S. Department of Health and Human Services. Chantala, K., & Tabor, J. (1999). Strategies to perform a design-based analysis using Add Health data. Retrieved February 10, 2005, from http://www.cpc.unc.edu/projects /addhealth/files/weight1.pdf Charles, C. Z. (2003). The dynamics of racial residential segregation. Annual Reviews of Sociology, 29 167-207. Cheng, T. L., Savageau, J.A., Sattler, A.L., De Witt, T.G. (1993). Conf identiality in health care. A survey of knowledge, percepti ons, and attitudes among high school students. JAMA, 269 1404-1407. Cohen, J. (1992). A power primer. Psychological Bulletin, 112 155-159. Cook, R.L., Wiesenfeld, H.C., Ashton, M.R ., Khron, M.A., Zamborsky, T., & Scholle, S.H. (2001). Barriers to screening sexually active adolescent women for Chlamydia: A survey of primary care physicians. Journal of Adolescent Health, 28 204-210.

PAGE 205

192 Cradock, A. L., Kawachi, I., Colditz, G. A., Hannon, C., Melly. S.J., Wiecha, J.L., Gortmaker, S.L. (2005). Playground safe ty and access in Boston neighborhoods. American Journal of Preventive Medicine, 28 357-363. Cubbin, C., Santelli, J., Brindis, C.D., & Braveman, P. (2005). Neighborhood context and sexual behaviors among adolescents: Fi ndings from the National Longitudinal Study of Adolescent Health. Perspectives on Sexual and Reproductive Health, 37 125-134. Danielson, R., Marcy, S., Plunkett, A., Wiest, W., &Greenlick, M.R. (1990). Reproductive health counseling fo r young men: What does it do? Family Planning Perspectives, 22 115-121. Darroch, J.E., & Singh, S. Why is teenage preg nancy declining? The roles of abstinence, sexual activity, and contraceptive use. Occasional Report New York: Alan Guttmacher Institute. Davidson, P.L., Andersen, R.M., Wyn, R., & Brown, E.R. (2004). A framework for evaluating safety-net and other community -level factors on access for low-income populations. Inquiry, 41 21-38. DeNavas-Walt, C., Proctor, B.D.,& Mills, R.J. (2005). Income, poverty, and health insurance coverage in the United States: 2003. U.S. Census Bureau: Current Population Reports (P60-226). Washington, DC: U.S. Government Printing Office.

PAGE 206

193 Diez-Roux, A. (2003). The examination of ne ighborhood effects on he alth: Conceptual and methodological issues related to the presence of multiple levels of organization. In L. F. & I. Kawachi & Berkman (Ed.), Neighborhoods and Health (pp. 45-64). New York: Oxfo rd University Press. Elixhauser, A., Machlin, S.R., Zodet, M.W ., Chevarley, F.M., Patel, N., McCormick, M.C., & Simpson, L. (2002). Health care for children and youth in the United States: 2001 annual report on access, utilization, qua lity, and expenditures. Ambulatory Pediatrics, 2 419-437. Elster, A., Jarosik, J., VanGeest, J., & Flem ing, M. (2003). Racial and ethnic disparities in health care for adolescents. Archives of Pediatrics and Adolescent Medicine, 157, 867-874. Ewing, R., Schmid, T., Killingsworth, R., Zlot, A., & Raudenbush, S. (2003). Relationship between urban sprawl and physical activity, obesity, and morbidity. American Journal of Health Promotion, 18 47-57. Finer, L. B., & Zabin, L.S. (1998). Does the tim ing of the first family planning visit still matter? Family Planning Perspectives, 30 30-33, 42. Fiscus, L.C., Ford, C.A., & Miller, W.C. (2004). Infrequency of sexually transmitted disease screening among sexually expe rienced U.S. female adolescents. Perspectives on Sexual and Reproductive Health, 36 233-238. Ford, C., Bearman, P.S., & Moody, J. (1999). Forgone health care among adolescents. JAMA, 282 2227-2234. Fox, J. (1997). Applied regression analys is, linear models, and related methods. Thousand Oaks, CA: Sage.

PAGE 207

194 Frost, J.J. (2001). Public or private provide rs? U.S. women’s use of reproductive health services. Family Planning Perspectives, 33 4-12. Geronimus, A. T. (2003). Damned if you do: Culture, identity, privilege, and teenage childbearing in the United States. Social Science and Medicine, 57 881-893. Goldsmith, W.W. & Gunn, C. (2003). Politic al economy. In K. Christensen & D. Levinson (Eds.), Encyclopedia of Community Thousand Oaks, CA: Sage. Grunbaum, J.A. Kann, L., Kinchen, K., Ross, J., Hawkin s, J., Lowry, R., et al. (2004). Youth risk behavior surv eillance -United States, 2003. MMWR, 53 1-29. Guagliardo, M. F., Ronzio, C.R., Cheung, I., Chacko, E., & Joseph, J.G. (2004). Physician accessibility: An urban ca se study of pediatric providers. Health & Place, 10 273-283. Haas, J. S., Phillips, K.A., Sonneborn, D., McCulloch, C.E., Baker, L.C., Kaplan, C.P., Perez-Stable, E.J., & Laing, S. (2004). Variation in access to health care for different racial/ethnic groups by the raci al/ethnic composition of an individual's county of residence. Medical Care, 42 707-714. Harris, K.M., Florey, F., Tabor, J., Bearman, P.S., Jones, J., & Udry, J.R. (2003). The National Longitudinal Study of Adolescen t Health: Research Design. Retrieved April 20, 2005, from http:// www.cpc .unc.edu/projects/addhealth/design. Henshaw, S. K. (2004). U.S. teenage pregnancy statistics w ith comparative statistics for women aged 20-24 New York: The Alan Guttmacher Institute. Henshaw, S.K. (1998). Unintended pregnancy in the United States. Family Planning Perspectives, 30 24-29 & 46.

