|USFDC Home | USF Electronic Theses and Dissertations||| RSS|
This item is only available as the following downloads:
Correlates of Weight in Adolescents: A Path Analysis by Angela T. Sheble A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Psychological and Social Foundations College of Education University of South Florida Co-Major Professor: George Batsche, Ed.D. Co-Major Professor: Kathy Bradley-Klug, Ph.D. John Ferron, Ph.D. Kathleen Hague-Armstrong, Ph.D. Date of Approval: November 3, 2006 Keywords: childhood obesity, obesity, depres sion, binge-eating, st ructural equation modeling Copyright 2006 Angela T. Sheble
Dedication This project is dedicated to my son, Dylan.
Acknowledgements I would like to acknowledge and thank several of my professors and professional colleagues who provided guidance and encourag ement to me on this research study: Kathy Bradley-Klug, Ph.D., George Bats che, Ed.D., John Ferron, Ph.D., Kathi Armstrong, Ph.D., and Marsha Luginbuehl, Ph.D. In addition, I would like to express appreciation for the generous assistance of Margaret Polk, R.N., Lynn Stoechicht, R.N., Rae Daly, Wendy Kahl, LCSW, Iravonia Rawls, M.A., and Raquel Doblas-Shaw, M.A., M.P.H.
i Table of Contents List of Tables ................................................................................................................ ....iii List of Figures ............................................................................................................... ......v Abstract ...................................................................................................................... .......vi Chapter I: Overview........................................................................................................... .1 The Epidemic of Obesity...........................................................................................1 Risk Factors of Adolescent Obesity ..........................................................................3 Other Variables Correlated With Adolescent Obesity ..............................................6 Research Objectives ................................................................................................. .7 Chapter II: Literature Review ..........................................................................................11 Obesity in Adolescents ............................................................................................11 Definition of Obesity ....................................................................................1 2 Prevalence of Adolescent Obesity ................................................................16 Outcomes of Adolescent Obesity ..................................................................20 Risk Factors of Adolescent Obesity ..............................................................22 Binge-Eating Disorder in Adolescents ....................................................................29 Definition of Binge-Eating Disorder ............................................................29 Prevalence of Binge-Eating Disorder ...........................................................31 Risk Factors for Binge-Eating Disorder .......................................................32 Relationship of Binge-Eating Disorder to Obesity .......................................34 Relatio nship of Binge-Eating Disorder to Dieting ........................................35 Depression in Adolescents ......................................................................................37 Definition of Depression ...............................................................................37 Prevalence of Depression ..............................................................................40 Risk Factors for Depression ..........................................................................40 Relations hip of Depression With Binge-Eating Disorder .............................42 Relationship of Depression With Physical Activity .....................................44 Relationship of Depression With Obesity .....................................................45 Conclusion .......................................................................................................... ....48 Research Questions .......................................................................................5 0 Path Model ................................................................................................ ....50
ii Chapter III: Research Methods ........................................................................................51 Design .............................................................................................................. .......51 Participants and Setting ...........................................................................................5 5 Measures ............................................................................................................ .....62 Demogra phic and Physical Activity Questionnaire (DPAQ) .......................62 Body Mass Index (BMI) .....................................................................64 Socio-Economic Status (SES) .............................................................64 Exceptional Student Education (ESE) Status .....................................64 Sex/Grade/Age/Ethnicity ....................................................................65 Physical Activity ............................................................................... ..65 Reynolds Adolescent Depression Scale-2nd Edition (RADS-2) ...................66 Eating Disorder Inventory 2 (EDI-2) .........................................................68 Du tch Eating Behavior Questionnaire (DEBQ) ............................................71 Procedure ........................................................................................................... .....73 Recruitment ............................................................................................... ....73 Obtaining Consent ........................................................................................7 6 Survey Packet Organization ..........................................................................77 Data Collection ........................................................................................... ..77 Post Data Collection Procedures ...................................................................79 Tracking of Participants ................................................................................80 Inter-Rater Agreement ............................................................................................81 Chapter IV: Results .......................................................................................................... 84 Data Analysis ....................................................................................................... ...84 Question 1 ................................................................................................ .....84 Question 2 ................................................................................................ ...105 Question 3 ................................................................................................ ...109 Reliability Estimates on Measurement Instruments .............................................111 Chapter V: Discussion ...................................................................................................113 Question 1 ................................................................................................ ...113 Question 2 ................................................................................................ ...118 Question 3 ................................................................................................ ...121 Implications .............................................................................................. ...122 Assumptions and Limitations .....................................................................124 Directions for Future Research ...................................................................127 Conclusion ................................................................................................ ..130 References .................................................................................................................... ...131 Appendices .................................................................................................................... ..137 Appendix A: Parental Informed Consent .............................................................138 Appendix B: Demographi c and Physical Activity Questionnaire ......................145 About the Author ..................................................................................................End Page
iii List of Tables Table 1 BMI Cutoff Points for Child, Adolescent, Adult Weight Classifications ...........................................................................................14 Table 2 Prevalence of Obesity in U.S. Children, Adolescents, and Adults ...........17 Table 3 Prevalence of Adolescent Obesity by Gender and Race, 1999-2000 .......18 Table 4 DSM-IV-TR Research Crite ria for Binge-Eating Disorder .....................30 Table 5 DSM-IV-TR Criteria for Major Depressive Episode ................................39 Table 6 Demographic Dive rsity of High Schools, District, and State ...................56 Table 7 Distribution of Partic ipants by Gender and School ..................................58 Table 8 Distribution of Particip ants by Ethnicity and Gender ...............................59 Table 9 Distribution of Participants by Ethnicity, Gender, and High School ........60 Table 10 Distribution of Par ticipants by Grade and School ....................................61 Table 11 Distribution of Participants by Free/Reduced Lunch and School .............62 Table 12 Distribution of Participants by ESE (Excep tional Student Education) Status .........................................................................................................63 Table 13 Inter-rater Agreement ................................................................................82 Table 14 Distribution of BMI ..................................................................................85 Table 15 Weight Ca tegory Distribution ...................................................................87 Table 16 Distribution of Age by School ..................................................................88 Table 17 Distribution of Age ...................................................................................89
iv Table 18 Distribution of Physical Activity ..............................................................90 Table 19 Distribution of De pression: RADS-2 T-Score .........................................92 Table 20 Depression Scores Above Cutoff ..............................................................93 Table 21 Distribution of Die ting: DEBQ Scaled Score ..........................................94 Table 22 Distribution of Binge Eating: EDI2 Raw Score .......................................95 Table 23 Distribution of Age at 1st Diet and Age at 1st Binge .................................96 Table 24 Distribution of Have Yo u Ever Been on a Diet Variable .........................96 Table 25 Distribution of Have You Ever Binged Variable ......................................97 Table 26 Distribution of Ev er Dieted and Ever Binged ...........................................98 Table 27 Correlation Matr ix for Entire Data Set .....................................................98 Table 28 Correlation Matrix for High School 1 .....................................................100 Table 29 Correlation Matrix for High School 2 .....................................................100 Table 30 MODEL2 Unstandardized Parameter Estimates a nd Standard Errors ...104 Table 31 MODEL3 Unstandardized Parameter Estimates and Standard Errors ...108 Table 32 Binge-Eating History of Obese and Non-Obese students .......................110 Table 33 Correlation of Binge-Eating a nd Depressive Symptoms for Obese and Non-Obese Students ................................................................................111 Table 34 Precedence of Dieting and Bingeing .......................................................111 Table 35 Reliability Estimates for Measurement Instruments ...............................111
v List of Figures Figure 1 Proposed Path Diagram, Co rrelates of Weight in Adolescents ..................8 Figure 2 BMI growth chart for boys aged 2-20 .......................................................13 Figure 3 MODEL1: Initial Path Diagram, Correlates of Weight in Adolescents ..53 Figure 4 MODEL2: Revised Path Diag ram, Correlates of Weight in Adolescents................................................................................................54 Figure 5 Path Analysis Results: MODEL2 ..........................................................103 Figure 6 Path Analysis Results: MODEL3 ..........................................................107
vi Correlates of Weight in Adolescents: A Path Analysis Angela T. Sheble ABSTRACT This study examined the interrelationships between adolescent weight and several other variables thought to impact weight an d obesity: physical ac tivity, depressive symptoms, binge-eating symptoms, dieting, socio-economic status, special education status, gender, and ethnicity. The sample consisted of 251 high school students in rural Florida who completed measures of depr ession, binge-eating, dieting, and physical activity. Measurement instruments include d the Reynolds Adolescent Depression Scale2nd Edition (RADS-2), the bulimia scale of th e Eating Disorder Inventory-2 (EDI-2), the dieting scale of the Dutch Eating Beha vior Questionnaire (D EBQ), and a physical activity questionnaire derived from the Y outh Risk Behavior Scale for Students (YRBSS). The study u tilized path an alysis, a group correlati onal design, to determine whether the proposed path model fit the data Obese and non-obese students also were compared with regard to a) the levels of binge-eating symp toms, and b) the relationship between binge-eating and depres sion. Path analysis results were not statistically or clinically significant, suggesting a poor fit of the model to the data. Results indicated 19% of participants were obese and 20% we re overweight. More than three times as many obese students than non-obese student s reported experiencing a binge-eating experience at some time in the past. Howeve r, on the bulimia scal e of the EDI-2, obese and non-obese participants did not differ statistically in their responses. Stat istically but
vii not clinically significant correlations were revealed between depr ession and binge-eating for the sample and also for non-obese studen ts. For the sub-sample of 13 students who had both binged and dieted, 7 had binged first, 3 had dieted first, and 3 binged and dieted for the first time at the same age. Future research should continue to investigate the relationships of the variables related to obesit y. Future directions might include a larger sample size and a modified sample selection pr ocess. Action research should continue in the areas of obesity prevention and interv ention, and student serv ices personnel should promote healthy lifestyle choices and a r ecognition of obesity as a socio-cultural problem.
1 Chapter I: Overview According to recent estimates by the Cent er for Disease Control (CDC), at least 15% of all children and adolescents in the United States, and over 30% of adults, are obese (National Center for Health Statistics, 2002). The rapidly escalating problem of obesity receives almost daily attention in th e news media, and many researchers continue to investigate its nature as well as the effi cacy of prevention and intervention techniques. The traditional view of obesity as an individu al problem is giving way to the perspective that obesity is a socio-cultura l problem and that responsibilit y for treatment therefore lies with society as a whole and not only with the individual. In order to effectively prevent and treat obesity, we must appreciate th e relationships among the constellation of variables that interact with obesity. The Epidemic of Obesity The CDC and most research studies defi ne obesity using body mass index (BMI), with obese individuals having a BMI of at least 30 or at least at the 95th percentile for their age and gender. BMI is an indirect m easurement of adiposity or body fatness and is calculated from an individual s body weight in kilograms divided by height in meters squared (kg/m2). Research indicates that, for the same BMI, the percentage of actual body fat tends to be higher for Asian-Ameri cans and lower for African-Americans when compared to Caucasians (Prentice & Jebb, 2001). Contrary to the common misperception that obesity is an eatin g disorder in the mental health realm of diseases, obesity is more accurately de scribed as a medical
2 condition that affects progressi vely more and more of our countrys adults and youth. Between 1980 and 2000, the prevalence of obesity doubled for children ages six to 11 and tripled for adolescents ag es 12 -19 (National Center fo r Health Statistics, 2002). These rapidly escalating prevalence rates have led to the common perception today in the United States of obesity as an epidemic. Obesity occurs at equal rates overall in adolescent boys and girls without respect to race, but prevalence varies between boys and girls of each race. Adolescent White males and females have similar rates of obes ity (12.8% vs. 12.4%), Black females have a higher obesity rate than Black males ( 26.6% vs. 20.7%), and Mexican males have a higher rate than Mexican females (27.5% vs. 19.4%). Research also indicates that obesity prevalence is higher in low socioeconomic status populat ions (Goodman, 1999; McMurray et al., 2000), and in children who receive special education services (Tershakovec, Weller, & Gallagher, 1994). Obesity is associated with many negativ e and dangerous outcomes, both medical and psychosocial in nature. The CDC recentl y identified obesity as the second leading preventable cause of death in the United States, second only to smoking, and according to the same study obesity caused 400,000 d eaths in 2000 (Mokdad, Marks, Stroup, & Gerberding, 2004). Obesity has been linked in adults to a host of medical problems including cardiovascular dis ease, cancer, coronary heart di sease, hypertension, diabetes, gallstones, osteoarthritis, and pr ostate enlargement. Interestin gly, up to one-half of adults with extreme obesity devel oped the condition in childhood or adolescence (Price, 2002; Steinbeck, 2001), and the more obese a child or adolescent is, the more likely he or she will be obese as an adult (Price, 2002). Obese youth suffer significant negative
3 psychosocial outcomes as well, including poor quality of life (Schwimmer, Burwinkle, & Varni, 2003), stigmatization (Latner & St unkard, 2003), and stereotypes of obese children and adolescents as mean, ugly, laz y, stupid, selfish, unhappy, socially isolated, subject to teasing, and dishonest (Schwart z & Puhl, 2003). Thus, the propensity of negative outcomes associated with obes ity clearly outlines the importance of understanding how to prevent and treat it effectively. Risk Factors of Adolescent Obesity Risk factors for obesity, or conditions th at increase the like lihood of obesity, fall into one of two categories: fixed and variab le. Fixed risk factor s are not manipulable and therefore are not typically targeted for di rect intervention. They may, however, point to particular demographically-defined groups of individuals that would benefit from prevention efforts. Research points to gene tics, heritability, gender, race, age of the child, and socio-economic status as fixed risk factors. Alternativel y, we can control and manipulate variable risk factors, and th ey present an excellent opportunity for intervention and prevention. Variable risk fact ors for obesity discusse d in the literature include diet and physical activ ity, prenatal malnutrition, caloric intake during infancy, cultural trends, and psychologica l factors including infant temperament, self-esteem, anxiety, depression, and body image. Genetics and heritability re present the most prominent fixed risk factors for obesity. Genetic research, lead ing to the discovery of speci fic genes and gene sequences related to obesity, has yielded evidence of some single-gene ob esities. More progress is expected in the pursuit of this line of re search (Price, 2002). In addition, heritability research over the last three decades, involvi ng dozens of twin and adoption studies, has
4 firmly established the genetic role in obesity. Research consistently points to biological factors as accounting for up to two-thirds of the variance in obesity and weight (Grilo & Pogue-Guile, 1991; Price, 2002). It also appear s that the influence of biological factors increases with the severity of obesity, and the risk of obesity if a family member is obese is much higher for extreme obesity (i.e., BMI ov er 40) than for moderate obesity (Price & Lee, 2001). Other research-based, non-manipulable risk factors for obesity include gender and race. As stated in the earlier discussion on th e prevalence of adolescent obesity, rates of obesity are similar for adolescent White males and females, adolescent obesity is higher for Black females than for Black males, and higher for Mexican males than for Mexican females. With regard to race, data also clearly indicate that obesity rates are higher for non-European populations such as African -American, Hispanic-American, and NativeAmericans. Some researchers (Price, 2002) pr opose a genetic thrifty gene theory that applies to groups such as the Pima Indians in Arizona and Mexico, whose rates of obesity and diabetes are drastically lower in the non-Westernized lifestyle and environment. According to this theory, these groups adap ted genetically in a manner that maximized the efficient use of energy, but in modern Westernized culture this metabolic tendency has become a liability. Genetic data do not ye t exist to support or refute this theory, however genetic research relate d to obesity is largely still in its early stages and is expected to progress (Price, 2002). Two final risk factors for adolescent obes ity that could be considered fixed in nature include the age of the adolescent and his or her socioeconomic status. Research indicates that the risk of obe sity is greatest at three point s before adulthood: early
5 infancy, prepuberty, and adolescence (Ste nbeck, 2001). Research is somewhat inconclusive on the role of socio-economic status in a dolescent obesity, however much evidence suggests that for adult women low econo mic status is associated with obesity. In developing countries elsewhere in the wo rld, however, high socio-economic status is correlated with obesity in children, women, and men (B erkowitz & Stunkard, 2002). Perhaps more interesting than fixed risk factors for obesity are variable risk factors because of their potential for use in prevention and interven tion. Cultural trends of the last several decades, in cluding decreasing physical activ ity levels, increased dietary fat, more sedentary lifestyles, increased use of automobiles, television viewing, and more frequent dining out especially in fast food establishments, have been blamed by many researchers for the rise of obesity (Stu nkard, 2002). It is lik ely that a genetic predisposition to obesity (i.e., a fixed risk factor) may interact w ith the recent cultural tendency towards increased dietary fat and sedent ary lifestyle (i.e., a va riable risk factor), resulting in the negative outcome of obesity or overweight for thos e individuals (Price, 2002). This negative interaction may serve to explain the drastic increase in obesity in America over the last several decades. Support also can be found for other contro llable risk factor s including prenatal malnutrition, high caloric intake during infancy, and parental attempts to control food intake with their children, all of which reportedly increa se the likelihood of obesity (Berkowitz & Stunkard, 2002). Prenatal malnutri tion in the first half of the pregnancy has been found to predict obesity. Resear ch also provides evid ence that a vigorous feeding style during infancy predicts chil dhood obesity, as does cal oric intake during infancy (Berkowitz & Stunkard, 2002). Interestingly, dietar y intake in childhood has not
6 been found to predict later body fat levels. However, prompting of children by parents to clean their plates or eat mo re has been associated with weight in several studies (Berkowitz & Stunkard, 2002). Other Variables Correlated with Adolescent Obesity In addition to the risk factors for obesity of genetics and heritability, gender, race, age, socio-economic status, diet and physical activity, a review of the obesity literature reveals support for correlations between obe sity and binge-eating behavior, depressive symptoms, and special education status. Evidence has accumulated in the past decade related to the relationship of binge-eating disorder w ith obesity. The American Psychiatric Association define s binge-eating disorder in th e Diagnostic a nd Statistical Manual of Mental Disorders-Fourth Editio n-TR (American Psychiatric Association, 2000) in an appendix for proposed diagnostic cat egories. Similar to bulimia nervosa, individuals with binge-eating disorder engage in binge-e ating episodes but do not use purging or other inappropriate we ight loss methods. Estimates of obese adults that have binge-eating disorder range from 3% to 30% (Stunkard, 2002), and European research has found that approximately 36% of obese chil dren and adolescents also are binge-eaters (Decaluwe, Braet, & Fairburn, 2002). Resear ch has also supported a relationship between binge-eating and dieting (Kinzl et al., 1999; Ross & Ivis, 1999), although it is not established which behavior precedes the ot her. Alternatively, research has firmly established dieting as a risk factor fo r anorexia nervosa and bulimia nervosa. The amount of research on the relations hip between depression and obesity over the past decades has ebbed and flowed, with a recent surge of interest. The urgency of understanding the nature of th is relationship is emphasized by estimates that up to 8.5%
7 of adolescents in the United States suffer from depression (National Institute of Mental Health, 2000), and 7% of depressed adoles cents may die by suicide as young adults (Weissman, Wolk, & Goldstein, 1999). Research from the 1960s and 1970s investigated depression and ot her psychological and behavior al constructs that were thought to underlie obesity. Later empirica l studies in the 1980s and 1990s focused on obesity as a genetically mediated medical condition, and it was accepte d that little or no relationship existed between obesity and de pression (Stunkard, Faith, & Allison, 2003). Recently, however, research has focused once again on the possibility of a direct relationship between the two conditions (G oodman & Whitaker, 2002; Stunkard et al., 2003). Little research has investigated the relationship between obesity and special education status. One study found that obese children were overrepresented in special education and remedial classes (Tershakovec, Weller, & Gallagher, 1994). However, this overrepresentation could be explained at least in part by the historic al overrepresentation of minorities in special educat ion (National Academy Press, 2002), along with the higher prevalence rates of obesity in minority child ren and adolescents (National Center for Health Statistics, 2002). Research Objectives In light of the serious and escalating nature of obesity as a public health concern, and the importance of identifying preventi on and intervention opportunities, further research on the relationships between obesity and related variables is critical. Most obesity research conducted to date has focu sed on the relationships among one or two variables, and not on a larger view that considers the constellation of variables thought to
8 be associated with obesity. The present study will investigate th e problem of obesity using a more comprehensive framework. The path diagram illustrated in Figure 1 portrays the relationships betw een obesity and correlated va riables as currently supported in the literature. The model was constructe d for the purpose of the present investigation, and draws upon research on the various inte rrelationships between adolescent obesity, binge-eating behavior, depressive sympto ms, dieting, physical exercise, gender and ethnicity, socio-economic status, and special education status. The relationships depicted reflect those receiving the str ongest empirical support, with the exception of genetics and heritability. The strong biol ogical contribution to weight and obesity has received consistent support through two decades of research, and cons equently the present model and study focus on other proposed relati onships with obesity and weight. Figure 1. Proposed Path Diagram, Co rrelates of Weight in Adolescents Depressive Symptoms Binge Eating Symptoms Physical Activity Socio-Economic Status (Yes/No) Obesity/Weight Special Education Status Dieting Gender/Ethnicity CA HI AS NA O AA
9 Further research into the relationship be tween obesity and depr ession is warranted given the lack of studies on this topic until very recently. If the rate of obesity continues to rise in adolescents, and depr ession is associated with obesity as a cause, effect, or correlate of obesity, it is critic al for mental health and medi cal professionals to have an accurate understanding of the nature of this relationship. In particular, research will inform intervention efforts for that subgroup of obese individuals who also have symptoms of binge-eating, and are most at-r isk for depression and other psychopathology compared to non-bingeing obese yout h (Wilson, Heffernan, & Black, 1996). Another shortcoming of existing obesity rese arch is that much of the research has been conducted in Europe, limiting its genera lizability to adolescents in America. Furthermore, much of the obesi ty literature neglects to differentiate between results for adolescents and adults. Gi ven the known trajectory of obesity from childhood and adolescence into adulthood, it is necessary to id entify the best potential areas for prevention and intervention, at th e earliest age possibl e. In order to achieve this goal, more research specifically targeting children and adolescents is necessary. In addition, many of the existing studies on obesity and eating disorders included only females in their samples, further limiting the ability to generalize results to adolescents in America. Especially given the similar rates of obesity for males and females in the United States, more obesity research is needed that is inclusive of males. The present study was designed to addre ss some of the shortcomings in the existing obesity research by answering th e following three rese arch questions. (1) To what degree do the data support the propos ed model of the correlates of weight in adolescents?