PAGE 208

195 Hofrichter, R. (2003). The politics of health inequalities: Contes ted terrain. In R. Hofrichter (Ed.), Health and social justice: Politi cs, ideology, and inequity in the distribution of disease. (pp. 1-56). San Francisco: Jossey-Bass. House, J. S., & Williams, D.R. (2003). Understanding and reducing socioeconomic and racial/ethnic disparities in h ealth. In R. Hofrichter (Ed.), Health and social justice: Politics, ideology, and inequ ity in the distri bution of disease (pp. 89131). San Francisco: Jossey-Bass. Hox, J. J. (1995). Applied multilevel analysis Amsterdam: TT-Publikaties. Institute of Medicine. (1993). Access to he alth care in America. Washington, DC: National Academy Press. Institute of Medicine. (2003). Unequal tr eatment: Confronting racial and ethnic disparities in healthcare. Washi ngton, DC: National Academy Press. Jargowsky, P.A. (2003). Stunning progress, hidden problem s: The dramatic decline of concentrated poverty in the 1990s Washington DC: The Brookings Institution. Jargowsky, P. A. (1996). Take the money and run: economic segregation in U.S. metropolitan areas. American Sociological Review, 61 984-998. Jencks, C. & Mayer, S.E. (1990). The soci al consequences of growing up in a poor neighborhood. In L.E. Lynn and M.G. H. McGeary (Ed.), Inner-City Poverty in the United States (pp. 111-186). Washington, D.C.: National Academy of Press. Jones, R. K., Purcell, A., Singh, S., & Finer, L.B. (2005). Adolescents' reports of parental knowledge of adolescents' use of sexual he alth services and their reactions to mandated parental notification fo r prescription contraception. JAMA, 293 340348.

PAGE 209

196 Kawachi, I., Daniels, N., & Robinson, D.E. (2005). Health disparities by race and class: Why both matter. Health Affairs, 24 343-352. Kirby, D. (2002). Effective approaches to reducing adolescent unprotected sex, pregnancy, and childbearing. The Journal of Sex Research, 39 51-57. Kirby, D., Coyle, K., & Gould, J.B. (2001). Manifestations of pove rty and birthrates among young teenagers in Calif ornia zip code areas. Family Planning Perspectives, 33, 63-69. Kirby, J. B., & Kaneda, T. (2005). Neighbor hood socioeconomic disadvantage and access to health care. Journal of Health and Social Behavior, 46 15-31. Komaromy, M., Grumbach, K., Drake, M., Vranizan, K., Lurie, N., Keane, D., & Bindman, A.B. (1996). The role of Black and Hispanic physicians in providing health care for underserved populations. New England Journal of Medicine, 334 1305-1310. Kreft, I.G.G. (1996). Are multilevel techni ques necessary? An overview, including simulation studies. Retrieved May 26, 2005, from http://www.calstatela.edu/faculty/ik reft/quarterly/quarterly.html Krieger, N. (2001). A glossary for social epidemiology. Journal of Epidemiology and Community Health, 55, (693-700). Krieger, N. (2000). Discrimination and healt h. In L. F. Berkman & I. Kawachi (Eds.), Social epidemiology (pp. 36-75). New York: Oxford.

PAGE 210

197 Krieger, N., Chen, J.T., Waterman, P.D., Rehkopf, D.H., & Subramanian, S.V. (2005). Painting a truer picture of US socioeconomi c and racial/ethnic health inequalities: The Public Health Disp arities Geocoding Project. American Journal of Public Health, 95, 312-323. Krieger N, Waterman PD, Chen JT, Re hkopf DH, Subramanian SV. Geocoding and monitoring US socioeconomic inequalities in health: an introduction to using area-based socioeconomic measures -Th e Public Health Disparities Geocoding Project monograph. (n.d.).Boston, MA: Harvard School of Public Health. Retrieved June 27, 2005, from http://www .hsph.harvard.edu/thegeocodingproject/ Krieger, N., Waterman, P.D., Chen, J.T., Soobader, M.J., & Subramanian, S.V. (2003). Monitoring socioeconomic inequalities in sexually transmitted infections, tuberculosis, and violence: Geocoding and choice of area-based socioeconomic measures--The Public Health Disparities Geocoding Project (US). Public Health Reports, 118, 240-260. Krieger, N., Williams, D.R., & Moss, N.E. ( 1997). Measuring social class in U.S. public health research: Concepts, methodologies, and guidelines. Annual Review of Public Health, 18 341-78. Kushell, M., & Bindman, A.B. (2004). Health care lobbying: Time to make patients the special interest. American Journal of Medicine, 116 496-7. Landers, S.H., & Sehgal, A.R. (2004). Health care lobbying in the United States. American Journal of Medicine, 116 474-477.

PAGE 211

198 Lewis, L.B., Sloane, D.C., Nascimento, L.M ., Diamant, A.L., Guinyard, J.J., Yancey, A.K., & Flynn, G. (2005). African Americans’ access to healthy food options in South Los Angeles restaurants. American Journal of Public Health, 95 668-673. Leventhal, T., & Brooks-Gunn, J. (2000). The ne ighborhoods they live in: The effects of neighborhood residence on child and adolescent outcomes. Psychological Bulletin, 126 309-337. Lieu, T. A., Newacheck, P.W., & McManus, M. A. (1993). Race, ethnicity, and access to ambulatory care among U.S. adolescents. American Journal of Public Health, 83 960-965. Liu, X., Spybrook, J., Congdon, R., & Raudenbus h, S. (2005). Optimal Design for multilevel and longitudinal research. Survey Res earch Center of the Institute for Social Research: University of Michigan. Lurie, N. (2004). Measuring disparities in access to care. In E. K. Swift (Ed.). Guidance for the National Healthcare Disparities Report Washington, D.C.: National Academies of Press. Lynch, J., & Kaplan, G. (2000). Socioeconomic position. In L. F. Berkman & I. Kawachi (Eds), Social epidemiology (pp. 13-35). New York: Oxford. McCanne, D.R. (2004). A national health in surance program for the United States. PLoS Medicine, 1 e39. Macintyre, S., & Ellaway, A. (2003). Neighborhoods and health: An overview. In L. F. Berkman & I. Kawachi (Eds.). Neighborhoods and health (pp. 20-44). New York: Oxford University Press, Inc.