10 (2) For the populations of obese and non-obese adolescents, what is the level of binge-eating symptoms, and what is the strength of the re lationship between binge-eating symptoms and depressive symptoms? (3) For the population of adolescents who e ngage in both binge-eating and dieting behavior, what is the order of preceden ce of binge-eating and dieting (i.e., which occurs first, binge-eating or dieting)? It was expected that the data collected in this research study would support the proposed path model. Based on existing researc h, it was also expected that similar levels of binge eating symptoms would be identifie d for obese and non-obese adolescents, and that the level of binge-eating symptoms w ould positively correlate with the level of depressive symptoms. Finally, it was expected that, for adolescents who have a history of both bingeing and dieting behavior, a higher percentage of subjec ts would report that dieting preceded bingeing. Information gained from this study helps to inform critically needed efforts towards prevention and interven tion for obesity by shedding more light on the nature of the relationships among obes ity, depression, binge-ea ting, physical activity, dieting, socio-economic status, special e ducation status, and gender and race.
11 Chapter II: Literature Review Obesity in both children and adults curre ntly poses a major public health threat in the United States. Recently identified by the Centers for Disease Control and Prevention as the number two preventabl e cause of death, obesity caused 400,000 deaths in 2000 and is solidly on track to overtake smoking as th e number one preventable killer in the U.S. (Mokdad, Marks, Stroup, & Gerberding, 2004). The Centers for Medicare and Medicaid Services has revised its Medicare obe sity coverage policy to eliminate language stating that obesity is not an illness, and intends to extend co verage for obesity treatments that are empirically based and shown to be eff ective in clinical trials (U.S. Department of Health and Human Services, 2004). The followi ng review of the lite rature examines the available research on variables related to obesity. First, obesity research will be summarized with respect to its definition, pr evalence in adolescents, outcomes, and risk factors including ethnicity, gende r, socioeconomic status, and special education status. A discussion will follow of several constructs th ought to be related to obesity, including binge-eating disorder, dieting, depression, and ph ysical activity. These constructs will be summarized within the major s ections of binge-eating disord er and depression. Together all of the research findings will be tied to a proposed path model for investigating and hypothesizing the relationships among the correlates of obesity. Obesity in Adolescents Obesity in both adults and adolescents represents an increasing problem in the United States and elsewhere in th e world, and it is often referred to in the literature as an
12 epidemic (Ebbeling, Pawlak, & Ludwig, 2002). Obesity in children and adolescents poses a serious health risk. The fields of public health, medicine, and pediatric psychology have given increased attention to the problem of obesity in children and adolescents, as has the media. Given the curr ent prominence of obesity as a health care concern throughout much of the developed and developing world, a firm understanding of the factors that co rrelate with and predic t obesity carries obvi ous importance. The following discussion explores the definition a nd prevalence of obesity in adolescents, the medical outcomes associated with obes ity, and the risk factors for obesity. Definition of Obesity In defining obesity, it is us eful to begin with what obe sity is not. Contrary to common belief, obesity is not an eating disord er, but rather a medi cal condition. In the Diagnostic and Statistical Manua l of Mental Disorders, Fourth Edition (DSM-IV-TR), the American Psychiatric Association defines diag nostic criteria for anorexia nervosa and bulimia nervosa, but not for obesity, because th ere is insufficient evidence that obesity is associated with a psychological or beha vioral syndrome (American Psychiatric Association, 2000). The DSM-IV-TR does, howev er, define research criteria for bingeeating disorder, a diagnosis proposed for furt her study that is quite prevalent in obese individuals. Obesity is usually defined in terms of body mass index (BMI), an index of weight and height equal to body weight in kilograms divided by height in meters squared (BMI = kg/m2). BMI is an indirect measurement of ad iposity or body fatness, and is intended as a screening tool for underweight or overwe ight. Epidemiological studies conducted around the world have used various definitions of obesity, including BMI at or above 95th
13 percentile, weight-for-height ratio greater th an two standard deviations from the median, and an age-adjusted BMI cutoff of 30 or greater (Ebbeling, Pawlak, & Ludwig, 2002). Average BMI varies by gender and age thr oughout childhood, and the Center for Disease Control (CDC) recommends the use of its BM I-for-age growth charts for children and adolescents age two and older (Center for Disease Control, n.d.). Figure 2 shows the Figure 2. BMI growth chart for boys aged 220. From Using the BMI-for-age Growth Charts, Center for Disease Control ( n.d.). Retrieved April 19, 2003, from http://www.cdc.gov/nccdphp/dnpa/growthcharts /training/modules/module1/text/intro.htm BMI-for-age growth chart for boys, with BMI decreasing until age 4 and then increasing each year thereafter. Table 1 depicts the cutoff points recommended by the CDC to identify children and adolescents who are a t risk of overweight and overweight. The 85th and 95th percentile cutoff points correspond to CDC criteria classifying adults as overweight (i.e., adult BMI of 25) and obese (i.e., adult BMI of 30), respectively (Center for Disease Control, n.d.). Similarly, the 5th percentile cutoff point is the
14 criterion for identifying children and adolescents who are underweight, and it corresponds to an adult BMI of 18 or lowe r (Field, Barnoya, & Colditz, 2002). Table 1 BMI Cutoff Points for Child, Adolescen t, Adult Weight Classifications Child/Adolescent Classification Percentile Adult BMI Classification Overweight >= 95th >=30 Obese At Risk of Overweight >= 85th >= 25 Overweight Normal 6th 84th 19 24 Normal Weight Underweight <= 5th <= 18 Underweight Note. Cutoff points recommended by Cent ers for Disease Control. Adapted from Using the BMI-for-age Growth Charts, Center for Di sease Control (n.d.). Retrieved April 19, 2003, from http://www.cdc.gov/nccdphp/dnpa/growthcharts /training/modules/module1/text/intro.htm According to the CDC, obesity cutoff cr iteria for children and adolescents are based upon the 2000 CDC BMI-for-age-growth char ts which are sex-specific. The adult BMI cutoff of 30 for obesity emerged in 1997 as an international standard when the World Health Organization (WHO) publis hed terminology and classifications for overweight and obesity. Thes e criteria have since been embraced by the CDC and the National Heart, Lung, and Blood Institute (NHLBI), among other federal agencies. The CDC cites several empirically-based advantages to using its BMI-for-age charts as a screening tool for obesity, overweight, and underweight in children and adolescents (Center for Disease Control, n.d.). First, the charts re present a previously
15 unavailable reference source for adolescents. The measure can be used for an individual continuously into adulthood because it is consis tent with the BMI index used for adults. BMI-for-age also correlates well with hea lth risks for cardiovascul ar disease including high blood pressure and elevated insulin. Finally, web-based CDC training materials state that BMI correlates significantly with total and subcutaneous body fat. While BMI serves as the commonly accepted measurement tool for obesity and other weight classifications in the United St ates and internationally, some researchers argue that more direct measures should be employed to determine levels of fat in the body (Prentice & Jebb, 2001). Direct methods for measuring body fat include waist-hip ratio, waist circumference, skinfold thic kness measurements (e.g., tricep skinfold thickness), underwater weighing, dual energy x-ray absorptiometry (DMX), and other complex laboratory-based methods. These m easures are necessarily more accurate and valid than the proxy method of BMI because they directly measure the amount of body fat. In some cases, BMI can even be inaccurate in its representation of body fat percentage. For example, the aging process fo r adults entails a continuous increase in the ratio between fat and lean body mass, even fo r individuals whose BMI remains the same throughout their lifespan. In addition, racial differences have been documented relating to the accuracy of BMI. Asian-American persons have a high er actual body fat percentage than Caucasians, and African-A mericans have a lower percentage than Caucasians, for the same BMI. Similarly, the same level of body fat for Caucasians translates to a higher BMI for African-Ame ricans and a lower BMI for Chinese, Thai, and Indonesians. Prentice and Jebb (2001) also point out the systematic error that occurs when utilizing BMI for athletes and othe rs who pursue rigorous physical training and
16 buildup of muscle mass, which weighs more th an body fat. Consequently, the high level of muscle mass leads to an overestimation of body fat in these individuals. Other researchers counter, however, that highly fit individuals ra rely have enough muscle mass to be misclassified as obese (Field, Bar noya, & Colditz, 2002). Thus, BMI provides an established, well-accepted, and simple but not always accurate measurement tool for body fat percentage. Prevalence of Adolescent Obesity At least 15% of children a nd adolescents in the U.S. ar e obese, according to data from the most recent National Health a nd Nutrition Examination Survey (NHANES), which reports data for 1999-2000 (National Cent er for Health Statistics, 2002). The 15% prevalence rate applies to children ranging fr om six to 11 years, and to adolescents ranging from 12 19 years. These data repr esent the latest of several NHANES studies conducted since 1963, and were gathered by the CDC through household interviews and physical examinations of each participant. Participant data for all NHANES studies were stratified by sex, age, race, and Hisp anic origin. The NHANES 1999-2000 study included 3,601 participants, a smaller sample size than either of the two previous studies from 1994 and 1980, which had 14,468 and 11,207 participants respectively. Consequently, standard errors of estimates cited for the latest study were somewhat higher than for the 1994 study, ranging from 1.2 to 3.3 percentage points for the 2000 data vs. standard errors of 0.9 to 2.5 percen tage points for the 1994 data. Children and adolescents were classified as obese if their BMI was greater than or equal to the 95th percentile for their age and gender. Table 2 shows the alarming increase in obesity prevalence between 1980 and 2000, during whic h obesity rates doubled for children age
17 six to 11 years, and tripled for adolescents. These prevalence rates for children and adolescents trail those for adult obesity. Th e adult obesity rate according to the same 1999-2000 NHANES study was 31%, up from 23% in 1994 and 15% in 1980. These escalating rates of obesity cast light on th e common reference to the condition as an epidemic in U.S. society. Table 2 Prevalence of Obesity in U.S. Ch ildren, Adolescents, and Adults Age in Years NHANES II 1976-1980 n = 11,207 NHANES III 1988-1994 n = 14,468 NHANES 1999-2000 n = 3,601 6 11a 7 11 15 12 19a 5 11 15 >20b 15 23 31 Note. Adapted from National Center for He alth Statistics, 2002. Retrieved April 19, 2003 from http://www.cdc.gov/nchs/produc ts/pubs/pubd/hestats/overwght99.htm and http://www.cdc.gov/nchs/products/p ubs/pubd/hestats/obese/obse99t2.htm a. Obesity defined for child ren and adolescents: BMI >= 95th percentile for gender, age b. Obesity defined for adults: BMI >= 30 The overall obesity prevalence for adoles cents closely represents the rates for boys and girls in the NHANES 1999-2000 survey (N ational Center for Health Statistics, 2002). Interestingly, boys and girls 12 19 years of age both have a 15.5 % prevalence rate. However, the prevalence of adolescen t obesity varies cons iderably among racial groups including non-Hispanic Whites, nonHispanic Blacks, and Mexicans. As
18 illustrated in Table 3, White non-Hispanic boys and girls have similar prevalence rates of 12.8 and 12.4 %, respectively. However, obesity is more common in non-Hispanic Black females (26.6%) than males ( 20.7 %), and it is more preval ent in Mexican males (27.5%) than females (19.4 %). Adolescent obesity is again defined as having a BMI at or above the 95th percentile for age and gender, accordi ng to the CDCs 2000 BMI-for-age growth charts. Table 3 Prevalence of Adolescent Obes ity by Gender and Race, 1999-2000 Total Non-Hispanic White Non-Hispanic Black Mexican Males 12 19 years 15.5% 12.8 % 20.7 % 27.5 % Females 12 19 years 15.5% 12.4 % 26.6 % 19.4 % Note. Adapted from National Center for Hea lth Statistics, 2002. Retrieved April 19, 2003 from http://www.cdc.gov/nchs/produ cts/pubs/pubd/hestats/overwght99.htm Recent studies on the relationship between socioeconomic status (SES) and weight status for adolescents indicate that overweight and obesity ar e inversely related to SES. A North Carolina school-based study with a sample of 2,389 a dolescents classified students as low, moderate, and high SES base d on parental education level (McMurray et al., 2000). Using a criterion of BMI over the 85th percentile, the low SES group had the highest rate of overweight or obesity (41%), as compared to the moderate SES rate (35%), and the high SES rate (24%). Si milar findings were reported by a large study
19 based on a sample of 15,483 adolescents from the National Longitudinal Study of Adolescent Health (Goodman, 1999). This st udy defined SES in terms of parental income, parental education, and resident pare nt occupation, and defined obesity as a BMI greater than the 95th percentile for age and sex. Obesity was linearl y related to parental occupation and was also significan tly related to household inco me and parent education. Little research exists on the relations hip between obesity and special education status, but there is some evidence that obe se children are overrepresented in special education settings. The historic al overrepresentation of mino rities in special education (National Academy Press, 2002), combined with the higher rates of obesity in minority children and adolescents, would seem to pred ict an overrepresentation of obese children and adolescents in special e ducation. In a study of blac k, inner-city elementary school students, obese students were twi ce as likely as non-obese studen ts to be placed in special education or remedial settings (Tersha kovec, Weller, & Gallagher, 1994). Given the homogeneous sample, this study inherently c ontrolled for race and socio-economic status, so that educational setting could not be related to those factor s. The results of this study should be interpreted with caution due to the facts that these data are 10 years old and data were only collected in elementary sc hools. In addition, the authors did not draw conclusions related to causality, and th e study was cross-sectional rather than longitudinal. Interestingly, another study has sugg ested a predictive link between learning problems and obesity. Lissau and Sorensen (1993) conducted a prospective longitudinal study whose participants consisted of a randomly selected sample of 1258 third graders in Copenhagen, Denmark, or 25% of the 1974 th ird grade population in the Copenhagen
20 municipality. The authors concluded that special education and learning difficulties in the third grade predicted obesity at age 2021. The study controlled for socio-economic status, BMI in childhood, and gender in their data analysis. Base line occurred in 1974 when subjects were in the 3rd grade, and follow-up was at age 20 or 21. Unfortunately these results are also somewhat dated, and ma y or may not generalize to Americans. In addition, the authors included no information re garding race or ethnic ity of participants, preventing any analysis or control of race as a factor. Outcomes of Adolescent Obesity Obesity poses a serious health risk to th e U.S. and other societies. Obesity in adults is associated with increased risk of mortality from all causes, especially cardiovascular disease. Obese adults are mo re likely than normal weight adults to develop a myriad of health problems includ ing coronary heart di sease, hypertension, Type 2 diabetes mellitus, numerous types of cancer (e.g., breast, endometrial, gastric, colon, esophageal, stomach), gallstones, osteoart hritis of the hip and knee, and prostate enlargement (Field, Barnoya, & Colditz, 2002). Given the risi ng prevalence of obesity in our country, it is also reasonabl e to expect the rate s of all of these medical conditions to rise in future decades. Since obesity is in most cases a preventable disease, health professionals should place a significant fo cus on successful efforts towards obesity prevention. Furthermore, research suggests that betw een one-third and one-half of extremely obese adults first develop obe sity in childhood or adolesce nce (Price, 2002; Steinbeck, 2001), and the more extremely obese an indivi dual is in childhood or adolescence, the more likely he or she will be obese as an adult (Price, 2002). Even if a child or
21 adolescent does not become an obese adult, he or she risks negative medical outcomes as a child or adolescent. For example, rising obe sity rates are likely related to the recent increase in Type 2 diabetes mellitus in early adolescence (Steinbeck, 2001). Poor quality of life can also be viewed as a negative outcome of severe obesity for children and adolescents. Although research ex ists on the health-related quality of life for adults with obesity, very little research has been conducted on the quality of life for children and adolescents with obesity. A recent hospital-based study of children and adolescents ages 5 to 18 years compared the health-related quality of life of 106 severely obese subjects with that of a group diagnosed with cancer and another group of healthy controls (Schwimmer, Burwinkle, & Varni, 2003). Using the PedsQL 4.0 (Varni, Seid, & Kurtin, 2001), a 23-item pediatric quality of life inventory, the rese archers reported that the scores of obese subjects indicated an impaired health-related quality of life in all the domains measured: physical, psychosocial, emotional, social, a nd school functioning. The authors defined impairment as a score fa lling at least one standard deviation below the mean score for healthy participants. Su rprisingly, the degree of impairment in all areas was similar to that reported by childre n and adolescents diagnosed with cancer and receiving chemotherapy. Obese children and adolescents face frequent stigmatization by both peers and adults in society. There is evidence that this bias against obese children has increased significantly since the 1960s. Latner and Stunkard ( 2003) recently replicated a study conducted in the 1960s in which 458 5thand 6thgraders ranked six drawings of one healthy child, one obese child, and four chil dren with disabilities such as facial disfigurement, a missing hand, and a child in a wheelchair. Participants indicated how
22 much they liked each child by ranking the drawings from one to six with one representing the child they liked the most. Results indicated that the obese child was liked the least, as in the 1960s study. The healthy child had th e highest mean rank of 1.97, the obese child had a mean rank of 4.97, and the children with disabilities had m ean ranks ranging from 3.09 to 3.86. In addition, the obese child was liked 40.8% less in the current study than in the original study, and the authors conclude d that stigmatization of obese children may have increased significantly since that time. According to attit udinal research, obese children and adolescents are associated with negative characteristics such as mean, ugly, lazy, stupid, selfish, unhappy, soci ally isolated, subject to teasing, and dishonest, and biased attitudes against obese peers appare ntly increases with age (Schwartz & Puhl, 2003). In summary, given the wide array of medical and psychosocial outcomes, prevention of adolescent obesity is critical. Risk Factors of Adolescent Obesity The success of prevention programs de pends to a great extent on the understanding of factors that increase the like lihood of obesity, and also those that reduce its likelihood. Factors that can be influenced or ma nipulated provide the best opportunities for prevention and intervention, su ch as physical exercise and prenatal nutrition. On the other hand, some factors su ch as genetics and heritability are more fixed in nature and do not offer much in the way of interventi on opportunities. Fixed factors thought to influence the likelihood of adolescent obes ity include genetics, gender, race, and the age of the child. Fixed risk factors will be discussed first, followed by controllable factors.