PAGE 212

199 Manlove, J., Terry-Humen, E., Papillo, A.R ., Franzetta, K., Williams, S., & Ryan, S. (2001). Background for community-level work on positive reproductive health in adolescence: Reviewing the literature on contributing factors Child Trends. Massey, D. S., & Denton, N.A. (1989). The di mensions of residential segregation. Social Forces, 67 281-315. Massey, D.S., & Fischer, M.J. (1999). Does risi ng income bring integration? New results for Blacks, Hispanics, and Asians in 1990. Social Science and Research, 28 (316326). Miller, B. C., Monson, B.H., & Norton, M.C. (1995). The effects of forced sexual intercourse on white female adolescents. Child Abuse and Neglect, 19, 12891301. Millstein, S.G., Igra, I.V., & Gans, J. (1996). Delivery of STD/HIV preventive services to adolescents by primary care physicians. Journal of Adolescent Health, 19 249257. Moore, K.A., Morrison, D.R., & Greene, A. D. (1997). Effects on the children born to adolescent mothers. In R.A. Maynard (Ed.) Kids having kids: Economic costs and social consequences of teen pregnancy (pp 145-180). Washington, DC: Urban Institute Press. Moore, K.A., Myers, D.E., Morrison, D.R., Brown, B.V., Nord, C.W ., & Edmonston, B. (1993). The effects of the timing of first birth on young women's poverty. Journal of Research on Adolescence 3, 393-422.

PAGE 213

200 Morland, K.,Wing, S., Diez-Roux, A., & Poole, C. (2002). Neighborhood characteristics associated with the location of f ood stores and food service places. American Journal of Preventive Medicine, 22 23-29. Morrison, K.R., Andersen, R., & Aday, L. A. (1998). Understanding the context of healthcare utilization: Assessing environm ental and provider-related variables in the behavioral model of utilization. Health Services Research, 33 571-594. Mosher, W.D., Martinez, G.M., Chandra, A ., Abma, J.C., & Wilson, S.J. (2004). Use of contraception and use of family planni ng services in the United States: 19821992. Advance Data (No. 350). Retrieved February 16, 2005, from http://www.cdc.gov/nchs/data/ad/ad350.pdf Neuman, W.L. (2003). Social research methods: Qualitative and Quantitative Approaches. Boston: Allyn and Bacon. Newacheck, P. W., Hughes, D.C., Hung, Y. Y., Wong, S., & Stoddard, J.J. (2000). The unmet health needs of America's children. Pediatrics, 105 989-997. Newacheck, P. W., Hung, Y.Y., Park, M.J ., Brindis, C.D., & Irwin, C.E. (2003). Disparities in adolescent health and h ealth care: Does socioeconomic status matter? Health Services Research, 38 1235-1252. Office of Management and Budget. (2000). Standards for defining metropolitan and micropolitan statistical areas. Federal Register 65 82228-82238. Orr, D.P., Langefeld, C.D., Katz, B.P., &Cai ne, V.A. (1996). Behavi oral intervention to increase condom use among high-risk female adolescents. Journal of Pediatrics, 128 288-295.

PAGE 214

201 Pamuk, E., Makuc, D., Heck, K., Reuben, C., & Lochner, K. (1998). Socioeconomic status and health chartbook Hyattsville, MD: National Center for Health Statistics. Porter, L.E., & Ku, L. (2000). Use of reproductive health se rvices among young men, 1995. Journal of Adolescent Health, 27 186-194. Prentice, J.C. (2006). Neighborhood effects on primary care access in Los Angeles. Social Science and Medicine, 62 1291-1303. Putsch, R. W., & Pololi, L. (2004). Distri butive justice in Am erican healthcare: Institutions, power, and the equitable care of patients. The American Journal of Managed Care, 10 SP45-SP53. Quadagno, J. (2004). Why the United States has no national health insurance: Stakeholder mobilization agai nst the welfare state, 1945-1996. Journal of Health and Social Behavior, 45S 25-44. Raj, A., Silverman, J.G., & Amaro, H. (2000) The relationship between sexual abuse and sexual risk among high school students: Findings from the 1997 Massachusetts Youth Risk Behavior Survey. Maternal and Child Health Journal, 4, 125-134. Raphael, D. (2003). A society in decline: The political, economic, and social determinants of health inequalities in th e United States. In R. Hofrichter (Ed.). Health and social justice: Politics, ideo logy, and inequity in the distribution of disease. (pp. 59-88). San Fran cisco: Jossey-Bass. Raudenbush, S.W. (1997). Statis tical analysis and Optimal Design for cluster randomized trials. Psychological Methods, 2 173-185.

PAGE 215

202 Raudenbush, S. W., & Bryk., A.S. (2002). Hierarchical linear mode ls: Applications and data analysis methods Thousand Oaks, Ca: Sage. Raudenbush, S.W., Spybrook, J., Liu, X.F ., & Congdon, R. (2005). Optimal Design for Longitudinal and Multilevel Research, Ve rsion 1.55: Software Documentation. Rawlings, L., Harris, L., Turner, M.A., & Padilla, S. (2004). Race and residence: Prospects for stable neig hborhood integration. Retrieved November 6, 2005, from http://www.urban.org/UploadedPDF/310985_NCUA3.pdf Ray, M.N., Wall, T., Casebeer, L., Weissman, N., Spettell, C., Abdolrasulnia, M. et al. (2005). Chlamydia screening of at-ris k young women in managed health care: Characteristics of top-perfo rming primary care offices. Sexually Transmitted Diseases, 32 382-386. Reddy, D. M., Fleming, R., & Swain, C. (2002). E ffect of mandatory pa rental notification on adolescent girls' use of sexual health care services. JAMA, 14 710-714. Robert, S. A. (1999). Socioeconomic position and health: The independent contribution of community socioeconomic context. Annual Review of Sociology, 25 489-516. Santelli, J. S., Abma, J., Ventura, S., Lindbe rg, L., Morrow, B., Anderson, J.E., Lyss, S., & Hamilton, B.E. (2004). Can changes in sexual behaviors among high school students explain the decline in t een pregnancy rates in the 1990s? Journal of Adolescent Health, 35 80-90. Sarigiani, P. A., Ryan, L., & Petersen, A.C. (1999). Prevention of high-risk behaviors in adolescent women. Journal of Adolescent Health, 25 109-119.

PAGE 216

203 Scher, L.S. (2004). Reducing the risk of teen pregnancy: Assessing the effectiveness of schooland community-based pregnancy prevention interven tions. (Doctoral dissertation, University of Pennsylvania, 2004). Dissertation Abstracts International, 65 843. Shafii, T. & Burstein, G.R. (2004). An overv iew of sexually transmitted infections among adolescents. Adolescent Medicine Clinics, 15 201-214. Shenkman, E., Youngblade, L., & Nackashi J. (2003). Adolescents' preventive care experiences before entry into the Stat e Children's Health Insurance Program (SCHIP). Pediatrics, 112 533-541. Shi, L., & Stevens, G.D. (2005). Vulnerable populations in the United States San Francisco, Jossey-Bass. Simpson, L., Owens, P.L., Zodet, M.W., Chev arley, F.M., Dougherty, D., Elixhauser, A., & McCormick, M.C. (2005). Health care fo r children and youth in the United States: Annual report on patterns of coverage, utilization, quality, and expenditures by income. Ambulatory Pediatrics, 5 6-44. Slade, E.P. (2003). The relationship between sch ool characteristics and the availability of mental health and related health services in middle and high schools in the United States. Journal of Behavioral Health Services and Research, 30 382-392. Smelser, N.J., Wilson, W.J., & Mitchell, F. (Eds.). (2001). America becoming: Racial trends and their consequences: Volume II. Washington, DC: National Academy Press. Smith, D. B. (2005). Racial and ethnic health disparities and the unf inished civil rights agenda. Health Affairs, 24 317-324.

PAGE 217

204 Snijders, T., & Bosker, R. (1999). Multilevel analysis: An introduction to basic and advanced multilevel modeling. Thousand Oaks, CA: Sage. Society for Adolescent Medicine. (2004). Acces s to health care for adolescents and young adults: Position paper for the Society of Adolescent Medicine. Journal of Adolescent Health, 35 342-344. South, S. J. & Baumer, E.P. (2000). D eciphering community and race effects on adolescent premarital childbearing. Social Forces, 78 1379-1407. South, S.J., & Crowder, K.D. (1999). Ne ighborhood effects on family formation: Concentrated poverty and beyond. American Sociological Review, 64 113-32. Starfield, B., Shi, L., & M acinko, J. (2005). Contribution of primary care to healthy systems and health. Millbank Quarterly, 83 457-502. Starfield, B., & Shi, L. (2004). The medica l home, access to care, and insurance: A review of evidence. Pediatrics, 113 1493-1498. Stevens, G. D., & Shi, L. (2003). Racial and ethnic disparitie s in the primary care experiences of children: A review of th e literature. Medical Care Research and Review, 60 3-30. Sturm, R., Ringel, J.S., & Andreyeva, T. (2003). Geographic disparities in children’s mental health care. Pediatrics, 112 e308-e315. Subramanian, S. V., Chen, J.T., Rehkopf, D.H., Waterman, P.D., & Krieger, N. (2005). Racial disparities in context: a multileve l analysis of neighborhood variations in poverty and excess mortality among bl ack populations in Massachusetts. American Journal of Public Health, 95 260-265.

PAGE 218

205 Subramanian, S. V., Jones, K., & Duncan, C. (2003). Multilevel methods for public health research. In L. F. Berkman & I. Kawachi (Eds.). Neighborhoods and Health (pp. 65-111). New York: Oxford University Press, Inc. Sucoff, C. A. & Upchurch, D.M. (1998) Neighborhood context and the risk of childbearing among metropolitan -area black adolescents. American Sociological Review, 63 571-585). Tabachnick, B.G., & Fidell, L.S. (2001). Using multivariate statistics Boston: Allyn & Bacon. Terry-Humen, E., Manlove, J., & Moore, K. A. (2005). Playing catch-up: How children born to teen mothers fare. Washington, DC: National Campaign to Prevent Teen Pregnancy. Thomas, R.K. (2003). Society and health: Sociology for health professionals New York: Kluwer Academic. Torkko, K.C., Gershman, K., Crane, L.A., Ha mman, R., & Baron, A. (2000). Testing for Chlamydia and sexual history taking in adolescent females: Results from a statewide survey of Colo rado primary care providers. Pediatrics, 106 e32. United Nations. (1948). Universal declarati on of human rights. Retrieved March 12, 2006, from http://www.un.org/ Overview/rights.html U.S. Census Bureau. (1994). Census tracts and block numbering areas. In Geographic areas reference manual (chap. 10). Retrieved from http://www.census.gov/geo/www/GARM/Ch10GARM.pdf U.S. Department of Health and Human Services. (2000). Healthy people 2010 (2nd Ed.). Washington, DC: U.S. Government Printing Office.