23 An increasing amount of research has been conducted in the last two decades on the genetic contribution to obesity. Re search surrounding the Human Genome Project has generated a prolific amount of data on DNA sequence and gene function as well as innovations in research technol ogy. Scientists have provided evidence of se veral singlegene obesities, or obesities which have been linked to a single gene. While these data and technology are impressive, little is still known regarding the specific genetic patterns that contribute to most common obesities. However, the field of genetics continues to operate in a period of rapid grow th. Thus, the state of its kno wledge base is expected to climb to new levels in the next several years, and will most likely illuminate the link between genetics and ob esity (Price, 2002). Genetic research can be di stinguished from studies on heritability of obesity. Some of the most interesting st udies that appear in the obes ity literature have strived to distinguish the influence of i nherited biological traits from that of environmental factors shared by family members. Many twin and adoption studies over the last two decades have demonstrated a strong influence of biological factors on obe sity. One commonly cited seminal review (Grilo & Pogue-Geile, 1991) provide s a good introduction to the importance of heritability in the occurre nce of obesity. Th e authors analyzed approximately 45 studies conducted between 1970 and 1990, with samp le sizes totaling approximately 28,000 pairs, or 56,000 subjects. The studies compared various pairs of related individuals, including bi ological siblings reared toge ther, adoptive siblings reared together, and twins reared togeth er versus reared apart. Tw in studies produced data for both dizygotic twins (i.e., fraternal twins or iginating from two different eggs), and the more rare monozygotic twins (i.e ., identical twins or iginating from the same egg). Also
24 included in their review were studies on corre lations of weight for spouses, and adoptive parents compared to their adoptees. Grilo and Pogue-Guile (1991) reported strong evidence for biological influence on obesity based on the collectiv e results of the studies revi ewed. The authors reported that non-shared experiences among family members were a much stronger influence on weight, fatness, and obesity than were sh ared experiences among family members. According to Grilo and Pogue-Guile, exam ples of non-shared experiences include perinatal insult, peer relationships, or differe nces in parenting. Examples of shared experiences might include similar child-re aring practices and li ving conditions. For example, average correlations for weight in monozygotic twins reared apart, .72, were similar to that of monozygotic twins reared together, .80. The average correlations for BMI were also similar, .62 for twins reared apart and .74 for twins reared together. The authors also reported that adop tive siblings were uncorrelated in weight or fatness, and adopted children did not resemble their adoptive parents in wei ght or fatness. Also, their review found that adopted child ren and biological children resembled their biological parents in weight to the same degree. Grilo and Pogue-Guile reported several methodological weaknesses in their metaanalysis, the most significant being the wide variation in subject age within many of the studies analyzed ranging from pairs of infants to elderly a dults. Some of the studies combined results for subjects of different ages making it difficult to identify patterns in weight occurring at different ages. In a ddition, the study makes no mention of race or ethnicity. These limitations notwithstandi ng, the authors found ev idence of only minor influence of shared environmental experiences on weight. They concluded that unshared
25 experiences account for approximately 20% of th e variability in weight, and that genetics accounts for most of the variabil ity in weight and obesity. A more recent review of the genetics a nd heritability research on obesity draws conclusions quite similar to those outlined in the seminal analysis by Grilo and PogueGuile (1991), one decade later. According to Price (2002), results fr om multiple twin and adoption studies suggest that up to two-thirds of the variance in obesity and weight in adults can be attributed to genes, and lit tle influence can be attributed to family environment. While the authors summary of the heritability literature concurs with earlier findings, it should be noted that the methods used to conduct this review of the literature were not well documented, and th e number of studies and their sample sizes were not clear. Price and Lee (2001) also conducted research on odds ratios that represent the risk of obesity if a family member is obese, as determined by the ratio between prevalence in the family and prevalence in the population. The odds ratio associated with extreme obesity, or BMI greater than 40, is much highe r than that for moderate obesity. Thus, it appears that the influence of genetics may incr ease with severity of the obesity. In sum, according to the literature the contribution of genes to overweight and obesity is clearly larger than any other factor, and while it obv iously cannot be manipulated, at least with todays technology, it also cannot be ignored as an ongoing to pic of research if one considers such possibilities as ge netic manipulation or gene therapy. Other non-manipulable risk factors for obe sity possibly related to genetics are those of gender and race. Prevalence rates in adolescents do not s how a clear difference between girls and boys, as shown in Table 3, but a much greater disparity exists between
26 adult women and men. In addition, the data in Table 3 support ex isting research on the higher obesity prevalence rates in non-Eu ropean races including African-American, Hispanic-American, and Native-Americans. One theory advanced by Price (2002) and others is that these groups developed genetically in a way to allow them to survive hardship conditions by maximizing the effi cient use of energy, however in modern Westernized culture it has become a cause of obesity. Groups such as the Pima Indians in Arizona are thought to perhaps have devel oped a thrifty genotype that mediates the storage and expenditure of energy. No genetic evidence exists to suppo rt this theory as yet. Another unchangeable risk factor for obesity that is highly rele vant to adolescent obesity is the age of the child. Research has shown that the risk of obesity increases at three points in the life cy cle prior to adulthood: ea rly infancy, prepuberty, and adolescence (Steinbeck, 2001). While the age of the child obviously cannot be altered, this knowledge provides good insigh t for intervention and preven tion in terms of the best times to implement a program. Socioeconomic status (SES), another poten tial fixed risk factor for obesity, has been found in some studies to correlate with obesity in children and adolescents, but the research is not consistent in this area. Berkowitz and Stunkard ( 2002) reported that 40% of the studies they reviewed found an inve rse relationship between obesity and SES for children and adolescents, 40% showed no relationship, and 25% showed a positive relationship. The relationshi p between obesity and SES for adult women is consistently inverse, however. Interesti ngly, studies in developing count ries showed a strong positive correlation between obesity and SES for children, as well as for women and men. Little
27 research exists on the eating behaviors of children of low SES in the U.S., but some researchers have theorized th at diet may play an importa nt role in weight status (McMurray et al., 2000). In a study on the influence of physical activity, socioeconomic status, and ethnicity on the weight status of 2389 adolescents in Nort h Carolina (McMurray et al., 2000), subjects in the low SES status group ha d a higher rate of overweight and obesity (41%) than participants in the moderate (35%) and high (24%) SES groups. The authors theorized that this higher rate of obesity wa s not related to physical activity, because individuals in the low SES group reported highe r levels of physical activity as well as high-intensity activit ies. However, members of this group reported twice as much time spent on television viewing and video game pl ay, suggesting that se dentary behavior may play a key role in weight status. Despite the body of research pointing to genetics as accounting for most of the variability in weight, it is clear that some factor or combination of factors in the environment must be responsible for the rapid increase in obesity rates in recent decades. One common theory is that western culture has brought changes in lifestyle, especially relating to diet and exercise. As some rese archers propose (Price, 2002) it is likely that the increase in dietary fat and the trend towards more sedentary lifestyles interact in a negative way for individuals who have a ge netic predisposition towards overweight and obesity. Several researchers have found th at lower levels of physical activity are associated with obesity in children (B erkowitz & Stunkard, 2002; Steinbeck, 2001). Physical activity also tends to decrease during adolescence in all children, and is influenced by parental physical activity leve ls. However, more research is called for
28 regarding physical activity as a risk factor for obesity, because no st udies exist to verify that physical activity levels have decreased in the same time frame in which obesity has dramatically increased (Steinbeck, 2001). Other controllable risk factors for adolescen t obesity appear in the literature as well. Prenatal malnutrition has been found to increase the likelihood of obesity, if it occurs in the first trimester or first half of the pregnancy (Berkowitz & Stunkard, 2002). While breast versus bottle feeding does not a ppear to influence development of obesity, caloric intake during infancy does predict early childhood adiposity. However, childhood dietary intake does not reliabl y predict later leve ls of body fat. Evidence also suggests that a vigorous feeding style during infancy (i .e., rapid, long sucks with shorter intervals between sucks) predicts obesity. Several studies have also s uggested a relationship between weight and parental attempts to cont rol food intake with their children, such as prompting of children to eat more or clean their plates (Berkowitz & Stunkard, 2002). In their analysis of risk factors for th e development of early obesity, Berkowitz and Stunkard (2002) also refer to cultural trends in recent decades such as the increasing tendency of Americans to eat meals outside th e home, dine in fast food establishments, a decline in physical education for students in the U.S., and an increased use of automobiles. Television viewing represents another controllable risk factor for obesity, due to its effects of reducing physical activ ity and increasing consumption of calories while watching TV or afterwards resulting fr om food advertising (Birch & Fisher, 1998). One final area of research on risk factor s for obesity is that of psychological factors, which could arguably be viewed as eith er fixed or variable f actors. Research is not conclusive on whether infant temperament is related to later body fatness, or whether
29 a relationship exists between obesity and se lf-esteem, depression, or anxiety (Berkowitz & Stunkard, 2002). However, some eviden ce suggests that obese adolescents and children have greater body image di ssatisfaction than non-obese peers. Binge Eating Disorder in Adolescents Eating disorders can be defined as sev ere disturbances in eating behavior, maladaptive and unhealthy efforts to contro l body weight, and abnormal attitudes about body weight and shape (Wilson, Heffernan, & Black, 1996, p. 541). In earlier decades, researchers often viewed obesity as an eating disorder caused by the same eating behaviors in all obese individua ls (Stunkard, 2002). However, more recent research has shed light on the wide variation in eating behaviors in the obese population. Increasingly, binge-eating disorder has emerged as a correl ate and contributor to obesity. Binge-eating disorder will now be discussed with regard to its definition and preval ence in adolescents, risk factors, and the relationship of binge-eating with di eting and obesity. Definition of Binge-Eating Disorder The DSM-IV-TR (American Psychiatri c Association, 2000) defines diagnostic criteria for three eating disorders: anorexia nervosa, bulimia nervosa, and eating disorder not otherwise specified. The la tter criteria are intended for disorders that do not satisfy the requirements for anorexia or bulimia, and the manual includes binge-eating disorder as an example. In 1994, the American Psyc hiatric Association fo rmally defined bingeeating disorder in Appendix B of the DS M-IV as a proposal for a new diagnostic category (American Psychiatric Association, 19 94). Those criteria remain in the DSMIV-TR, as shown in Table 4.
30 Table 4 DSM-IV-TR Research Criteria for Binge-Eating Disorder A. Recurrent episodes of binge eating. An ep isode of binge eating is characterized by both of the following: (1) eating, in a discrete period of time (e.g., within any 2-hour period), an amount of food that is definitely larg er than most people would eat in a similar period of time under similar circumstances (2) a sense of lack of control over eating dur ing the episode (e.g., a feeling that one cannot stop eating or contro l what or how much one is eating) B. The binge-eating episodes are associated with three (or more) of the following: (1) eating much more rapidly than normal (2) eating until feeling uncomfortably full (3) eating large amounts of food when not fee ling physically hungry (4) eating alone because of being embarrassed by how much one is eating (5) feeling disgusted with oneself depressed, or very guilty after overeating C. Marked distress regarding binge eating is present. D. The binge eating occurs, on average, at least 2 days a week for 6 months. E. The binge eating is not associated with the regular use of inappropriate compensatory behaviors (e.g., purging, fasting, excessive ex ercise) and does not occur exclusively during the course of Anorexia Nervosa or Bulimia Nervosa. Note: From American Psychiatric Association (2000, p. 787) Binge-eating disorder and bulimia nervosa share some of the same diagnostic criteria. Episodes of binge eating occur in both binge-eating disorder and bulimia nervosa. However, binge eating frequency is defined differently for the two disorders. Bulimia nervosa requires a frequency of two or more binge eating episodes a week for three months, while binge-eating disorder requi res episodes to occur on two or more days per week for six months. In addition, bulimia requires the regular us e (i.e., at least two times a week) of compensatory behaviors such as self-induced vomiting, misuse of
31 laxatives, diuretics, or enemas. The nonpurgi ng type of bulimia specifies the regular use of fasting or excessive exer cise, but not purging behaviors such as vomiting, laxatives, diuretics, or enemas. In order to meet th e criteria for binge-eating disorder, a person could not regularly engage in any of these compensatory behaviors, although they may sometimes engage in them. Researchers on bi nge-eating disorder ha ve varied in their criteria for regular compensatory behavi ors, ranging from epis odes occurring twice a week to not at all. Prevalence of Binge-Eating Disorder Many of the prevalence figures found in th e research on bingeeating disorder do not provide separate rates for adolescents. Some research suggests that onset for bingeeating disorder occurs in late adolescence or in the early 20s, that females are 1.5 times more likely to develop the disorder than male s, and the overall prevalence for adults in the general population ranges from 0.7%-4% (A merican Psychiatric Association, 2000). A limited number of studies have yielded pr evalence estimates among adolescents. One of the most recent, a school-based epid emiological study, found a 1% prevalence of binge-eating disorder in Norwegian 15-year olds (Rosenvinge, Borgen, & Borresen, 1999). It is also evident that research on binge eating in a dolescents often fails to use a strict definition of binge-eating disorder, so that prevalence rates reflect subclinical levels of binge eating. For example, in a Canadi an school-based study of 1031 girls and 888 boys, 18.9% of the girls and 17.5% of the boys reported engaging in bi nge eating at least once in the last year without compensatory weight loss behaviors (Ross & Ivis, 1999).