PAGE 219

206 U.S. Department of Health and Human Servic es. (2003). National hea lthcare disparities report. Rockville, MD: Agency for Healthcare Research and Quality. Ventura, S.J., & Bachrach, C.A. (2000). No nmarital childbearing in the United States, 1940-99. National Vital Statistics Report, 48 1-40. Whiteis, D.G. (1998). Third world medicine in first world cities: Capital accumulation, uneven development and public health. Social Science and Medicine, 47 795808. Whiteis, D.G. (1997). Unhealthy cities: Corporate medicine, community economic underdevelopment, and public health. International Journal of Health Services, 27 227-242. Wickrama, K.A.S., & Bryant, C. (2003). Comm unity context of social resources and adolescent mental health. Journal of Marriage and Family, 65, 850-866. Williams, D., & Collins, C. (2001). Racial resi dential segregation: A fundamental cause of racial disparities in health. Public Health Reports, 116 404-416. Wilson, K., & Klein, J.D. (2000). Adolescents who use the emergency department as their usual source of care. Archives of Pediatrics and Adolescent Medicine, 154 361-365. Winter, L., & Breckenmaker, L.C. (1991). Ta iloring family planning services to the special needs of adolescents. Family Planning Perspectives, 23 24-30. Yu, S.M., Bellamy, H.A., Kogan, M.D., Dunba r, J.L., Schwalberg, R.H., & Schuster, M.A. (2002). Factors that influence th e receipt of recommended preventive pediatric health and dental care. Pediatrics, 110 e73.

PAGE 220

207 Zenk, S.N., Schulz, A.J., Israel, B.A., Ja mes, S.A., Bao, S., & Wilson, M.L. (2005). Neighborhood racial composition, neighborhood poverty, and the spatial accessibility of supermarkets in metropolitan Detroit. American Journal of Public Health, 95 660-667. Zierler, S., Krieger, N., Tang, Y., Coady, W., Siegfried, E., DeMaria, A., & Auerbach, J. (2000). Economic deprivation and AI DS incidence in Massachusetts. American Journal of Public Health, 90 1064-1073. Zimmer-Gembeck, M. J., Alexander, T., & Ny strom, R.J. (1997). Adolescents report their need for and use of health care services. Journal of Adolescent Health, 21 388-399.

PAGE 221

208 Appendices

PAGE 222

209 Appendix A: Measures Construct Items Source Dependent Variables Access to Health Care (1) Receipt of Routine Physical (2) Receipt of Contraceptive Services (3) Presence of Unmet Health Needs (1) In the past year have you had a routine physical examination? Response Options o Yes or No (Dichotomous) (2 ) Have you ever received a birth control method from a doctor or clinic? Response Options o Yes or No (Dichotomous) (3) Has there been any time over the past year when yo u thought you should get medical care, but you did not? Response Options o Yes or No (Dichotomous) (1) Adolescent (2) Adolescent (3) Adolescent Level I Independent Variables Race (1) What is your race? You may se lect more than one response. Response Options o White, Black or African American, American Indian or Native American, Asian or Pacific Islander, other, or don’t know (2) Which one category best describes your racial category? Response Options o White, Black or African American, American Indian or Native American, Asian or Pacific Islander, other, or don’t know (3 ) Are you of Hispanic or Latino origin? Response Options o Yes or No ** Created race variable fr om these three questions** Non-Hispanic Black and Non-Hispanic White Response Options o Yes or No (Dichotomous) (1) Adolescent (2) Adolescent (3) Adolescent Continued on Next Page

PAGE 223

210 Appendix A: Measures (continued) Construct Items Source Socioeconomic Position (1) Household Income (2) Income to Poverty Ratio (3) Parental Education (1) About how much total income, before taxes did your family receive in 1994? Incl ude your own income, the income from everyone else in your household, and income from welfare benefits, dividends, and all other sources. Response Options o Continuous in thousands (Continuous) (2) About how much total income, before taxes did your family receive in 1994? Incl ude your own income, the income from everyone else in your household, and income from welfare benefits, dividends, and all other sources household size Response Options Created o 0-100%, 101-200%, 201-300% 301-400%, and >400% (Dichotomous) (3) How far did you go in school? And How far in school did she or he (mom or dad) go? Response Options o 8th grade or less, more than 8th grade, but did not graduate from high school, went to a business, trade, or vocational school instead of high school, high school graduate, completed a GED, went to a business, trade or vocational school after high school, went to college, but did not graduate, graduated from a college or university, and professional tr aining beyond a 4-year college or university Response Options Created o Less than high school, high school degree or GED, so me college or technical training, college degree or more (Dichotomous) (1) Parent (2) Parent & Adolescent (3) Parent & Adolescent Health Needs (1) History of STI (2) History of Pregnancy (3) General Health Status (1) Have you ever been told by a doctor or a nurse that you had…chlamydia, syphilis, gonorrhea, HIV or AIDS, genital herpes, genital warts, or trichomoniasis? Response Options o Yes or No ** Created STI variableHistory of STI if adolescents reported ever having one of the above infections Response Options o Yes or No (Dichotomous) (2) Have you ever been pregnant? Be sure to include if you are currently pregnan t and any past pregnancy that ended in an abortion, stillbirth, miscarriage, or a li ve birth after which the baby died Response Options o Yes or No (Dichotomous) (3) In general, how is your health? Would you say… Response Options o Excellent, very good, good, fair, and poor (continuous) Continued on Next Page

PAGE 224

211 Appendix A: Measures (continued) Construct Items Source Adolescent Health Beliefs (1) Pregnancy Undesirable (2) Barriers to Contraceptives (3) Mom Disapprove of Sex (4) Mom Disapprove of Birth Control (1) Composite made from following items: (a) G etting pregnant at this time in y our life is one of the worst th ings that could happen to you (b) If you got pregnant, it would be embarrassing for your family (c) If you got pregnant, it would be embarrassing for you (d) If you got pregnant, you would be forced to grow up too fast (e) If you got pregnant, you have had to decide whether or not to have the baby, and that would be stressful and difficult (2) Composite made from following items: (a) In general, birth control is too much of a hassle to use (b) In general, birth control is too expensive to buy (c) It takes too much planning ahead of time to have birth control on hand when you’re going to have sex (d) It is too hard to get a boy to use birth control with you (e) For you, using birth control inte rferes with your sexual enjoyment (f) Using birth control is morally wrong (3) Single Item: How would she (mom) feel about your hav ing sex at this time in your life? (4) Single Item: How would she feel about your using birt h control at this time in your life? Response Options for all items o Strongly disagree, disagr ee, neither agree nor disagree, ag ree, and strongly agree (Continuous) Length of Residence in Neighborhood Variable constructed from three questions… (1) How old were you when you moved here to your current residence? (2) Has {child’s name} always lived, since (he/she) was born, in the house or apartm ent building where (he/she) lives now? (3) In what month and year did (he/she) move to the hous e or apartment building (he/ she) lives in now? **Used these questions in combination with adolescent’s age and date of interview to create variable representing length of residence in years that adolescen t lived in neighborhood (continuous) (1) Adolescent (2) Parent (3) Parent Adolescent’s Age Variable constructed from the interview da te and adolescent’s year and month of birth. Equation provided by Add Health (continuous) Adolescent Continued on Next Page

PAGE 225

212 Appendix A: Measures (continued) Construct Items Source Health Insurance (1) What kind of health insurance does {Name} have? Check off all that apply. Response Options o Medicare/SSI, Medicaid, individual or group private co verage (such as Blue Cross or Cigna), prepaid health plan (suc h as an HMO or Cigna), other, none, or don’t know ** Due to small cell sizes in several of these categories, hea lth insurance was dichotomized and dummy coded into yes for those adolescents with health insurance and no for those adolescents without health insurance. (1) Parent Transportation (1) Do you have a valid driver’s li cense (not a dr iver’s permit)? Response Options o Yes or No (Dichotomous) (1) Adolescent Family Structure Variable created from household roster that asked adolescent to name all persons living in househol d and their relationship to the adolescent. Due to small ce ll sizes on some of the response options, the variable was dichotomized as adolescents who lived in a two parent househ old and those who did not. Adolescent School Health Education (1) Re Medical Attention (2) Re Sex Education Adolescents were asked… Please tell me whether you have learned about each of the followi ng things in a class at school… Response Options o Pregnancy, AIDS, and where to go for help with a medical problem **Two variables created: school health ed ucation re medical attention (yes or no) and school health education re sex (pregnancy and AIDS combinedyes or no). Both variables dichotomous. Sex education was not included in multivariate analyses due to small cell sizes of those adolescents who did not receive sex education. Adolescent School Health Care Services (1) Non-Athletic Physical (2) Reproductive Health Care Services School health administrators were asked… For each of the following health-related services, please indicate whether it is provided at your school, is provided by your school district but not at your sc hool, referred to other providers, or neither provided nor referred Response Options o Non-athletic physical STI treatment, family planning services ** The three items were dichotomized and coded as yes if the sc hool administrator reported the service was offered on-site at the school or no if the service was not offered on the site. Two variables created: availability of non-athletic physical (yes or no) and availability of reproductive health care services (combined availability of STI treatment and family planning servicesyes or no). Both items were disaggregated from level of school to the level of the in dividual, and represent whether or not the adolescent attends a school in which thes e health care services are offered. Reproductive health care services were not included in multivariate analyses due to small cell sizes of adolescents who attended a school that offered these services. School Administrator Continued on Next Page

PAGE 226

213 Appendix A: Measures (continued) Construct Items Source Level II Independent Variables Neighborhood Racial Composition (1) Proportion of residents in census tract that were Black-used for al l analyses (continuous) (a) Categories were also developed fo r use with descriptive analyses o 10% or less Black residents living within the census tract, 11% to 50% Black residents living within the census tract, 51% to 90% Black residents living within the census tract, and 90% or more Black residents living within the census tract Contextual 1990 Census Data Neighborhood Socioeconomic Composition (1) Proportion of adults in census tract who were 25 year s of age & older with < high school degree (continuous) (a) Categories were also developed for use with descri ptive analyses o < 25% and 25% of adults who are over 25 years old in census tract with < a high school degree. (2) Proportion of families in the census tract with income belo w the 1989 poverty level (continuous) (a) Categories were also developed fo r use with descriptive analyses o < 20% and 20% of families in census tract with income below 1989 poverty level Contextual 1990 Census Data Total Population (1) Total population of persons residing in the census tract (continuous) Contextual 1990 Census Data Median Age (1) Median age of persons residing within the census tract (continuous) Contextual 1990 Census Data Residential Stability (1) Proportion over persons over 5 years of age living in the same house for past 5 years (continuous) Contextual 1990 Census Data Urbancity (1) Proportion of persons in the census tr act residing within an urban area in wh ich there was a minimum of 50,000 persons (continuous) Contextual 1990 Census Data Geographic Region (1) Four categorical variables repr esenting geographic region of school adolescent attended (categorical) Midwest, Northeast, West & South School Administrator