32 Risk Factors for Binge-Eating Disorder Similar to the findings of research on obe sity, evidence exists for the contribution of both fixed and variable f actors in binge-eating disorder Literature supports the influence of the fixed risk factor of genetics, as well as a myriad of variable risk factors. In the area of heritability research, a recen t large twin study involving 2163 female twins (Bulik, Sullivan, & Kendler, 2002) examined genetic factors, common environmental factors, and unique environmental factors. Us ing bivariate twin analysis, the researchers found a moderate heritability factor for binge eating, a strong herita bility link for obesity, and a moderate overlap in the genetic contri bution to obesity and bi nge-eating disorder. The study used a broad definition of binge-e ating, however, and it included only white women. The first controlled study to examine ri sk factors of bingeeating disorder, often cited in the recent eating diso rder literature, was conducted in England in 1998 (Fairburn et al.). This community-based case control study differen tiated personal and environmental factors that were reported to occur prior to th e onset of disordered eating in the lives of 52 females age 16-35 with binge-eating disorder 104 healthy control subjects, 102 with other psychi atric disorders, and 102 with bulimia nervosa. Compared to the healthy controls, those with binge-eating disorder were more likely to have had a pregnancy prior to the age of onset, and th ey reported significantly higher levels of negative self-evaluation, parental depre ssion, major depression, conduct problems, deliberate self-harm, critical comments by fa mily about shape, weight, or eating, and teasing about shape, weight, eating or appe arance. Significantly higher exposure to several parent-related risk f actors were reported includi ng parental criticism, high
33 expectations, minimal affecti on, under involvement, matern al low care, and maternal overprotection. These subjects we re also significantly more li kely to report a history of sexual abuse, severe repeated physical abuse, and bullying. The authors also reported that subjec ts with binge-eating disorder were differentiated from their matched subjects w ith other psychiatric disorders by increased exposure to low parental contact, critical comments by family about shape weight or eating, and the incidence of childhood obesity. Interestingly, ther e was no significant difference in individual risk factors between the group with binge-eating disorder and the group with bulimia nervosa, although levels of exposure to some factors were higher in subjects with bulimia nervosa than thos e with binge-eating disorder. The most significant drawback to this st udy is that adolescent subjects were not differentiated from adults, with generic results pr ovided for the entire sample. However, since the mean age of onset of disordered eating was 16.8 year s, it can reasonably be assumed that the risk factors occurred before or during adol escence for many or most of the subjects. Other limitations of this study include the fact that adolescent boys and men were not included in the sample, as well as the retros pective self-report natu re of the study which prevents conclusions about the pred ictive ability of risk factors. Interestingly, a recent prospective lo ngitudinal study on the association of psychiatric disorders with the onset of binge-eating disorder concurred with the finding of Fairburn et al. (1998) that de pression predicted binge-eati ng disorder for adolescents (Zaider, Johnson, & Cockell, 2002). When controlling for age, sex, ethnicity, SES, and comorbid psychiatric disorders, dysthymic disorder was the only psychiatric condition found to predict onset of binge-eating disorder or bulimia nervosa in the sample of 201
34 adolescents. There was no evidence that s ubstance use was a risk factor for eating disorder symptoms in the sample. The authors concluded th at individuals who experience chronic depressive symptoms in ear ly adolescence are at significant risk for development of binge-eating disorder or bulimia nervosa during adolescence. Relationship of Binge-Eating Disorder to Obesity There is a general consensus in the litera ture that a sizeable proportion of obese adults and adolescents have bi nge-eating disorder. In adult patients seeking treatment for obesity, prevalence estimates fo r binge-eating disorder have varied widely from 3.4% to 30%, with interview-based methods of iden tification yielding lo wer prevalence rates (Stunkard, 2002). It should be noted that while most individu als with binge-eating disorder identified from community sample s are overweight, some have never been overweight. One community-bas ed study found that of adul t women with binge-eating disorder, 45% were of normal weight, 39% we re overweight, and 12% were obese (Kinzl et al., 1997). Studies have reported that, co mpared with obese persons who do not binge, obese bingers have more severe obesity, ear lier onset of overweight, earlier onset and more frequent dieting, and higher levels of psychopathology including depression, substance use, and emotional disorders. In his recent analysis of the literature, Stunkard (2002) found continued support for these findings. One recent European study of 126 childre n and adolescents ages 10-16 seeking inpatient treatment for obesity found that 36.5% had engaged in bi nge-eating episodes over the previous month (Decaluwe, Braet & Fairburn, 2002). However, of the 126 patients, 6.1% reported binge-eatin g at least two times per week as required to meet the DSM-IV criteria of binge-eati ng disorder. Females and males had similar rates of binge-
35 eating, or 37.3% and 35.3% respectively. Obese bingers were younger than obese nonbingers, with mean ages of 12.24 years a nd 13.23 years respectively. Unlike obese adults, obese bingers and obese non-bingers in this population did not differ significantly in degree of overweight. Relationship of Binge-Eati ng Disorder to Dieting From a theoretical standpoint, some h ealth experts argue that for those on a restrictive diet that excludes fat, bingeing is sometimes natures way of fighting back in order to allow the body to obtai n the nutrients it needs and is being denied (Hartley, 1998). Research over the last two decades has documented many relationships between nutrients, levels of neurotra nsmitters such as serotonin, dopamine, and norepinephrine, and related effects on mood and behavior. Se vere restrictions in diet can lead to deficiencies in various nutr ients such as Vitamin B12, B6, Thiamin, Riboflavin, Folic Acid, and Vitamin C, then resulting in a bnormal symptoms and behavior ranging from depression to aggression. Conversely, bingei ng episodes, often tri ggered by a stressful event and involving high fat, high calorie f oods, may parallel drug or alcohol abuse in their immediate ability to provide co mfort and gratifica tion (Hartley, 1998). Research on both adolescents and adults has found a correlat ion between dieting behavior and binge eating status. One study found that women with binge eating behaviors dieted significantly more frequently than those who did not binge eat; and that those who went on one or more diets in the la st year or engaged in chronic restrained eating behavior also engaged in more frequent binge eating episodes (Kinzl et al., 1999). In another study of 1919 high school adolescent s, compared to the non-bingeing control group, both girls and boys who had engaged in binge eating in the last year were more
36 likely to report skipping meals and trying to lose weight (Ross & Ivis, 1999). The authors described a gradient in the prevalence of dieting behaviors th at increased with the severity of bingeing/purging behavior, with th e most dieting occurr ing in students with symptoms of bulimia nervosa. This study, how ever, used loose criteria for identifying bingeing, resulting in a sample group with s ubclinical levels of binge eating, limiting generalizability to adolescents with binge-eating disorder. Though it seems clear that a relationship be tween dieting and binge eating exists, the literature shows some disagreement re garding whether or not dieting should be considered a risk factor for binge-eating. Becau se research has suggested dieting is a risk factor for bulimia nervosa and anorexia nervos a, and binge-eating diso rder is similar to bulimia, early research on binge-eating diso rder centered around a similar theoretical relationship between dieting and binge-eating disorder. In one outpatient study of 88 women and 2 men with binge-eating disord er or nonpurging bulimia nervosa, 60% of those with binge-eating disorder report ed that dieting preceded binge eating (Santonastaso, Ferrara, & Favaro, 1999). These results were mo re variable than those of subjects with nonpurging bulimia nervosa, 88.9% of whom reported that they began bingeing after dieting. This study was limited for the purposes of research on adolescents because it included only adults in the sample a nd relied on retrospective self-report of the order of occurrence of dieting and bingeing fo r each patient. According to Stunkards comprehensive review (2002), recent resear ch indicates that more often than not, bingeing behavior precedes die ting rather than the other wa y around. It seems likely that the relationship between binge-eating disorder may be a bidirecti onal one, with dieting preceding binge eating for some adolescents but bingeing leading to dieting for others.
37 Depression in Adolescents Interestingly, as recently as the 1970s, researchers generally held that depression in children and adolescents was a transitory state and developmentally normal, or did not exist at all (Hammen and Rudolph, 1996). Since th at time, the research literature on child and adolescent depression has grown enormous ly and continues to receive attention. Depression also has received varying amounts of attention in the literature on obesity over the past several decades. In th e 1960s and 1970s, many studies focused on psychopathology such as depression that was a ssumed to accompany or cause obesity. In the 1980s and 1990s obesity research moved towards a focus on the contribution of genetics and heritability to variations in weight, along with non-shared environmental causes of obesity. More recently, some studi es have again examined the relationship between obesity and depression, as well as betw een binge-eating disorder and depression. In the interest of preventi ng depression, binge-eating, and obesity, a need exists to understand the nature of these relationships The need to understand the depression component of the model is further highli ghted by the finding that up to 7% of adolescents who develop depression may die by suicide as young adults (Weissman, Wolk, and Goldstein, 1999). The following s ections outline the defi nition of depression, its prevalence in adolescence, risk factors, and the relationship of depression with bingeeating disorder, physical activity, and obesity. Definition of Depression The Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (American Psychiatric Association, 2000) pl aces depression within the Mood Disorders section in Axis I of the DSM-IV-TR manual. It lists criteria for three discrete depressive
38 disorders: major depressive disorder, dyst hymic disorder, and depressive disorder not otherwise specified. The depr essive disorders may present differently in children and adolescents than adults, with an irritable or cranky mood rather th an a depressed mood. Table 5 lists the criteria for a major depressive episode, one or more of which is required for a diagnosis of major depressive disorder. In addition to a consistently depressed or irritable mood or loss of interest in pleasurab le activities over a peri od of two weeks, an adolescent experiencing a major depressive episode also may have trouble sleeping, eating too much or too little, trouble with concentr ation, motor agitation, fatigue, feelings of worthlessness, or suicidal thoughts. The average length of a depressive episode ranges between 16 and 36 weeks (Hammen & Rudolph, 1996). Dysthymic disorder constitutes a somewhat less severe but very persistent form of depression, as compared to major depressive disorder. Rather than symptoms occurring most of the day every day for two weeks as in major depressive disorder, dysthymic disorder in an adolescent is characterized by a depressed or irritable mood for more days than not for at least one year, or two years for adults. The individual is not symptom-free for more than two months and has never ha d a major depressive episode or a manic episode. Depressive symptoms for dysthymic disorder are similar to those of major depressive disorder and may include eating t oo much or too little, sl eeping too much (i.e., hypersomnia) or too little (i.e., insomnia), fatigue, low self-esteem, poor concentration, and feelings of hopelessness. The third variety of depression as ou tlined in the DSM-IV-TR, depressive disorder not otherwise specifie d, serves as a category for depressive symptoms that donot fully meet the criteria for major depressive disorder or dysthymic disorder. This
39 Table 5 DSM-IV-TR Criteria for Major Depressive Episode A. Five (or more) of the following symptoms have been present during the same 2-week period and represent a change from prev ious functioning; at least one of the symptoms is either (1) depressed mood or (2) loss of interest or pleasure. (1) depressed mood most of the day, nearly every day, as indicated by either subjective report or obser vation made by others. Note : in children and adolescents, can be irritable mood. (2) markedly diminished interest or pleasure in all, or almost all, activities most of the day, nearly every day (as indica ted by either subjective account or observation made by others) (3) significant weight loss when not dieting or weight gain (e.g., a change of more than 5% of body weight in a month), or decrease or increase in appetite nearly every day. Note : in children, consider failure to make expected weight gains. (4) insomnia or hypersomnia nearly every day (5) psychomotor agitation or re tardation nearly every day (6) fatigue or loss of en ergy nearly every day (7) feelings of worthlessness or excessive or inappropr iate guilt (which may be delusional) nearly every day (not mere ly self-reproach or guilt about being sick) (8) diminished ability to think or concentrat e, or indecisiveness, nearly every day (either by subjective account or as observed by others) (9) recurrent thoughts of death (not just fear of dying), recurrent suicidal ideation without a specific plan, or a suicide a ttempt or a specific plan for committing suicide B. The symptoms do not meet criteria for a Mixed Episode. C. The symptoms cause clinically signifi cant distress or impairment in social, occupational, or other impor tant areas of functioning. D. The symptoms are not due to the direct physiological effects of a substance (e.g., a drug of abuse, a medication) or a gene ral medical condition (e.g., hypothyroidism). E. The symptoms are not better accounted for by Bereavement, i.e., after the loss of a loved one, the symptoms persist for longer than 2 months or are characterized by marked functional impairment, morbid preo ccupation with worthl essness, suicidal ideation, psychotic symptoms, or psychomotor retardation. Note: From American Psychiatric Association (2000, p. 356)
40 category covers several provisi onal disorders for which re search criteria are provided, including premenstrual dysphoric disorder, mi nor depressive disord er, recurrent brief depressive disorder, and postpsychotic depr essive disorder of schizophrenia. By definition, symptoms present fo r depressive disorder not ot herwise specified are similar to those discussed for major depressive diso rder and dysthymic disorder, but are present to a lesser degree. Prevalence of Depression According to the National Institute of Mental Health (NIMH), epidemiological studies indicate that up to 8.5% of adolescents in the United States suffer from depression (National Institute of Mental Health, 2000). Research by the NI MH also indicates that in recent decades, onset of depression has become ear lier, and that earlier onset is related to more severe depressive sympto ms in adulthood. The rate of depression in adolescents is higher than the 2-3% prevalence rate estimated for children aged six to eleven (Hammen and Rudolph, 1996). During childhood, boys and gi rls have similar rates of depression, but the rate of depression for adolescent gi rls is much higher than for adolescent boys, with the biggest gender differen ce occurring at about age 14. This trajectory continues into adulthood, when depressed females ou tnumber males approximately 2:1 to 3:1 (Hammen & Rudolph, 1996). Risk Factors for Depression Risk factors represent conditions or even ts which have been found to co-occur or correlate with the construct in question. Although the direction of causality is not always clear, the more risk factors and the fewer protective factors that are present for the adolescent, the more likely depressive symptoms will also be present. Risk factors
41 identified in the literature for adolescent depression incl ude stress, cigarette smoking, loss of a parent or loved one, breakup of a roma ntic relationship, attention, conduct, or learning difficulties, abuse or neglect, a nd other trauma including natural disasters (National Institute of Mental Health, 2000). Cognitive factor s that have been found to predict depressive symptoms include pessi mism, low self-concept, and a depressive attributional style, or the te ndency to attribute positive outcomes to external, specific, and unstable factors and negative outcomes to internal, global, and stable causes (Hammen & Rudolph, 1996). Although boys and gi rls are equally at-risk for depression during childhood, by adolescence girl s are twice as likely to de velop depression as boys (National Institute of Mental Health, 2000). This increa sed vulnerability for girls emerges in early to middle adolescence, at around the age of 14 (Hammen & Rudolph, 1996). Some evidence exists that socio-economic status (SES) acts as a risk factor for depression in adolescents. The National L ongitudinal Study of A dolescent Health looked at the effects of SES on the self-rated health of a nationally repres entative sample of 15,483 subjects (Goodman, 1999). SES was defi ned in terms of parental education, occupation, and household income; dependent variables included depression, obesity, asthma, suicide attempt in the last year, and prior sexually transmitted disease. Depression was measured using the Center for Epidemiological Studies Depression Scale (CES-D) with separate cutoff scores for male and female subjects. According to the results, 9.3% of the sample scored above th e cutoff level for depression. Using multiple logistic regression analysis to analyze the data, the research er reported that the indicators of SES were consistently linea rly associated with measures of depression, as well as with
42 obesity, but not for asthma. The author hypothe sized that factors co ntributing to the SES effect on health included h ealth-related behaviors, psyc hological characteristics, residential characteristics, social support, a nd access to health care. One limitation of this study was a reliance on the CES-D as a measure of depression, rather than a diagnostic interview. Also, because the study was school -based, the sample included only enrolled students and not adolescents w ho did not attend school, preven ting generalizability to all adolescents. In addition, research suggests that genetic s and heritability play an important role in early-onset depression, defi ned as two or more episodes of depression before age 25. According to preliminary results from one study on early-onset depression, in which the mean age of onset was 15.6 years, 87% of the sample reported either a firstor seconddegree relative with affec tive disorder (Smith, Muir, and Blackwood, 2003). However, the preliminary results of this ongoing study do not indicate use of a non-depressed control group. Until a specific genetic marker can be located, or twin studies are conducted to rule out environmental influe nces, it will remain unclear whether the familial transmission of depression is due to biological or psychosocial causes, or both. Based on available research, it seems reasona ble to hypothesize that depression results from some combination of the two, with adolescents biologically predisposed to depression and then exposed to a triggering event. Relationship of Depression with Binge-Eating Disorder Recent research points to a probable bidi rectional relationship between depressive symptoms and binge-eating disorder. A pros pective longitudinal st udy (Zaider, Johnson, & Cockell, 2002) examined the association of psychiatric disorders with the onset of
43 binge-eating disorder and bulimia nervosa dur ing adolescence. The researchers reported that, in the sample of 201 adolescents re ferred from school nurse offices and clinics, dysthymic disorder predicted the onset of bi nge-eating disorder or bulimia nervosa. Conversely, adolescents with eating disorder symptoms at baseline were significantly more likely to have dysthymic disorder at follow-up, suggesting a cyclical relationship between the two disorders. Regarding age of onset of binge eating and depressive symptoms, 40% of the sample reported that depression or anxiety had preceded their eating problems, 20% reported that eating pr oblems preceded their depression or anxiety, and 40% reported that their difficulties with eating and depression or anxiety had started within the same year. While they stopped short of claiming evidence of causality, the authors concluded that a bidir ectional association most likel y exists between depression and binge-eating disorder. Another recent study examined the correlation between binge eating and depressive symptoms (Ackar d, Neumark-Sztainer, Story, & Perry, 2003). A sample of 4746 middle and high school girls and boys fr om Minnesota completed a depressive mood scale and a 221-item questionnaire on nu trition and eating habits. The authors declined to draw conclusions about causality because the st udy was cross-sectional, but they reported that students who met the cr iteria for binge eating disorder scored significantly higher on a depressi ve mood scale than those w ho indicated subclinical or no binge eating. One limitation of this st udy was the questions used to assess binge eating. The questions were adapted from an adult measurement scale for binge-eating disorder and therefore reliability and validity data were not available.
44 Relationship of Depression with Physical Activity It seems reasonable that because an adoles cent with depression is highly likely to experience the depressive symptom of fa tigue (Hammen and Rudolph, 1996), that it would follow that his or her level of physical activity would be low. Unfortunately, little research exists on the relationship betw een depression and physical activity in adolescents in regard to exer cise as a treatment for depr ession or as a preventative measure for depression (Paluska, 2000). Howe ver, according to a re cent review of the literature on treatment of clin ically depressed adults with physical exercise (Brosse, Sheets, Lett, and Blumenthal, 2002), interven tion studies have provi ded evidence that exercise treatment is more effective than no treatment and it is as effective as psychotherapy and medication treatment. In addition, both cr oss-sectional and prospective studies found signi ficant relationships between regular physical activity and lower scores on depression questionnaires. This correlation of exercise with fewer depressive symptoms has been suggested in samples of clinically depressed adults, as well as in healthy, non-depre ssed adults, cardiac patients, patients with chronic obstructive pulmonary disorder (COPD), and patients with neuromuscular disorders, rheumatoid arthritis, and osteoarthritis. The authors caution that the existing research base on adults is characterized by poor design and many methodol ogical flaws. In addition, it is uncertain whethe r these correlations and effe cts of physical activity will generalize to adolescents. German research on adolescents does s upport the idea that phys ical exercise is associated with lower levels of anxiety and depression, as measured by the anxiety/depression scale of the German ve rsion of the Achenbach Child Behavior
45 Checklist (Kirkcaldy, Shephard, & Siefen, 2002). A group of 1000 adolescents were given a questionnaire about phys ical activity level and as ked to rank their regular continuous involvement in endur ance sports such as cycl ing, swimming, and running. The authors reported that the adolescents w ho engaged in regular physical activity also had significantly lower anxietydepression scores as compared to less active groups in the sample. However, the generalizability of th is study to American youth is questionable since the study was conducted in Europe. Relationship of Depression with Obesity Theory and research on obesity in the 1960s and 1970s often focused on psychological and behavioral aspects that were thought to underlie or co-occur with obesity. In the following decades, obesity came to be viewed as predominantly a genetically mediated, biol ogically-based, physical di sorder, and research on psychological correlates of obesity declin ed (Wilson et al., 1996). For many years, researchers of obesity and depression have had little communication and it was taken for granted that little or no rela tionship existed between the tw o disorders (Stunkard et al, 2003). Recently, however, seve ral well-regarded medical re searchers in the field of obesity and eating disorders have turned their attention towards understanding the relationship between depression and obesity (Goodman & Whitaker, 2002; Stunkard et al., 2003). Thus, in a sense, the focus of re search on depression and obesity has come full circle with recent renewed at tention on a possible direct relationship between the two disorders. Stunkard et al. (2003) proposes a moderator/mediator framework for investigating the link between depression and obesity in adults, characterized by
46 moderator variables which define who is affected by the independent variable, and mediator variables which define to what ex tent the dependent vari able is affected. According to this research-based theory, moderating variables in clude severity of depression, severity of obesity, gender, socioeconomic status gene-environment interactions, and adverse childhood experiences Mediating variables acting between the constructs of depression and obesity in clude eating and physic al activity, teasing, disordered eating, and stress. This theoreti cal framework is valuable and especially useful for guiding future research, however it is based on research with adults. A similar model based on research in adolescent depr ession and obesity would be helpful in understanding these relationships for adolescents. Stunkard et al. (2003) view severity of depression as a moderator variable for the association between adolescent depression and obesity because, while studies have not found a relationship between subc linical levels of depression and obesity in adults, one longitudinal study found a predictive relationshi p between major depression in children six to 17 years and their adult BMI as measur ed 10-15 years later (P ine, Goldstein, Wolk, and Weissman, 2001). This study used an expe rimental group of 90 children with major depression and a control group of 87 children with no psychiatric disorder. Children and adolescents with depression at baseline had a mean adult BMI of 26.1, compared to 24.2 for the control group. Although these result s make a valuable contribution to the research, the study failed to differentiate child ren from adolescents at baseline, and they did not examine shorter-term effects of depr ession on BMI that may have been evident before the participants reached adulthood. It should also be noted that childhood depression did not specifically predict adul t obesity, but rather a higher BMI. Also,
47 depression was not isolated as the sole predic tor of BMI, because adult poverty was also found to predict adult BMI. The study contro lled for socioeconomic status, age, gender, cigarette and alcohol use, and pregnancy a nd medication use, but it did not control for parental BMI. Some recent evidence does point to a po ssible causal role of depression in the occurrence and persistence of obe sity within adolescence. Researchers used data from the 1995 and 1996 waves of the National L ongitudinal Study of Adolescent Health (Goodman and Whitaker, 2002). The school-based sample consisted of 9374 adolescents and was nationally representative of students in grades seven through 12. There was no significant correlation between baseline depres sion and baseline obesity. However, when controlling for socioeconomic status, physical activity, BMI at baseline, age, race, gender, parental obesity, nu mber of parents in the home, self-esteem, delinquent behavior, and smoking, researchers reported that depressive symptoms predicted obesity at follow-up. Conversely, no evidence was found that obesity at baseline predicted depression at follow-up. Some limitations of this study were the use of self-reported rather than measured height and weight at baseline, the use of a self-report measure of depressive symptoms rather than a diagnostic interview, and the non-inclusion in the sample of youth un-enrolled in school. In summary, little research exists on th e correlational or the causal relationship between depression and obesity in adolescen ts. Some longitudina l evidence supports a causal relationship between childhood depression and adult BMI, and between adolescent depression and adolescent obesit y. Theoretical explanations of a causal relationship from obesity to depression in adolescents have not yet received empirical support. Thus, the
48 available research suggests that the causal relationship between depression and obesity is unidirectional, with depression as a predictor of weight and obesity. Conclusion In conclusion, there are many risk factor s and critical medical and psychosocial outcomes associated with obesity in adoles cence. As emphasized by Steinbeck (2001), obesity has traditionally been viewed as a medical problem in the individual, caused by the individual, and which should be solved by th e individual, rather th an as a social issue with negative medical and psychosocial outcom es. Rather than taking this limiting view, we should strive to treat obe sity as a preventable problem placing responsibility for prevention not on one person or instituti on, but on society as a whole through community-based programs. A dditional research is necessary to confirm the proposed risk factors discussed in th e literature, and pr evention programs should focus on those risk factors that can be infl uenced or manipulated. It is only through effective, widereaching prevention programs that the epidem ic of childhood and adolescent obesity and its debilitating comorbid conditions ca n be slowed and ultimately stopped. Research indicates that adolescents w ho develop binge-eating disorder represent the subpopulation of obese adoles cents who are perhaps the most in need of intervention, due to their higher rate of depression and other psychopathology. Some treatments for binge-eating disorder have shown promise, pa rticularly cognitive-behavioral therapy and interpersonal psychotherapy, but research show s they are not effective for approximately 40% of those who seek help (Stice, 1999). Additional knowledge must be gained in the mental health and health care fields about this disorder and its rela tionship to depression and obesity in order to improve treatment success rates.
49 Much of the literature on obesity, ea ting disorders, and depression fails to differentiate between results for adolescents a nd adults. In order to effectively plan for assessment and intervention for adolescents wi th disordered eating behaviors, it is necessary to be aware of the differences be tween adolescents and adults. For example, one study found that, compared to adults treate d for an eating disorder, adolescents more often presented with a diagnosis of eating di sorder not otherwise specified, had a lower global severity score, greater level of denial, less desire for help, and a history of fasting (Fisher et al., 2001). A need exists for addi tional research that focuses on adolescents. Furthermore, many of the studies on eating disorders included only females in their sample, and much of the literature on binge -eating disorder has been conducted in Europe. Additional research should be conduc ted that looks at both girls and boys, and more investigations are needed in the United States. More varied samples in terms of age, gender, nationality and culture will improve the external validity of that research for those demographic variables. The research described in this revi ew indicates the presence of several relationships among obesity, phys ical activity, depressive symptoms, binge eating, and dieting. If it can be determined that binge-e ating disorder truly ha s a causal relationship with both depression and obesity, then ta rgeting binge eating for intervention and prevention research would potentially also benefit adolescents with depression and obesity. Similarly, if there is evidence that physical activ ity has a causal relationship with both depression and obesit y, then support will be given to targeting physical activity for intervention and prevention research. In addition, evidence that special education status has a high correlation with weight and obesity would support a focus on this
50 population for intervention and prevention. The same concept applies to correlations of weight with SES and gender/race, the othe r non-manipulable risk factors for obesity. Research Questions The objective of the pres ent study was to answer three research questions: (1) To what degree do the data support the propos ed model of the correlates of weight in adolescents? (2) For the populations of obese and non-obese adolescents, what is the level of binge-eating symptoms, and what is the strength of the re lationship between binge-eating symptoms and depressive symptoms? (3) For the population of adolescents who e ngage in both binge-eating and dieting behavior, what is the order of preceden ce of binge-eating and dieting (i.e., which occurs first, binge-eating or dieting)? Path Model Figure 1, appearing in Chap ter 1, illustrates the proposed path diagram that depicts the relationships between the correlate s of obesity as curre ntly supported by the literature. Straight lines with arrows re present relationships with exogenous outcome variables to be analyzed in th e study. The direction of the arrow symbolizes the direction of causality most clearly supported in th e literature. Curved dotted lines with bidirectional arrows represent assumed co rrelations between e xogenous variables. Additional details related to the path model and the data analysis to be performed will be addressed in the data anal ysis section of Chapter 3.
51 Chapter III: Research Methods Design The study used a path analysis design, which is a group correlational design in which the independent variables are continuous and are not manipulated. The path model represents hypothetical relations hips between pairs of variable s in the model. The model included exogenous variables, which are defi ned as those that are caused by forces outside the model, and endogenous variables, or those that are influenced by other variables within the model. Some relations hips were hypothesized to be unidirectional, with the flow of causality from one variable to another in only one direction, and other relationships were thought to be bidirecti onal, with the flow of causality in both directions. The path model in its entirety represents the hypo thetical relationships of the constellation of variables most significantly related to weight a nd obesity, as supported by theory and research. The path diagram in Figure 3 includes four exogenous variables (i.e., not predicted by other variables on the diagram): socio-economic status, special education status, gender/ethnicity, and physical activity. The model al so includes four endogenous variables (i.e., predicted by variables on the diagram): weight, dieting, binge-eating, and depression. All variable data types ar e continuous except for three dichotomous variables: socio-economic status, special education status, and gender/ethnicity. Figure 3 contains MODEL1, the initial model for the correlates of weight in adolescents. The model was constructed by the primary investigat or for the purpose of
52 the present study. The model shows the in itially proposed relationships between adolescent obesity, binge-eating behavior depressive symptoms, dieting, physical exercise, gender and ethnicity, socio-economic status, and special education status. The relationships depicted in the model represen t those that receive the most empirical and theoretical support in the literature. Path analysis, a type of structural eq uation modeling, was used to analyze the model and to evaluate the models goodness of f it with the data. Each of the rectangular boxes in the path diagram represen ts a measured variable which is an indicator of a latent variable (Hatcher, 1996). For example, depr essive symptoms repres ent the total scaled score on the RADS-2, which in tu rn is an indicator of the latent variable depression. Straight lines on the diagram symbolize pred iction, leading from predictor variable to outcome variable (Hatcher, 1996). For ex ample, the model proposed that physical activity, binge-eating symptoms depressive symptoms, soci o-economic status, special education status, and gender/ethni city act as predictors of we ight. Curved, dotted lines on the path diagram are bidirectional with two arrows and represented an expected correlation between two exogenous variables (Hatcher, 1996). For example, a correlation was expected between socio-economic status and special education status (Tershakovec, Weller, & Gallagher, 1994), and between soci o-economic status and gender/ethnicity (National Research Council, 2002 ). In the statistical anal ysis, the mathematical model assumed a correlation of zero between any two unrelated variables on the model. Following data collection, it became necessa ry to modify the path model to be used in data analysis to the one shown in Fi gure 4. Initially, gender and ethnicity were to be combined as a single vari able to allow analysis of ge nder-ethnic specific groups such
53 Figure 3. MODEL1: Initial Path Diagram, Correlates of Weight in Adolescents. as Caucasian females, African-American female s, Latino-Hispanic males, etc. However, the small sample sizes of many of these gende r-ethnicity groups precl uded using them in the path analysis. As discussed in the fo llowing section, only three sub-groups had more than 10 participants: White Females, White Ma les, and Latino/Hispanic Females. It was decided not to exclude gender entirely from the model because to do so would cause the path estimate results to be biased if gender did influence other variables in the model. Consequently, the gender and ethnicity variab les became separate va riables within the model. Both gender and ethnic ity were treated as exogenous variables in the model, with ethnicity but not gender depicted as influenc ing weight directly, as suggested in the literature (National Center fo r Health Statistics 2002). The MODEL2 diagram also illustrates the coding of dummy variables used in the data set for the variables socioDepressive Symptoms Binge Eating Symptoms Physical Activity Socio-Economic Status (Yes/No) Obesity/Weight Special Education Status Dieting Gender/Ethnicity CA HI AS NA O AA MODEL1
54 Figure 4. MODEL2: Revised Path Diagram, Correlates of Weight in Adolescents. economic status, special education status, gender, and ethnicity. Because they are categorical in nature, these variables had to be treated as dummy-coded variables, in order to include them in the path analysis, and ethnicity had several possible values while the other categorical va riables each had two possible valu es. Consequently, a reference group had to be chosen for ethnicity, to whic h all other ethnicities w ould be statistically compared during the data analysis. Caucas ian, which was chosen by the greatest number of participants, was chosen as the reference group. The selection of Caucasian as a reference group often occurs in research, as in the National Research Councils use of Depressive Symptoms Binge Eating Symptoms Physical Activity Socio-Economic Status (1=Free/Red 0=Not Free/Red) Obesity/Weight Special Education Status (1=ESE, 0=Not ESE) Dieting Ethnicity (AA) Gender (1=Male 0=Female Ethnicity (LH) MODEL2
55 risk indices when calculating odds ratios for various ethnic groups when examining special education placement rates (National Research Council, 2002). The path model reflects that the Caucasian et hnicity served as the refere nce group, so that the path analysis essentially compared the relations hips of the African-American group and the Latino/Hispanic group to those of the Caucasian group. Thus the model contains boxes for the African-American and Latino/Hispanic groups but not for the Caucasian group. Participants and Setting The study was conducted during Spring, 2005 of the 2004-2005 school year and Fall, 2005 of the 2005-2006 school year in two central Florida high schools, hereafter referred to as High School 1 and High School 2. Both schools were within the same county and school district. High School 1 had an approximate enrollment of 1,638 9th 12th graders in 2004-2005, and 1,700 in 2005-200 6. The school was located in a growing, suburban/rural area of west-central Florida. As illustrated in Table 6, High School 1 was less demographically diverse th an the district or state, with a 2005-2006 student population that was 92.2% White, 0.9% Black, 4.5% Hispanic, 0.9% Asian, 0.5% American Indian, and 1.1% Multiracial. Approximat ely 43.1% of High School 1s enrollment was reported to be economically disadvantaged, a rate similar to the 43.5% rate for the district and 45.9% rate for the state of Florida (Florida Department of Education, 2006b). High School 1 received a grade of D for the 2004-2005 school year and a C for 2005-2006 from the state of Florid a as part of the st atewide school grading accountability system. According to the Fe deral Governments No Child Left Behind accountability system, High School 1 failed to ma ke Adequate Yearly Progress in either the 2004-2005 or 2005-2006 school year, mee ting 73% and 77% of the criteria
56 respectively (Florida Department of Educati on, 2006a). According to the most recent data available from the Florida School I ndicators Report, for the 2004-2005 school year, 0.5% of students were designate d as Limited English Proficie nt. The report indicated a stability rate of 90.7% of students that re mained enrolled throughout the year, and a graduation rate of 73.7%. The school al so reported that 23.5% of students had disabilities, and 5.3% of students took the SAT college admission test (Florida Department of Education, 2005). Table 6 Demographic Diversity of High Schools, District, and State Ethnicity High School 1 High School 2 District State White 92.2% 64.1% 78.9% 47.7% Black 0.9% 14.3% 4.5% 23.4% Hispanic 4.5% 19.7% 11.3% 23.4% Asian 0.9% 0.5% 1.7% 2.2% American Indian 0.5% 0.2% 0.3% 0.3% Multiracial 1.1% 1.3% 3.3% 3.0% Note. Adapted from Florida Department of Education, 2006b. Retr ieved September 24, 2006 from http://doeweb-prd.doe.state.fl .us/eds/nclbspar/main0506.cfm. High School 2 had an approximate enrollment of 1,294 students in 2004-2005 and 1,352 in 2005-2006. The school was located in a ru ral area of west-central Florida. Demographically, High School 2 was more di verse than High School 1 and the school population in 2005-2006 was reported to be 64.1% White, 14.3% Black, 19.7% Hispanic,
57 .5% Asian, .2% American Indian, and 1.3% Mult iracial. Table 6 contrasts the ethnic makeup of High School 1 with that of High Sch ool 2, along with the district and state of Florida. Approximately 50.7% of stude nts were reported to be economically disadvantaged in that same school year (F lorida Department of Education, 2006b). High School 2 also received a D grade for bot h the 2004-2005 and 2005-2006 school year from the state of Florida as part of the statew ide school grading accountability system. According to the Federal Governments No Child Left Behind accountability system, High School 2 also failed to make Adequate Yearly Progress in either the 2004-2005 or 2005-2006 school year, meeting 60% and 64% of the criteria respectively (Florida Department of Education, 2006a). The most recent data available from the Florida School Indicators Report show that for the 2004-2005 school year, 2.4% of High School 2 students were designated as Limited English Proficient. The report indicated a stability rate of 91.2% of students that remained enrolled throughout the year, and a graduation rate of 72.6%. The school al so reported that 26.4% of st udents had disabilities, and 4.2% of students took the SAT college admission test (Florida Department of Education, 2005). The participants of this study consiste d of 252 students in grades 9 through 12. One participants data were excluded from the data set because she was pregnant, resulting in an N of 251. Th is students data were excl uded due to the weight gain associated with pregnancy, because her weight was not representative of her normal weight. As shown in Table 7, participants were 56% female and 44% male. Fifty-three percent attended High School 1 and 47% High School 2. Table 8 shows the distribution of participants across ethnicity and gender. The participants were 79% Caucasian, with Latino/Hispanics identifying themselves as the largest minority. Seven students
58 identified themselves as a combination of tw o or three ethnicities, such as Caucasian and Latino/Hispanic. Table 9 depicts the ethnic di stribution of the participants within each school, and this distribution resembled school enrollment. As shown in Table 10, Table 7 Distribution of Participan ts by Gender and School Gender N Percentage N=Total Participants 251 100% Female 141 56% Male 110 44% High School 1 133 53% High School 2 118 47% High School 1 Female 66 50% Male 67 50% High School 2 Female 75 64% Male 43 36%
59 Table 8 Distribution of Participants by Ethnicity and Gender Ethnicity N Percentage Total Participants 251 100% AA= African American 12 4.8% Female (50%) 6 Male (50%) 6 AS= Asian 6 2.4% Female (100%) 6 Male 0 CA= Caucasian 200 79.7% Female (52%) 104 Male (48%) 96 LH= Latino/Hispanic 21 8.4% Female (81%) 17 Male (19%) 4 NA= Native American 3 1.2% Female (33.3%) 1 Male (66.7%) 2 OT= Other 2 0.8% Female (100%) 2 Male 0 AA+CA+LH 1 0.4% AA+LH 1 0.4% AA+OT 1 0.4% CA+LH 3 1.2% CA+NA 1 0.4%
60 Table 9 Distribution of Participants by Ethnicity, Gender, and High School* Ethnicity High School 1 High School 2 N Percentage N Percentage AA= African American 2 1.5% 10 8.5% Female 1 5 Male 1 5 AS= Asian 4 3% 2 1.7% Female 4 2 Male 0 0 CA= Caucasian 116 87.2% 84 71.2% Female 55 49 Male 61 35 LH= Latino/Hispanic 7 5.3% 14 11.9% Female 5 12 Male 2 2 NA= Native American 1 0.8% 2 1.7% Female 0 1 Male 1 1 OT= Other 0 0% 2 1.7% Female (100%) 0 2 Male 0 0 Note: Percentages do not sum to 100%. Three participants from High School 1 and four from High School 2 provided more than one ethnicity and are excluded from this table.
61 Table 10 Distribution of Particip ants by Grade and School Overall High School 1 High School 2 Grade N Percentage N Percentage N Percentage Total Participants 251 100% 133 53% 118 47% 9 89 35% 54 40.6% 35 29.7% 10 63 25% 31 23.3% 32 27.1% 11 42 17% 18 13.5% 24 20.3% 12 57 23% 30 22.6% 27 22.9% participants came from all grad es but overall twice as many 9th graders participated than 11th graders. The participants were also sim ilar to the schools enrollment in terms of the percentage of special educa tion students. Approximately 45% of participants overall were eligible for free or reduced lunch, with 50% eligible at High School 1 and 38% at High School 2. Table 11 illustrates free and reduced lunch eligibility by school. As shown in Table 12, 26% of Hi gh School 1s participants and 19% of High School 2s participants were enrolled in one or more special education programs at the time of data collection. The initial goal for particip ant recruitment was 200 students from each school. Following the 1st round of data collection at both schools in Spring 2005, a 2nd round of data collection was conducted at Hi gh School 2 in Fall 2005 in an effort to increase the numbers of minority participants in the study.