PAGE 227

214 Appendix B: Summary of Model Building Pro cess and Contribution to Hypothesis Testing Model Features Contribution to Hypothesis Testing Hypothesis I Model I (1) No independent variables in the model so there is only the fixed intercept. (2) The Level I intercept is random (1) The fixed intercept in the null model represents the average odds of access to health care among the sexually experienced adolescen t females in this study. (2) Determines if there is variation in access to health care acro ss neighborhoods and the need to continue HGLM. Tests Hypothesis ID. The Level I intercept remains random for Models I-VI to examine the extent to which the independent variables explain the varia tion in access to health care across neighborhoods. Model II (1) Adolescent race variable is entered into the model (2) Level I intercept is random (3) Random slopes of adolescent race variables entered into equation. (1) Begins testing of Hypothesis IA to determine if th ere are racial disparities in access to health care. (2) Decreases in size and significance of the Level I ra ndom parameter estimate woul d indicate that adolescent race explained some of the variation acro ss neighborhoods in access to health care. (3) Begins testing Hypothesis IE to determine if the re lationship between race and access to health care varied across neighborhoods. Model III (1) Adolescent socioeconomic position variables are entered into the model (parental education and ratio of household income to poverty) (2) Level I intercept is random (3) Random slopes of a dolescent socioeconomic variables entered into re gression equation one at a time. (1) Continues testing Hypothesis IA to determine if ra cial disparities in access to health care exist after adjusting for adolescent socioeconomic position. (2) Begins testing Hypothesis IB to determine the extent to which socioec onomic disparities in access to health care exist after adjusting for adolescent race. (3) Decreases in size and significance of the Level I ra ndom parameter estimate woul d indicate that adolescent socioeconomic position explained some of the variation across neighborhoods in access to health care. (4) Continues testing Hypothesis IE to determine if the relationship between socioeconomic position and access to health care varies across neighborhoods. Model IV (1) Enter in the Level I control variables. (2) Level I intercept is random (1) Completes testing of Hypothesis IA and IB to dete rmine the extent to which racial and socioeconomic disparities in access to health care exist after ad justing for one another and the control variables. (2)Decreases in the size and significance of the Leve l I random parameter estimat e would indicate that the control variables explaine d some of the variation across nei ghborhoods in access to health care. Model V (1) Enter in interacti on term for adol escent race and adolescent socioeconom ic position (parental education attainment). (2) Random slopes of inter action term entered into regression equation. (3) Level I intercept is random. (1) Begins testing of Hypothesis IC to determine the ex tent to which the relations hip between adolescent race and access to health care is a function of adolescent socioeconomic position. (2) Continues testing of Hypothesis IE to determine if the interaction between a dolescent race, socioeconomic position and access to health care varies across neighborhoods. (3) Decreases in the size and signif icance of the Level I random paramete r estimate would indicate that the interaction variables explained some of the variat ion across neighborhoods in access to health care. Continued on Next Page

PAGE 228

215 Appendix B: Summary of Model Building Process and Contribution to Hypothesi s Testing (continued) Model Features Contribution to Hypothesis Testing Model VI (1) Enter in interacti on term for adol escent race and adolescent socioeconomic position (ratio of household income to poverty). (2) Random slopes of inter action term entered into regression equation. (3) Level I intercept is random. (1) Completes testing of Hypothesis IC to determine th e extent to which the relationship between adolescent race and access to health care is a function of adolescent socioeconomic position (2) Completes testing of Hypothesis IE to determine if the interaction between a dolescent race, socioeconomic position and access to health care varies across neighborhoods. (3) Decreases in the size and signif icance of the Level I random paramete r estimate would indicate that the interaction variables explained some of the variat ion across neighborhoods in access to health care. Hypothesis II Model I (1) No independent variables in the model so there is only the fixed intercept. (2) The Level I intercept is random (1) The fixed intercept in the null model represents the average odds of access to health care among the sexually experienced adolescen t females in this study. (2) Determines if there is variation in access to health care acro ss neighborhoods and the need to continue multilevel modeling. Tests Hypothesis ID. Level I inter cept remains random for Models I-VI to examine the extent to which the independent variables explain the variation in access to health care across neighborhoods. Model II (1) Neighborhood race variable (proportion of Black residents in neighborhood) is entered into the model (2) Level I intercept is random (1) Begins the testing of Hypothesis IIA to determine if the neighborhood racial context is associated with the average odds of access to health care. (2) Decreases in the size and sign ificance of the Level I random para meter estimate would indicate that neighborhood racial context explained some of the vari ation across neighborhoods in access to health care. Model III (1) Neighborhood soci oeconomic variables are entered into the model (proportion of low educated and poor residents in the neighborhood) (2) Level I intercept is random (1) Continues testing of Hypothesis IIA to determine if the neighborhood racial context is associated with access to health care exist after adjusti ng for the neighborhood socioeconomic context. (2) Begins testing of Hypothesis IIB to determine th e extent to which the neighborhood socioeconomic context is associated with access to health care afte r adjusting for the neighborhood racial context. (3) Decreases in size and signific ance of Level I random parameter estimate would indi cate the neighborhood socioeconomic context explained some of the vari ation across neighborhoods in access to health care. Model IV (1) Enter in the Level II control variables. (2) Level I intercept is random (1) Completes testing of Hypothesis IIA and IIB to dete rmine the extent to which the neighborhood racial and socioeconomic context is associated with access to health care after adjusting for one another and the neighborhood control variables. (2)Decreases in the size and significance of the Leve l I random parameter estimat e would indicate that the neighborhood control variables explained some of the vari ation across neighborhoods in access to health care. Model V (1) Enter in interaction term between neighborhood racial and socioeconomic context (proportion of low educated residents in neighborhood). (2) Level I intercept is random. (1) Begins testing of Hypothesis IIC to determine th e extent to which the relationship between neighborhood racial context and access to heal th care is a function of the ne ighborhood socioeconomic context. (2) Decreases in the size and signif icance of the Level I random paramete r estimate would indicate that the neighborhood interaction explained some of the varia tion across neighborhoods in access to health care. Model VI (1) Enter in interaction term between neighborhood racial and socioeconomic context (proportion of poor residents in neighborhood). (2) Level I intercept is random. (1) Completes testing of Hypothesis IIC to determine th e extent to which the rela tionship between adolescent race and access to health care is a function of adolescent socioeconomic position (2) Decreases in the size and signif icance of the Level I random paramete r estimate would indicate that the neighborhood interaction explained some of the varia tion across neighborhoods in access to health care. Continued on Next Page