62 Table 11 Distribution of Participants by Free/Reduced Lunch and School Overall High School 1 High School 2 Free/Reduced Status N Percentage N Percentage N Percentage Total Participants 251 100% 133 118 Elig. Free or Reduced 112 45% 67 50% 45 38% Not Elig. Free or Reduced 139 55% 66 50% 73 62% Measures Four self-report measures were utilized to gather data for this study: the Demographic and Physical Activity Ques tionnaire (DPAQ), the Reynolds Adolescent Depression Scale-2nd Edition (Reynolds, 2002), the Eating Disorders Inventory-2 (Garner, 1991), and the Dutch Eating Behavi or Questionnaire (DEBQ). In addition, the school nurse, health assistant, and the exam iner measured height and weight for each participant, which they used in turn to calculate body mass index (BMI). Each measurement tool is reviewed in this section with regard to the test components, test administration, and technical adequacy. The data elements provided by the instrument for the present study are defined. Demographic and Physical Activity Questionnaire (DPAQ) School nurses and clinic staff admini stered the Demographic and Physical Activity Questionnaire (DPAQ, Appendix B) th at contains 10 demographic and personal information items, six physical activity items and requires approximately 5 minutes to complete. The present study derived seven data elements from this questionnaire, including BMI, free or reduced lunch status, special education services received, gender,
63 Table 12 Distribution of Participants by ESE (E xceptional Student Education) Status Special Education Status N Percentage Total Participants (N) 251 100% Not Special Education 195 78% Special Education 56 22% *A=EMH (Educable Mentally Handicapped) 8 14% *B=TMH (Trainable Mentally Handicapped) 2 4% *C=OI (Orthopedically Impaired) 4 7% *F=Speech Impaired 13 23% *G=Language Impaired 12 21% *I=VI (Visually Impaired) 2 4% *J=EH (Emotionally Handicapped) 6 11% *K=SLD (Specific Learning Disability) 29 52% *L=Gifted 4 7% *V=OHI (Other Health Impaired) 2 4% High School 1 133 100% Not Special Ed 99 74% Special Ed 34 26% High School 2 118 100% Not Special Ed 96 81% Special Ed 22 19% *Note: Some students classified with >1 special education exceptionality; A-V percentages calculated as percent of students classified as ESE.
64 ethnicity, and two types of phys ical activity engaged in with in the last seven days. Following is a discussion of each data element gathered via the questionnaire. Body Mass Index (BMI) Weight was measured using a beam balance scale, and height was measured with the participant standing straight against a ve rtical scale attached to the beam balance scale. Both measurements were taken w ith participants wearing no shoes or heavy jackets. BMI was determined after measuring each students height in inches and weight in pounds. BMI was calculated using an instrument that used the generally accepted formula ( kg / m2 ). At High School 1, the nurse util ized a standard BMI wheel on which the height and weight were li ned up to yield the BMI. At High School 2, the nurse used a BMI calculator which produced the BMI after th e height and weight were keyed in. All measurements and calculations were performe d by the school nurse, he alth assistant, or the examiner. The BMI, stated as a decima l number with a resolu tion of 0.1 (e.g., 18.5), was then recorded on the survey for e ach participant in the appropriate box. Socio-Economic Status (SES) Free or reduced lunch status, which served as the socio-economic status for the study, was determined for each student from reports generated by the data entry employee at High School 2. The status was represented by a Y to indicate that the student was eligible for free or reduced lunch, or N to indicate that the student was not. Exceptional Student Education (ESE) Status The special education codes for each student were obtained from reports generated by the data entry employee at Hi gh School 2. ESE Status had possible labels of A-Educable Mentally Handicappe d, B-Trainable Ment ally Handicapped, C-
65 Orthopedically Impaired, F-Speech Impa ired, G-Language Impaired, I-Visually Impaired, J-Emotionally Handicapped, K-Speci fic Learning Disabil ity, L-Gifted, and VOther Health Impaired. Sex/Grade/Age/Ethnicity Students provided demographic information on the survey about themselves. Each participant marked the checkbox for Male or Female, a checkbox for 9th, 10th, 11th, or 12th grade, and a checkbox for ages ranging from 13 to 20 years. Ethnicity was provided by marking the appropriate ch eckbox for the question, Do you think of yourself as ? Choices consisted of Af rican American, Asian, Caucasian/White, Latino/Hispanic, Native American/Alaskan, a nd Other, with a space for the student to write an ethnicity not liste d. For the purposes of this study, sex and ethnicity were initially to be combined into one variable to allow analysis of data for each combination of sex and ethnicity (e.g., Af rican American girls, Afri can American boys, Caucasian girls, and Caucasian boys). However, due to the small sizes of so me of the groups it became necessary to treat ethnicity and gende r as two separate variables during data analysis. The path model was modified accordingly and given the name MODEL2. Physical Activity The DPAQ questionnaire measured physical activity via seven questions, with the first two questions each provi ding the number of days duri ng which the student engaged in exercise that either a) made the student sweat and breathe hard, or b) did not make the student sweat and breathe hard. The replie s for those two items were summed and used as the physical activity variable in the study. These two it ems were selected to represent
66 physical activity because together they make up all types of exercise, assuming that all exercise either does or does not make one sweat and breathe hard. The physical activity questions on the re vised survey were derived from the Youth Risk Behavior Survey Surveillan ce (YRBSS), developed in 1990. The YRBSS was used to monitor behaviors linked to deat h, disability, and social problems in the United States. Test-retest reliability for the physical activity it ems on the YRBSS was found to be somewhat low at approximately .55 (Brener et al., 1999). However, it is unclear what time interval was used in this analysis. Reynolds Adolescent Depression Scale-2nd Edition (RADS-2) The Reynolds Adolescent Depression Scale-2nd Edition, or RADS-2 (Reynolds, 2002), was used to screen participants fo r depressive symptoms. The RADS-2 was published in 2002 and is the most recent vers ion of the Reynolds A dolescent Depression Scale, originally published in 1987 by Willia m Reynolds. The RADS-2 is a self-report measure designed for adolescents 11-20 years and can be administered in 5-10 minutes, either individually or in a group. It consists of 30 items written at the 3rdgrade level, and it provides standard T-scor es and percentile ranks for to tal depression and each of four subscales: Dysphoric Mood, Anhedonia/ Negative Affect, Negative Self-Evaluation, and Somatic Complaints. The T-Score for total depression provided the data element for depressive symptoms in the current study. The item conten t is designed to reflect DSMIV criteria for depression. Empirically supported cut-off sc ores indicate the clinical severity of depressive symptoms (e.g., normal, mild, moderate, or severe) and assist in identification of adolescents who may have Ma jor Depressive Disord er. In addition, six critical items are included to inform the clin ician that the adolescent may need immediate
67 intervention. The materials include a ha nd-scorable test book let and an optional summary/profile form that can be used to char t results of the four subscales. The present study used the hand-scorable te st booklets but not the summary/profile forms, which are helpful for analyzing and communicating re sults but are not n ecessary for scoring. The RADS-2 was standardized on a samp le of 3,300 students. The sample was stratified by gender, age, and ethnicity to reflect the 2000 U.S. Census, and normative data are provided in the professional ma nual for each subgroup. The normative sample included an equal number of males and fema les, and each age group consisted of 1,100 adolescents. Considerable evidence of relia bility and validity for the RADS-2 is also provided in the manual, and includes a school-based study with a sample of 9,000 adolescents and a clinic-based study with a sa mple of 297 adolescents Reliability data exist for the total scale as well as each of the four subscales, and include internal consistency, test-retest, and standard errors of measurement. Internal consistency estimates for the RADS-2, based on a school sample of over 9,000 students, was high on the Depression Total scale (r = .93), modera tely high (r = .86) for the subscales, and moderately high (r = .86) for males and female s. Test-retest reliability, measured for 1750 students at a two week time interval, was determined to be high for the Depression Total scale (r = .85), and moderately high for the subscales (r = .82). Standard errors of measurement on the total school sample of over 9000 students ranged from 2.71 for the Depression Total to 4.5 for the Somatic Comp laints subscale, based on T Scores which have a mean of 50. Validity data on the RADS-2 include re sults of content, criterion-related, convergent, discriminant, and clinical validity analyses. Content validity was assessed
68 for the standardization sample of 3,300 adol escents through item-with-total Depression Total scale coefficients, and was considered to be adequate (median r = .53). The median item-with-total correlation coefficients for the four subscales also were considered to be high and ranged from .53 to .66. The RADS-2 manual also reports high criterion validity as measured by correlations between the R ADS-2 and the Hamilton Depression Rating Scale, a clinical interview fo r depression (r = .82). Criteri on validity also is supported by high correlations between the RADS-2 and othe r self-report measures of depression such as the Adolescent Psychopathology Scale (r = .74 to .76), the Minnesota Multiphasic Personality Inventory (r = .78) and the Beck Depression Inve ntory (r = .80). Convergent validity of the RADS-2 has b een demonstrated with measur es of related constructs including self-esteem, anxiety, and suicidal be haviors. RADS-2 discriminant validity was assessed using scales of social desirabilit y, IQ, conduct disorder, substance abuse, and mania and determined to be low and therefor e acceptable (r = .11 to .37). Finally, the RADS-2 manual reported support for contrasted clinical groups valid ity with a difference of over two standard deviations (sd = 2.19) in standard scores betwee n a clinical sample of adolescents with depression and a school-based control group. Eating Disorder Inventory -2 (EDI-2) The Eating Disorder Inventory-2 (ED I-2) was published in 1991 and is the expanded form of the Eating Disorder Inve ntory (EDI), originally published in 1983. The EDI and EDI-2 were designed to assess sy mptoms of anorexia nervosa and bulimia nervosa in adolescents and adu lts 12 years and older. It is intended as a screening instrument and not as an exclusive method for diagnosing eating disorders. The EDI-2 requires about 20 minutes to complete a nd consists of 91 self-report items on 11
69 subscales. Items were written on a 5th grade level (Netemeyer & Williamson, 2001). Eight subscales originate from the EDI and re main unchanged in the EDI-2: 1) drive for thinness, 2) bulimia, 3) body dissatisfaction, 4) ineffectiveness, 5) perfectionism, 6) interpersonal distrust, 7) intero ceptive awareness, and 8) maturity fears. Three subscales were added to the 64-item EDI to create the ED I-2: 1) asceticism, 2) impulse regulation, and 3) social insecurity. The EDI-2 manual contains references to over 250 studies, reflecting the large amount of research involving the EDI and eating disorders. Reliability and validity data on the EDI-2 are strong (Ash, 1994; Shinke, 1994). Internal consistency was estimated at .8 and higher for the eight original subscal es, for both eating disorder samples and nonpatient comparison groups (Ash, 1994). Test-re test reliability es timates for oneand three-week intervals between the tests are in the .8 and higher range. Validity research has found that the original ED I differentiates between patient s with eating disorders and non-patient samples. Other research ha s shown evidence of the EDIs convergent validity with correlations between the EDI and other eating disorder scales such as the Eating Attitudes Test (Ash, 1994). Most students participating in the study were observed to fill out the one bulimia scale on the EDI-2 in less than five minutes. The current study utilized only the EDI-2 bulimia scale for the binge-eating symptoms va riable. The study did not make use of the other ten subscales. The bulim ia scale is made up of six bingeing items and one purging item. Each item is on a six-point Likert sc ale with responses ra nging from always to never. All subscale scores on the EDI-2 are calculated by adding the raw item scores together for that subscale. The bingei ng subscale score for the present study was
70 calculated in this manner. Because the study examined binge eating symptoms rather than purging symptoms, the one purging item was not included in the subscale raw score used for data analysis. Raw scores are convert ed to percentile ranks for each subscale of the EDI-2, and the manual does not report standa rd scores, T-scores, or scaled scores. The manual does list per centile ranks for normative groups of bulimia nervosa patients, anorexia nervosa patients, high-school boys and girls, and female and male college students. However, the high-sc hool percentile ranks were inappropriate for data analysis in the current study because of the exclusion of the purging ite m from the subscale score. Similarly, Perez and Joiner (2003) measured disordered eating using only the bingeing items from the bulimia scale and reported that the raw bulimia subscale without the purging item had a kappa reliab ility coefficient of 0.82. In order to address the research questi on related to the orde r of precedence of dieting and binge-eating, one question was a dded on the EDI-2. On the bulimia scale of the EDI-2, the question was added: Have you ever experienced a binge eating episode -a time when you ate an amount of food in less th an 2 hours that was definitely larger than most people would eat in the same situ ation, and you felt like you could not control yourself or stop yourself? ______ (Yes/No) If Yes, how old were you when you first experienced an episode: ______ year s old (ex: 14 years old). The wording for this question was based on the definition of a binge in the Diagnostic and Statistical Manual of Mental Disorders (American Psyc hiatric Association, 2000) and was intended to clearly and simply define b inge eating for participants.
71 Dutch Eating Behavior Questionnaire (DEBQ) The Dutch Eating Behavior Questionnaire (Van Strien, 2002) provided the dieting data for the study. This questionnaire was de signed for adults, adolescents, and children nine years old and up. The DEBQ was crea ted in 1986 and consists of 33 items on 3 scales that measure eating be havior: 1) emotional eating, 2) external eating, and 3) restrained eating. Questions ar e presented on a five point Li kert scale and are estimated to be on a reading level between 5th and 8th grades. It is availabl e in Dutch or English. Each scale in the DEBQ correlates to a theo retical cause of overeating, and the author further recommends guidelines for intervention and treatment of disordered eating based on the assessment results for each scale. A ccording to the theory underlying the restraint scale, individuals overeat afte r a period of dietary restricti on and the decision to eat less than wanted is no longer governing behavi or. The emotional eating scale, which corresponds to psychosomatic theory, assumes that one overeats in response to negative emotions. The external eating scale is relate d to externality theory which proposes that individuals overeat in response to external cues such as the smell and sight of food. Only the DEBQs restraint scale, which has 10 items, was utilized in the present study. The restraint scale was se lected to measure dieting beha vior because it is regarded as a preferred measure of dieting behavior a nd it has been used in several studies with children and adolescents to examine dieting behavior (Netemeyer & Williamson, 2001). The other two scales of the DEBQ were not used in the present st udy because they do not provide information about the subjects dieti ng behavior. The DEBQ yields raw scores and norm-referenced scaled scores for each of the 3 scales. The scale score was calculated by dividing the raw score by the num ber of items completed; the score is not
72 valid if more than one item is omitted. De scriptive categories for test score ranges include very high, high, above mean, mean, be low the mean, low, and very low. The entire DEBQ requires 10 minutes to complete and participants in this study required no more than five minutes to complete the restraint scale. Research supports the reliability and validity of the scores of the DEBQ restraint scale (Netemeyer & Williamson, 2001). The DEBQ is described as having excellent factorial validity, satisfactory to good reli ability, and satisfactory concurrent and discriminant validity (Thames Valley Test Company, 2004). Reliability for the DEBQ restraint scale ranged from .92 to .95 (Cronbach s alpha) in a sample of 1170 subjects. The DEBQ manual also reports strong factoria l validity for the restraint scale, with 68% of the variance among the items explained by a single factor when us ing exploratory and confirmatory factor analysis. The DEBQ was normed in the Netherla nds on samples consisting of over 2689 Dutch adults and adolescents including high sc hool females, female college students, female eating disorder patients, obese me n, obese women, non-obese men, and non-obese women. Additional normative statistics are provided for smaller subgroup samples, including 53 female obesity clinic patients 154 female subscribers of a ladies weekly magazine, 77 female athletes, 54 female da nce students, 104 high school boys, and 68 non-patient obese females. The fact th at the DEBQ was not normed on Americans presents a limitation for the study and effectiv ely calls into question the validity of the measurement tool for an American sample of adolescents, despite the evidence of validity for the original Dutch samples. However, it should be emphasized that, normative data issues notwithstanding, the DE BQ is regarded by researcher s as a preferred measure of
73 dieting behavior in the United St ates and it has been used in several studies with children and adolescents (Netemeyer & Williamson, 2001). In order to address the research questi on related to the orde r of precedence of dieting and binge-eating, prior to distribu ting the questionnaires one question was added on the DEBQ. On the DEBQ restraint scale, the question was added: At what age did you first go on a diet (i.e., deliber ately eat less in order to lo se weight)? ____ years. Note: Leave this question bla nk if you have never been on a diet. The wording for this question resembled the wording for eating rest raint items on the DEBQ and was intended to clearly and simply define diet for participants. Procedure The investigator for the present study obtained a pprovals from the school districts research office and from the Univ ersity of South Florid a Institutional Review Board (IRB) in January 2005. Recruitment After approvals were granted, the process to recruit participants began in both high schools in February and continued th rough May, 2005. The initial objective was to recruit approximately 200 students per school fo r a total of 400 partic ipants. A larger sample of 1000-5000 would provide more reli ability but would limit the practicality of the study, while a smaller sample of less th an 200 would increase the likelihood of nonnormality and reduce the ability to accurately calculate goodness of fit (West, Finch, & Curran, 1995). All participants for th e study who provided parent consent were accepted into the study, therefore constituting a convenience sample. The first round of data collection yielded 134 participants fr om High School 1 and 60 participants from
74 High School 2, for a total of 194 participan ts. According to the proposed study, if recruitment goals were not met then the study might be expanded to a 3rd school. However, by the time it became evident that the total number of participants would not meet the recruitment goal, it was too late in the school year to begi n the recruitment and data collection process at an additional schoo l. Rather than expand the study to a third high school in the fall, recruitment began ag ain at High School 2 in Fall, 2005. Because High School 2 represented the most demographi cally diverse school in the district, and High School 1 one of the least diverse, it was expected that th e number of minority participants could be maximized by recrui ting at High School 2. According to the Florida Department of Education (2006b), th e third high school under consideration was less diverse than High School 2 and the sc hool population in 20052006 was reported to be 84.6% White, 3.6% Black, 9% Hispanic, 0.7% Asian, 0.6% American Indian, and 1.5% Multiracial. This contrasted with de mographics at High School 2 of 64.1% White, 14.3% Black, 19.7% Hispanic, 0.5% Asian, 0.2% American Indian, and 1.3% Multiracial. Students signed up for the st udy at High School 2 from October through December 2005, and this second round of data collection yielded 58 new participants for an overall total of 252 students. None of the new particip ants for the second round had participated in the first round of data collection. Several recruitment strategies were empl oyed in each school. Due to different restrictions specified by the school princi pals, recruitment was school-wide at High School 1 but not at High School 2. Within th is framework, recruitment procedures were followed as similarly as possible between the two schools. Recruitment procedures used
75 exclusively for High School 1 will be discusse d first, followed by procedures used at High School 2, and lastly the procedures identical to both schools will be outlined. In one method of recruitment at High School 1, the morning news show ran a daily announcement for approximately four week s to request volunteers for the study. In addition, flyers were posted around High School 1 and on the c linic door advertising the study to students. Several te achers at this school agreed to make announcements about the study in their classes, incl uding the ROTC teacher and some of the special education teachers. The examiner also made announcements in other classrooms such as the performance-based diploma program and a sp ecial education classroom for mentally handicapped students. Students were invited to come to the clinic between classes at High School 1 to sign up for the study. The c linic assistant maintain ed the sign up sheets in a secure location and once a week ga ve them to the primary investigator. At High School 2, recruiting of participan ts was initially limited to Physical Education (P.E.) and Life Skills Management classes. The school nurse at High School 2, who served as the primary school cont act for the primary investigator, made announcements in these classes to recruit participants. In addition, teachers made announcements in the E.S.O.L. (English for Sp eakers of Other Languages) class and the EMH (Educable Mentally Handicapped) class. Sign-up sheets were distributed in each class by the school nurse or teac her, and students were invite d to sign up for the study at that time. The school nurse collected the si gn up sheets and then forwarded them to the primary investigator. Several recruitment procedures were iden tical at both schools. In both schools, P.E. and Life Skills Management teachers provided extra class credit towards their grade
76 to students for participating in the study. In addition, participants from both schools were informed that, if they participated in the study, they would be elig ible for a drawing to win one of many prizes donated by local busin esses. Businesses donated a total of 64 prizes for this study, which included movie passes, bowling passes, and gift certificates for restaurants, clothing stores, haircuts, and manicures. Letters used to solicit business donations were typed on school letterhead and approved by both school principals. Obtaining Consent At both schools, each student interest ed in participating in the study was instructed to write their st udent ID, name, address, and whether their parents spoke Spanish on the sign-up sheet. Teachers, clinic staff, and nurses forwarded the completed sign-up sheets to the principal i nvestigator on at least a weekly basis. Data entry staff supplied printed address labels for the student population, and the pr incipal investigator mailed a consent form home to each parent with a self-addressed stamped return envelope. A short note was included explaini ng that their son or daughter had signed up for the study. Parents signed and returned the consent form in the mail to the investigator. The consent form also include d a child assent form to be signed by the student at the school prior to data collection. If the student indicated his or her parents spoke Spanish, the consent form was sent to the parent in both Spanish and English. Students who were 18 years or older did not re quire parent consent but were required to sign the informed consent themselves. Pl ease refer to Appendix A for a copy of the English and Spanish parent consent forms. At both schools, many students signed up for the study but provided no parent consent. Teachers followed up on students missing consent by hand-distributing consent
77 forms to return directly to the school. The number of consent forms that were returned by hand was not tracked, but most forms were returned in response to the mailing. In addition, a small number of students returned their mailed consent forms directly to the school, rather than mailing them. In those cases, students handed the forms to their teacher or to clinic staff who in turn gave them to the primary investigator. At High School 1, a total of 219 students signed up and 134 (61%) provided parent or adult consent. At High School 2, a total of 348 students signed up and 118 (34%) provided consent. High School 1 had a higher return rate possibly due to the fact that the primary investigator provided school ps ychological services to this school, and thus had more regular and frequent contact with teachers, staff, and students. Survey Packet Organization The packets were pre-numbered consecutiv ely beginning with 1, with the packet numbers written on each questionnaire in th e packet. To counterbalance any potential order effects, the four questionnaires appear ed in a pre-selected, counterbalanced order for each package. The four questionnaires ha d a total of 24 possible orders, and packets one through 24 were each in a unique order. That unique order of questionnaires was then repeated for packets 25 through 48, and so on for the rest of the packets. In this manner, an unintentional effect of the orde r of the questionnaires such as unreliable results on the last questionnaire due to fa tigue of the participant, was minimized. Data Collection During data collection, participating stude nts were called to the school clinic at High School 1, where one office and the exam room were reserved for the study. Participants came first to the office which c ontained a round table and chairs. Similarly,
78 participants were called to a conference room at High School 2, occupied by a large conference room with chairs and the nurses m easuring station at the far end of the room for privacy. In both schools, students arrived in groups of five to ten at the beginning of a class period. As the students arrived, they were first directed to fill out the packet of questionnaires: the Nutrition/Physical Activ ity Questionnaire, the Reynolds Adolescent Depression Scale 2nd Edition (RADS-2), the Eating Diso rders Inventory 2, (EDI-2), and the Dutch Eating Behavior Questionnair e (DEBQ). The principal investigator remained in the room with the students as th ey filled out the questionnaires. A standard list of instructions was posted and explained to assist the students in filling them out. As each student completed the packet of questionnaires, he or she was instructed to proceed to the measuring station to have his or her height and weight measured and BMI calculated. Weighing, measuring, and BM I were typically conducted by the nurse at both schools. If the nurse was not avai lable at High School 1, the health assistant conducted the weighing and measuring and BM I calculation, and in a few cases at High School 1 when neither the nurse nor health assistant was available, the principal investigator completed the height, weight, and BMI calculation. If the nurse was not available at High School 2, the principal i nvestigator performed this role. When weighing and measuring were comp lete, the students BMI (kg/m2) was calculated using a BMI wheel (at High School 1) or a special BMI calculator (at High School 2). The calculator takes pounds and inches as input, converts to kilo grams and meters, and then displays the BMI. The BMI was th en written on the students packet. Following the weighing and measuring, each participant turned in his or her completed packet to the primary investigator at which time the student received a ticket
79 for the drawing. Tickets for the drawings consisted of two identical parts with the students name and ID written on each half; one half was given to the student and the other half placed in a box to be used during the drawing at the completion of that round of data collection for that school. At this poi nt the student was instructed to return to class. Overall, each student missed a porti on of one class period, and it took each student approximately 15-30 minutes to complete the fo rms and have their BMI calculated. This time requirement was consistent to expecta tions, based upon an early field test of the packet in which a student comp leted the packet in 9 minutes. Post Data Collection Procedures After students returned to class, the primary investigator hand scored the depression protocols the same day students completed them The school psychologist, social worker, and/or guidance counselors we re notified immediately, usually the same day, of any students whose scores fell in the at-risk range for depression. Standard procedures for students reported to have depressive symptoms were then followed whereby a school mental health professional ta lked with the student in confidence, and the students parents were notif ied if deemed appropriate by the professional. Generally, according to school procedures, if a counselor, social worker, or psychologist speaks with the student and the student pres ents as depressed and in n eed of mental health care, parents are notified and a referral for outside counseling is offere d by the school social worker. If the offer is declined by the famil y, the student is monitored on a weekly basis by staff, and with an understanding that the student can request to see the counselor or other mental health professional at school. It should be noted that the primary investigator was the school psychologist for High School 1, but not for High School 2.
80 Approximately 10% of participants scored a bove the cutoff for depr ession, and these data are further discussed in Chapter 4. The prizes involved in the study attracted attention from students and staff, and the final step regarding participant prizes was the drawing for thos e prizes. Drawings were held in each school after each round of data collection was complete: once at High School 1 and twice at High School 2. Prizes were awarded by randomly selecting tickets from the box until all prizes had been awarde d. Clinic staff in each school notified student winners and gave the prizes to the students. High School 1 participants received 38 prizes after the first round of data collection, and High School 2 participants won a total of 26 prizes including bot h rounds of data collection. In addition, participating clinic staff at both schools were provided with gift certificates. Tracking of Participants Several procedures were used in the tracki ng of participant packets. With regard to packet numbers, the packet number and the students name were recorded in a separate packet log during data colle ction at both schools. Packet s and packet numbers were distributed to students in numeri cal order as students arrived in the clinic, so that the first student participant received packet number one, the second received packet number two, and so on. The students names did not appear on the packet or que stionnaires, but the packet log identified what student complete d a particular packet of questionnaires. The primary investigator kept all p acket logs at a home office. The primary investigator used the pack et number on the questionnaires to link the questionnaires to a student name and ID on th e packet log, so that the remaining data could be accessed in student records for each participant. A shaded box entitled for
81 school use appeared at the top of the Nutrit ion/Physical Activity questionnaire with the labels BMI, SES, and ESE. The socioeconomic status was coded as Y or N at the top of the questi onnaire in the box labeled SES, depending on the free or reduced lunch status of the student. Likewise, if th e student was served in any special education program, the codes of the programs were writte n at the top of the page in the box labeled ESE. Data were recorded for each particip ant in an Excel file format which was then exported and used for analysis. Each student was identified in the data file by their packet number (1-252). Neither the students name or school ID was entered in the data file. Inter-rater Agreement Inter-rater agreement was assessed for 36 participants in a randomly selected, 14% subset of the sample. Inter-rater agreement was calcu lated for the measurement of BMI, including height, weight and calculation of BMI, a nd for the determination of socio-economic status and ESE status, the RADS-2 depression survey, and both eating behavior questionnaires, the ED I-2 and DEBQ. Packets were marked to identify those to be used for the inter-rater agreement calcu lation. Inter-rater agreement for BMI was calculated at High School 1 betw een the school nurse and hea lth assistant if both were available, or between one of the two a nd the principal investigator. At High School 2, inter-rater agreement for BMI was calculated between the school nurse and the primary investigator. BMIs were considered in disagreement if the height, weight, or BMI index were not the same for both indi viduals. Inter-rater agreement on socioeconomic status, special education status, and protocols for the RADS-2, EDI-2, and DEBQ were calculated between the prim ary investigator and a school psychology
82 practicum student, trained in psychoeducatio nal assessment. The student looked up the free and reduced lunch status and special education status on printouts provided by High School 2, and compared them to those recorded by the principal investigator. Similarly, the student scored the protocols for the RADS-2, EDI-2, and DEBQ and compared the scores to those of the princi pal investigator. Table 13 show s inter-rater agreement rates for each variable. For BMI, height and weight were measured and BMI was then Table 13 Inter-rater Agreement Variable Measured Percent BMI 94% Depression (RADS-2) 100% Dieting (DEBQ) 100% Bingeing (EDI-2) 97% Socio-economic Status 100% Exceptional Student Education Status 100% calculated, and the two BMIs were compared. For agreement on socio-economic status, the free/reduced lunch eligibility was compar ed. For ESE status, the ESE categories were compared. Finally, agreement on th e RADS-2, EDI-2, and DEBQ was based upon the standard scores calculated on the protocols. Inter-rate r agreement ranged from 94% 100% across the variables. In the few cas es of disagreement, the two individuals resolved the disagreement and determined whic h rater was correct. In order to eliminate the disagreement, in the case of BMI, the stud ents height and weight were checked again
83 by both individuals; in the cas e of the Eating Disorders I nventory-2, the individuals reviewed the scoring protocol.
84 Chapter IV: Results Data Analysis Data analysis techniques employed to an swer the studys three research questions included descriptive statistics, structural equation modeling, and correlation analysis. The Statistical Analysis Software (SAS) soft ware package was utilized to perform all analyses. The following discussion outlines data analysis results for each research question. Question 1: To what degree do the data suppor t the proposed model of the correlates of weight in adolescents? A descriptive analysis was done first on the data in order to assess normality, possible outliers, and the appropriateness of doing the path analysis. Descriptive statistics on the variables include a summary of the data and a correlation matrix for each variable in the model overall and for both sc hools. The reader may also refer to the description of partic ipants in Chapter 3 for summaries of data on gender and school, ethnicity and gender, grade, socio-economic status, and special ed ucation status. The distributions of the following continuous va riables are now described: BMI, age, physical activity, depressive symptoms, dieti ng, binge eating symptoms, age at first diet and first binge, have you ever been on a diet, and have you ever binged. As shown in Table 14, the mean BMI ove rall for participants was 24.2, which falls at the high end of the normal range. Acco rding to the descriptive statistics in Table 14, the mean BMI was slightly higher at High School 2 (M=24.5) than High School 1
85 Table 14 Distribution of BMI Variable Name Mean Median Mode Min Max Std. Dev. Skewness Kurtosis BMI (N = 250*) 24.2 23.0 23.0 16.0 50.4 5.5 1.7** 4.4** Age 14 23.0 22.0 20.0 17.0 38.1 5.0 1.5 2.4 Age 15 23.2 22.0 22.0 16.0 44.0 5.0 1.7 4.8 Age 16 24.2 23.0 21.0 17.0 37.0 4.4 0.9 0.1 Age 17 26.0 24.5 21.0 18.0 50.4 7.5 1.9 3.8 Age 18 24.6 23.0 23.0 18.0 47.0 5.4 2.0 6.0 Age 19 26.6 26.9 ----20.5 32.0 5.0 -0.3 -1.6 High School 1 24.0 23.0 21.0 16.0 50.4 4.9 1.7 5.8 Age 14 22.9 21.8 20.0 13.0 38.1 5.0 1.8 4.2 Age 15 23.1 23.0 22.0 16.0 32.0 3.5 0.3 0.5 Age 16 25.3 24.3 22.0 18.5 37.0 4.9 0.7 -0.1 Age 17 25.0 23.3 21.0 18.0 50.4 7.5 2.3 6.5 Age 18 23.1 21.5 20.0 19.0 31.0 3.8 1.0 -0.4 Age 19 27.2 29.0 -20.5 32.0 6.0 -1.3 -High School 2 24.5 23.0 23.0 17.0 48.0 6.1 1.7 3.3 Age 14 23.0 22.0 18.0 17.0 37.0 5.2 1.4 2.0 Age 15 23.2 21.0 19.0 17.0 44.0 6.5 1.9 3.5 Age 16 23.5 22.8 21.0 17.0 33.0 4.0 0.9 0.2 Age 17 27.3 25.0 23.0 20.0 48.0 7.5 1.7 3.0 Age 18 26.4 25.0 23.0 18.0 47.0 6.5 1.8 4.7 Age 19 24.7 24.7 24.7 24.7 24.7 ---*note: missing BMI data for 1 participant ** note: distribution has moderate to ex treme positive skew, leptokurtic distribution