PAGE 229

216 Appendix B: Summary of Model Building Process and Contribution to Hypothesi s Testing (continued) Model Features Contribution to Hypothesis Testing Hypothesis III Model I Builds on Hypothesis I to test cross-level interactions. (1) All Level I variables are in the model. (2) Level II racial context (proportion of Black residents in the neighborhood) is entered into the model for crosslevel interaction with adolescent race and socioeconomic position. (3) Slopes of cross-level interaction are random. (1)–(2) Completes testing of hypothesis IIIA1 and IIIA 2 to determine the extent to which the relationship between adolescent race and access to health care is a function of the proportion of Black residents within the neighborhood and the extent to which the relati onship between adolescent socioeconomic position and access to health care is a function of the propor tion of Black residents within the neighborhood. (3) Partially tests Hypothesis IIIC to determine if th e cross-level relationship between race, socioeconomic position and access to health care varies across neighborhoods. Model II (1) All Level I variables are in the model. (2) Level II socioeconomic context (proportion of low educated residents in the neighborhood) is entered into the model for cross-level in teraction with adolescent race and socioeconomic position. (3) Slopes of cross-level interaction are random. (1)–(2) Partially tests hypot hesis IIIB1 and IIIB2 to determine the extent to which the relationship between adolescent race and access to health care is a function of the proportion of low educated residents within the neighborhood and the extent to which the relations hip between adolescent socioeconomic position and access to health care is a function of the proportion of low educated residents within the neighborhood. (3) Partially tests Hypothesis IIIC to determine if th e cross-level relationship between race, socioeconomic position and access to health care varies across neighborhoods. Model III (1) All Level I variables are in the model. (2) Level II socioeconomic context (proportion of poor residents in the neighborhood) is entered into the model for cross-level interaction with adolescent race and socioeconomic position. (3) Slopes of cross-level interaction are random. (1)–(2) Completes testing of hypothesis IIIB1 and IIIB2 to determine the extent to which the relationship between adolescent race and access to health care is a function of the proportion of poor residents within the neighborhood and the extent to which the relations hip between adolescent socioeconomic position and access to health care is a function of the propor tion of poor residents within the neighborhood. (3) Completes testing of Hypothesis IIIC to determ ine if the cross-level relationship between race, socioeconomic position and access to he alth care varies across neighborhoods.

PAGE 230

About the Author Jodi Nearns received her Bachelor and Master of Science in Nursing from the University of Florida in 1989 and 1993, respec tively. She worked as a Registered Nurse in the neonatal intensive care unit for four years while pursuing her master’s degree, and then as a Pediatric Nurse Practit ioner in a private pediatric pr actice for nearly nine years. In 2001, Ms. Nearns entered the Ph.D. in Pub lic Health program at the University of South Florida, and focused her program of st udy on the social determinants of adolescent health. During her doctoral training, Ms. Nearns was involved in a variety of mixed methods research projects, taught several under graduate courses, and co-authored several publications and conference pres entations related to child an d adolescent health. Upon graduation, Ms. Nearns will be employed as a Research Assistant Professor in the Institute for Families in Society at the University of South Carolina.


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 001801024
003 fts
005 20070802133148.0
006 m||||e|||d||||||||
007 cr mnu|||uuuuu
008 070802s2006 flu sbm 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0001645
040
FHM
c FHM
035
(OCoLC)162131774
049
FHMM
090
RA425 (ONLINE)
1 100
Nearns, Jodi.
4 245
The contribution of the neighborhood context to social disparities in access to health care among sexually experienced adolescent females
h [electronic resource] /
by Jodi Nearns.
260
[Tampa, Fla] :
b University of South Florida,
2006.
3 520
ABSTRACT: Access to health care is an important resource for sexually experienced adolescent females in the prevention of unintended pregnancy and sexually transmitted infections, including HIV. However, a paucity of research exists regarding the extent to which social disparities in access to health care exist among this vulnerable population of adolescents, including the potential contribution of the neighborhood context. Therefore, the primary aims of this dissertation were to examine (1) the extent to which racial and socioeconomic disparities in access to health care exist among sexually experienced adolescent females, (2) the extent to which access to health care among sexually experienced adolescent females varies across neighborhoods, and (3) the extent to which the neighborhood racial and socioeconomic context contribute to racial and socioeconomic disparities in access to health care among sexually experienced adolescent females.A multilevel design was employed for this di ssertation utilizing secondary data from Wave I of the National Longitudinal Study of Adolescent Health (Add Health). Analyses included hierarchical generalized linear modeling to examine the receipt of a routine physical, the receipt of contraceptive services, and reported unmet health needsamong the dissertation sample of 1,526 sexually experienced Non-Hispanic Black and Non-Hispanic White adolescent females between 15 years to 19 years of age who were dispersed across 546 neighborhoods. After adjusting for a variety of factors that may influence access to health care, the findings revealed no racial disparities and few socioeconomic disparities in access to health care among this sample of adolescents. No significant relationship was noted between the neighborhood racial and socioeconomic context and access to health care or social disparities in access to health care among this sample of adolescents. However, the findings revealed that access to health care among this sample of sex ually experienced adolescent females varied across neighborhoods, above and beyond the individual composition of the neighborhood. Further studies are indicated to explore the underlying factors that contribute to socioeconomic disparities in access to health care among sexually experienced adolescent females, and the potential neighborhood characteristics that may contribute to differential access to health care across neighborhoods among this vulnerable population of adolescents.
502
Dissertation (Ph.D.)--University of South Florida, 2006.
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 216 pages.
Includes vita.
590
Adviser: Melinda S. Forthofer, Ph.D.
653
Disparities.
Racial segregation.
Poverty.
Neighborhoods.
Health care.
Adolescent health.
Multilevel modeling.
690
Dissertations, Academic
z USF
x Public Health
Doctoral.
773
t USF Electronic Theses and Dissertations.
0 856
u http://digital.lib.usf.edu/?e14.1645