86 (M=24.0). The mean BMI was highest for 19-ye ar olds (M=26.6) and lowest for 14-year olds (M=23.0). The BMI data distribution showed a positive skew. Table 15 illustrates the prevalence of obesity, overw eight, normal weight, and under weight in the data. The weight categories are based upon thos e used for the 1999-2000 NHANES 1999-2000 study: the body mass index-fo r-age percentiles developed by the National Center for Health Statistics, published in 2000 and curren tly recommended for use with children and adolescents by the Centers for Disease Control. As described in Chapter 2, these criteria define obesity as a BMI above the 95th percentile for age and sex, however, it is important to understand that the cutoffs for each weight category have remained the same since the 2000 publishing date. The present da ta show a higher prevalence of obesity in adolescents than reported in the 2000 NHANE S study, at a rate of 19% compared to the earlier NHANES rate of 15%. Table 15 also displays weight categories for the three ethnicities with the largest sample sizes: Caucasian (N=198), Latino-Hispanic (N=21), and African-American (N=11). Table 16 shows the data distribution for ag e, with a mean participant age of 16 years overall and for both schools and particip ants ranging from 14 to 21 years. Table 17 defines the ages of the participants in the study. Table 18 lists the distribution of physical activity, or the sum of the first two physical activity answers on th e DPAQ (how many of the last 7 days did you participate in exercise that either did or did not make you sweat and breathe hard?). The mean total for the data set was 6.3 days, with High School 2 showing a slightly lower mean than High Sc hool 1, 6.1 days vs. 6.5 days. It should be noted that the responses for the two questions could have referred to the same days of the week, or to different days of the week. Consequently the total number of days for a
87 Table 15 Weight Category Distribution N **Obese (%) **Overweight (% ) **Normal (%) **Underweight (%) *Total Participants Ages 14-19 247 46 (19%) 50 (20%) 149 (60%) 3 (1%) Males 108 22 (20%) 21 (19%) 64 (59%) 2 (2%) Females 139 24 (17%) 29 (21%) 85 (61%) 1 (1%) High School 1 130 24 (18%) 29 (22%) 75 (58%) 3 (2%) High School 2 117 22 (19%) 21 (18%) 74 (63%) 0 Age 14 35 6 (17%) 8 (23%) 21 (60%) 0 High School 1 18 3 4 11 0 High School 2 17 3 4 10 0 Age 15 71 11 (15%) 16 (23%) 42 (59% 2 (3%) High School 1 41 6 12 21 2 High School 2 30 5 4 21 0 Age 16 58 12 (21%) 11 (19%) 35 (60%) 0 High School 1 24 7 5 12 0 High School 2 34 5 6 23 0 Age 17 36 9 (24%) 6 (16%) 20 (57%) 1 (3%) High School 1 21 4 4 12 1 High School 2 15 5 2 8 0 Age 18 42 6 (16%) 8 (19%) 28 (65%) 0 High School 1 24 3 3 18 0 High School 2 18 3 5 10 0 Age 19 4 1 (25%) 1 (25%) 2 (50%) 0 High School 1 3 1 1 1 0 High School 2 1 0 0 1 0 note: BMI data missing for 1 particip ant, Age data missing for 1 participant ** note: Weight category definitions base d on body mass index-for-age pe rcentiles developed by Nationa l Center for Health Stat istics in collaboration with the National Center for Chr onic Disease Prevention and Health Promotion (2000). http://www.cdc.gov/growthcharts
88 Table 15 (Continued) Weight Category Distribution N Obese (%) Overweight (%) Normal (%) Underweight (%) AA=AfricanAmerican*** 11 4 (36%) 0 7 (64%) 0 Female 6 4 (67%) 0 2 (33%) 0 Male 5 0 0 5 (100%) 0 CA=Caucasian*** 198 36 (18%) 39 (20%) 121 (61%) 2 (1%) Female 102 15 (15%) 21 (21%) 66 (65%) 0 Male 96 21 (22%) 18 (19%) 55 (57%) 2 (2%) LH=LatinoHispanic*** 21 3 (14%) 9 (43%) 9 (43%) 0 Female 17 3 (18%) 6 (35%) 8 (47%) 0 Male 4 0 3 (75%) 1 (25%) 0 *** note: Weight category data not shown for the following et hnicities: AS=Asian, NA=Native American, OT=Other, AA+CA+LH, AA+LH, AA+OT, CA+LH, CA+NA Table 16 Distribution of Age by School Variable Name Mean Median Mode Min Max Std. Dev. Skewness Kurtosis Age (N=250)* 16.0 16.0 15.0 14 21 1.4 0.4 -0.4** High School 1 16.0 16.0 15.0 14 21 1.5 0.5 -0.4 High School 2 16.0 16.0 16.0 14 20 1.4 0.4 -0.5 *note: missing age data for 1 participant **note: distribution considered to be fairly normal
89 Table 17 Distribution of Age Age N % Total Participants 250 100% Age 14 35 14% Age 15 71 28% Age 16 58 23% Age 17 37 15% Age 18 43 17% Age 19 4 2% Age 20 1 .4% Age 21 1 .4% note: missing age data for 1 participant participant did not necessarily correspond with ac tual days of the week on which exercise took place, but rather an arit hmetic sum of the two responses. Fourteen-year olds had the highest mean of 7 days, and 19-year olds had the lowest mean of 3.5 days. The distribution of the physical activity variable was fairly norma l. Table 18 also contains the distribution of physical activity by school for each age level. Table 19 illustrates the dist ribution of the depression va riable, as represented by T-Scores obtained on the RADS-2 screener. The depression data had a slight positive skew, with a mean T-Score of 47.4 which fe ll within the normal range and below the mean of 50. The mean depression T-Scores for the two schools were almost identical. Girls (M=47.8) had a slightly higher mean T-Score than boys (M=46.8). As shown in Table 20, 10% of all particip ants scored above the cutoff on the RADS-2 and were referred to a counselor, social worker, or psychologist. High Schools 1 and 2 had very similar referral rates, with 9.8% of High Sc hool 1 participants and 10.2% of High School
90 Table 18 Distribution of P hysical Activity Variable Name Mean Median Mode Min Max Std. Dev. Skewness Kurtosis Total Physical Activity (N=251) 6.3 6 7 0 14 3.7 0.3 -0.4* High School 1 6.5 7.0 7.0 0 14 3.8 0.2 -0.5 High School 2 6.1 6.0 7.0 0 14 3.7 0.4 -0.2 Age 14 7.0 7.0 7.0 1 14 3.5 0.2 -0.5 Age 15 6.4 6.0 7.0 0 14 3.3 0.4 0.2 Age 16 6.9 7.0 7.0 0 14 4.2 0.3 -0.8 Age 17 6.1 7.0 7.0 0 14 3.8 -0.2 -0.6 Age 18 5.1 5.0 2.0 0 14 3.8 0.8 0.2 Age 19 3.5 2.5 2.0 2 7 2.4 1.8 3.1 Age 20 10.0 10.0 10.0 10 10 ------------Age 21 8.0 8.0 8.0 8.0 8.0 ------------High School 1** 6.5 7.0 7.0 0 14 3.8 0.2 -0.5 Age 14 7.1 7.5 4.0 1 13 3.6 -0.1 -0.7 Age 15 6.5 7.0 7.0 0 14 3.0 0.2 0.3 Age 16 7.1 6.5 14 0 14 4.8 0.2 -1.1 Age 17 6.6 7.0 7.0 0 14 4.0 -0.2 -0.5 Age 18 5.6 5.0 5.0 0 14 3.9 0.7 -0.02 High School 2** 6.1 6.0 7.0 0 14 3.7 0.4 -0.2 Age 14 6.8 7.0 7.0 1 14 3.5 0.4 0.1 Age 15 6.3 6.0 4.0 0 14 3.7 0.6 0.2 Age 16 6.7 7.0 7.0 1 14 3.9 0.5 -0.4 Age 17 5.6 6.5 7.0 0 11 3.6 -0.5 -0.8 Age 18 4.5 4.0 2.0 0 14 3.6 0.9 1.0 *note: Distribution considered to be fairly normal **note: Age data not broken down by sc hool for ages 19-21 due to small number of participants ages 19-21
91 Table 18 (continued) Distribution of P hysical Activity Variable Name Mean Median Mode Min Max Std. Dev. Skewness Kurtosis High School 1 6.5 7.0 7.0 0 14 3.8 0.2 -0.5 Grade 9 6.8 7 7 0 14 3.2 0.2 -0.1 Grade 10 7.3 7 7 0 14 4.3 0.9 -0.8 Grade 11 6.3 7 4 0 14 4.1 -0.1 -0.8 Grade 12 5.4 5 2 0 14 3.8 0.6 0.1 High School 2 6.1 6.0 7.0 0 14 3.7 0.4 -0.2 Grade 9 6.7 6 7 0 14 3.3 0.8 0.3 Grade 10 6.3 6.5 7 0 14 4.3 0.4 -0.6 Grade 11 5.7 6.5 7 0 14 3.8 0.3 -0.3 Grade 12 5.4 6 2 0 14 3.4 0.4 -0.1 2 participants referred to school mental health service providers. Girls were somewhat more likely to score above the cutoff than boys as indicated by the 11% referral rate for girls compared to 9% for boys. Interestingly, the rate of students scoring above the cutoff (86th percentile) was lower than that in the standardization sample of the RADS-2. The distribution of dieting responses appears in Table 21, and these data were positively skewed. The overall mean scaled score of 2.2 fell within th e Mean range on the Restrained Eating norms for Dutch high sc hool females. The DEBQ manual lists no norms for high school males. Participants at the two high schools had similar mean scaled scores on the DEBQ for dieting behavi or, and both fell within the Mean range. Table 22 shows the distribution of binge eati ng data on the EDI-2 bulimia scale, which
92 Table 19 Distribution of Depressi on: RADS-2 T-Score Variable Name Mean Median Mode Min Max Std. Dev. Skewness Kurtosis RADS2 T-Score (N=251) 47.4 46 38 31 74 9.4 0.5* -0.3* High School 1 47.3 46 45 31 74 9.4 0.6 -0.2 High School 2 47.4 47 38 31 74 9.4 0.4 -0.4 Female 47.8 47 51 31 74 9.6 0.3 -0.4 Male 46.8 45 45 31 74 9.1 0.7 0.1 Age 14 46.4 45 36 32 64 9.1 0.5 -0.6 Age 15 46.4 45 36 31 74 10.0 0.6 -0.2 Age 16 47.7 47.5 38 31 67 8.7 0 -0.9 Age 17 50.3 50 43 33 73 9.8 0.4 -0.3 Age 18 45.7 44 44 31 74 8.7 1.2 1.8 Age 19 51.8 50.5 49 49 57 3.8 1.3 0.8 Age 20 46.0 46.0 46.0 46 46 ------------Age 21 53.0 53.0 53.0 53.0 53.0 ------------High School 1 47.3 46 45 31 74 9.4 0.6 -0.2 Grade 9 46.6 46 46 31 68 9.3 0.3 -0.7 Grade 10 48.1 45 45 33 69 10 0.5 -0.6 Grade 11 48.1 45.5 38 36 73 9.8 1.1 1.0 Grade 12 47 46.5 39 34 74 9 .9 1.3 High School 2 47.4 47 38 31 74 9.4 0.4 -0.4 Grade 9 45.7 44 36 32 74 9.5 1.1 1 Grade 10 47.8 49 51 31 66 10.1 0 -1 Grade 11 49.8 50 50 37 67 7.6 0.3 -0.1 Grade 12 47.3 45 41 31 69 10 0.6 -0.2 *note: distribution has Slight positive skew
93 Table 20 Depression Scores Above Cutoff Total Participants (N) Above Cutoff (N) Above Cutoff (%) Total Participants 251 25 10% High School 1 133 13 9.8% High School 2 118 12 10.2% Female 141 16 11% Male 110 10 9% also was positively skewed. As discussed in Chapter 2, in the present study EDI-2 data are reported using raw scores due to the exclusion of the one purging item in the subscale, so the percentiles in the manual fo r high school students are inappropriate. Table 23 shows the distribution of responses fo r Age at First Diet and Age at First Binge, completed by participants who replied yes to the questions, Have you ever experienced a binge eating epis ode? and, Have you ever been on a diet? Participants responded how many years old they were at those times. St udents indicated they first dieted as young as 9 years old, and first bi nged as young as 10 years old, and the median ages for dieting and bingeing were 14.5 a nd 14.0, respectively. The mean age for first binge was 9 months earlier at High School 2 (M=13 years, 5 months) than High School 1 (M=14 years, 2 months), and the mean age fo r first diet was 7 months earlier at High School 1 (M=14 years, 1 month) than Hi gh School 2 (M=14 years, 8 months). According to Table 24, 33% of participants responded that they had been on a diet, and Table 25 illustrates that 12% responded they had experienced a bingeing episode. Table 26 further illustrates that 6% re ported they had both binged and dieted at some time in the
94 Table 21 Distribution of Dieting: DEBQ Scaled Score Variable Name Mean Median Mode Min Max Std. Dev. Skewness Kurtosis DEBQ Scaled Score (N=250)* 2.2 1.95 1.0 1.0 14.0 1.3 4.0** 29.6** High School 1 2.3 2.2 1.0 1.0 14.0 1.4 4.5 35.3 High School 2 2.1 1.9 1.0 1.0 10.0 1.2 2.9 15.4 Age 14 2.5 2.2 2.9 1.0 10.0 1.6 3.1 13.9 Age 15 2.1 1.8 1.0 1.0 14.0 1.6 5.6 40.0 Age 16 2.3 2.2 1.0 1.0 5.0 1.1 0.7 -0.4 Age 17 2.3 2.1 1.0 1.0 5.0 1.0 0.8 -0.1 Age 18 1.9 1.7 1.0 1.0 4.1 0.9 0.7 -0.5 Age 19 1.9 1.6 ---1.0 3.4 1.1 1.0 -0.4 Age 20 3.3 3.3 3.3 3.3 3.3 ------------Age 21 3.1 3.1 3.1 3.1 3.1 ------------High School 1 2.3 2.2 1 1 14 1.4 4.5 35.3 Grade 9 2.4 2.2 1 1 14 1.8 4.9 30.7 Grade 10 2.4 2.6 3 1 4.6 1.1 0.5 -0.5 Grade 11 2.2 2 1 1 4.6 1.1 0.7 -0.6 Grade 12 2.1 2.2 1 1 4.1 0.9 0.4 -1.1 High School 2 2.1 1.9 1 1 10 1.2 2.9 15.4 Grade 9 2.2 1.8 1 1 10 1.6 3.9 19.1 Grade 10 2.2 1.9 1.2 1 5 1.1 0.8 -0.1 Grade 11 2.0 1.9 1.2 1 5 1 1.6 3.3 Grade 12 2.1 1.9 1 1 4.1 1 0.7 -0.4 *note: missing data for one DEBQ score **note: distribution was leptokurtic with extreme positive skew
95 Table 22 Distribution of Binge Eati ng: EDI2 Raw Score Variable Name Mean Median Mode Min Max Std. Dev. Skewness Kurtosis EDI2 Raw Score (N=251) 1.0 0.0 0.0 0 18 2.2 3.6* 18.6* High School 1 1.1 0 0 1 10 1.9 2.1 4.5 High School 2 1.0 0 0 0 18 2.4 4.4 24.4 Age 14 1.6 1 0 0 6 2.1 1.0 -0.7 Age 15 0.8 0 0 0 10 1.9 3.0 10.0 Age 16 0.6 0 0 0 3 1.0 1.7 1.2 Age 17 1.5 0 0 0 18 3.5 3.6 14.3 Age 18 1.0 0 0 0 11 2.2 3.1 11.3 Age 19 2.0 2 ---0 4 1.8 0 -3.3 Age 20 0 0 0 0 0 ------------Age 21 6.0 6 6 6 6 ------------High School 1 1.1 0 0 1 10 1.9 2.1 4.5 Grade 9 1.5 0 0 0 10 2.3 1.7 2.8 Grade 10 0.5 0 0 0 3 1.0 1.6 1.3 Grade 11 1 0 0 0 6 1.7 2 4.1 Grade 12 1.1 0 0 0 7 1.9 1.9 2.9 High School 2 1.0 0 0 0 18 2.4 4.4 24.4 Grade 9 0.8 0 0 0 6 1.7 2.1 3.3 Grade 10 0.8 0 0 0 11 2.1 4 18.5 Grade 11 1.3 0 0 0 18 3.7 4.3 19.8 Grade 12 1 0 0 0 11 2.3 3.5 14.1 *Note: distribution is leptokurtic with extreme positive skew
96 Table 23 Distribution of Age at 1st Diet and Age at 1st Binge Variable Name Mean Median Mode Min Max Std. Dev. Skewness Kurtosis AGE at 1st Diet* (N=84) 14.4 14.5 15.0 9 18 1.6 -0.5 0.8 High School 1 14.1 14.0 13.0 9 18 1.8 -0.3 0.8 High School 2 14.7 15.0 15.0 12 17 1.3 -0.3 -0.3 AGE at 1st Binge** (N=26) 13.9 14.0 14.0 10 18 1.9 -0.2 0.1 High School 1 14.2 15.0 15.0 10 18 2.2 -0.7 0.3 High School 2 13.4 13.0 13.0 11.5 6 1.4 0.5 -0.2 note: Age at 1st Diet distribution showed slight negative skew, ** note: Age at 1st Binge distribution was fairly normal Table 24 Distribution of Have You Ever Been on a Diet Variable Reply N Percentage *N 250 100% No 167 67% Yes 83 33% *High School 1: 132 No (65%) 86 Yes (35%) 46 High School 2: 118 No (69%) 81 Yes (31%) 37 *Note: missing data for 1 participant
97 Table 25 Distribution of Have You Ever Binged Variable Reply N Percentage *N 250 100% No 221 88% Yes 29 12% High School 1: 133 No (86%) 115 Yes (14%) 18 *High School 2: 117 No (91%) 106 Yes (9%) 11 *Note: missing data for 1 participant past. The bingeing and dieting data will be fu rther discussed with the results of research question number 3. Correlation matrices were cr eated to determine the corr elation of each continuous variable with other continuous variables: BMI, physical activity, depressive symptoms, dieting behavior, and binge eating behavior. Table 27 shows the correlation matrix of variables across the entire data set, and Tabl es 28 and 29 show the correlation matrix for Schools 1 and 2, respectively. For the aggregated data, the analysis re vealed statistically significant correlations between depression and BMI (r = .25, p = .0001), between dieting and BMI (r = .21, p = .0007), between depres sion and dieting (r = .20, p = .0017), and between depression and binge eating (r = .14, p = .0228). The clinical significance of these correlations could be view ed as minimal, however. For example, the correlation of .25 between depression and BMI would explain about 6% (R2 = .0625) of the variance of
98 Table 26 Distribution of Ever Di eted and Ever Binged Reply N Percentage N* 249 100% Have Neither Binged nor Dieted 152 61% Have Dieted but not Binged 68 27% Have Binged but not Dieted 14 6% Have Dieted and Binged 15 6% *note: missing Ever-Dieted data for one partic ipant, Ever-Binged data for one participant the two variables. The correlation matrices for the two schools re vealed statistically similar relationships between the variab les at the two locations, providing some justification for continuing with the path analysis of the aggreg ated data set. A Boxs M Table 27 Correlation Matrix for Entire Data Set BMI Physical Activity Depressive Symptoms Dieting Behavior Binge Eating Behavior BMI (N=250) 1.0 .02 .25 p = .0001 .21 p = .0007 -.01 Physical Activity (N=251) 1.0 -.08 .06 -.11 Depressive Symptoms (N=251) 1.0 .20 p = .0017 .14 p = .0228 Dieting Behavior (N=250) 1.0 .07 Binge Eating Behavior (N=251) 1.0
99 analysis of the two samples, using an alpha level of .05, did show that the populations of the two schools were not identical, chi square (15) = 29.7, p=.0132. However, in a test of homogeneity of within covarian ce matrices, the natural logs of the determinants of the sample covariance matrices suggested that the differences between the two schools was small (High School 1 natural log = 1 1.95687; High School 2 natural log = 12.66374). Appropriateness of doing the path analysis was determined by examining whether the assumptions required for path analysis were met by the data. Assumptions for a path analysis include a) normality of variable distributions, as measured by kurtosis (g1), skewness (g2), and outliers in the distribution; b) independence, as determined by comparing demographic characteristics of th e sample to the schools (e.g., race, gender, age); and c) minimal measurement error. With regard to independence, the sample closely resembled the school populations al ong demographic variables. The small amount of missing data also supported th e independence assumption. Reliability estimates addressed the measurement error assumption and will be discussed in more detail at the end of this chapter. Sa tisfaction of the firs t assumption of normal distribution was more difficult since the dieting and bingeing variable distributions were extremely positively skewed. Non-normal variable distributions are known to produce inflated chi-squared values and an increased pos sibility of a rejection of a true model, and this tendency increases with the extent of the non-normality. Only the dieting and bingeing variables had clearly non-normal dist ributions, however the chi-square results should be viewed with some level of caution. Once it was deemed appropriate to pro ceed with the path analysis, maximum likelihood estimation was used to estimate the model parameters including path
100 Table 28 Correlation Matrix for High School 1 BMI Physical Activity Depressive Symptoms Dieting Behavior Binge Eating Behavior BMI (N=132) 1.0 .05 .39 .00 .13 Physical Activity (N=133) 1.0 -.06 -.02 -.34 Depressive Symptoms (N=133) 1.0 -.03 .04 Dieting Behavior (N=132) 1.0 .22 Binge Eating Behavior (N=133) 1.0 Table 29 Correlation Matrix for High School 2 BMI Physical Activity Depressive Symptoms Dieting Behavior Binge Eating Behavior BMI 1.0 -.03 .15 .28 -.05 Physical Activity 1.0 .00 .06 -.10 Depressive Symptoms 1.0 .25 p = .0074 .10 Dieting Behavior 1.0 .03 Binge Eating Behavior 1.0 Note: N= 118 for all variables
101 coefficients and variance estimates. The chi-square (X2) statistic is a test of the null hypothesis that the model fits the data. Th e likelihood ratio effec tively tests the null hypothesis for goodness of fit of the model, and statistical significance was determined at the p=.05 level. Three alternative fit indices, the RMSEA Estimate, Bentlers Comparative Fit Index (CFI), and McDonalds Centrality Index, also were considered in the path analysis. Figure 5 shows the resu lts of the path analysis, including path coefficients or parameter estimates for each pa ir of variables in the equation. Table 30 shows the unstandardized parameter estimat es and standard e rror for each path coefficient, as well as the va riance and standard error of ea ch exogenous variable, and the covariances among each pair of exogenous variab les that had a relationship in MODEL2. Upon the initial path analysis in SAS, it became necessary to modify the model to remove one of the relationships between th e binge eating and dieting variables. The complexity of a two-directional relationshi p between the variables created impossible negative R squared values in the output. Consequently, as reflected in Figure 5, the weaker of the two relationships was re moved to remedy this problem in SAS. The chi-square results in Figure 5 (chi-square(19) =34.1123, p=.0178, alpha = .05) indicate that the null hypothesis should be rejected a nd that the model does not fit the data. Because the RMSEA Estimate should be < .05 to be considered indicative of a good fit of the data, the RMSEA Estimate of .0588 also failed to provide support for the model. Bentlers CFI should be larger than .9 in order to i ndicate a good fit of the model, and since Bentlers CFI was .7533 for the present model it offered no support. McDonalds Centrality (.9678) provides the only support for goodne ss of fit, with a value
102 Figure 5. Path Analysis Results: MODEL2. greater than .9. In summary, the path analysis suggests a poor fit of the model to the data, with only one of the four indices pr oviding support for the null hypothesis. Figure 5 also shows the R2 values for each endogenous variable, which indicate the percentage of variance in that variable accounted for by the variables that directly affect it. The R2 values suggest 6% of the varian ce in BMI is accounted for by the variables affecting it in the model. Similarly, the R2 values for physical activity (R2=.0508), depressive symptoms (R2=.0129), binge eating symptoms (R2=.0199), and dieting (R2=.0113) indicate the percentage of vari ance of those variab les explained by the variables in the model. Depressive Symptoms Binge Eating Symptoms Physical Activity Socio-Economic Status (1=Free/Red 0=Not Free/Red) Obesity/Weight Special Education Status (1=ESE, 0=Not ESE) Dieting Ethnicity (AA) Gender (1=Male 0=Female MODEL2 Results Ethnicity (LH) .04277 .17825 1076 1468 .01649 -.09571 ..23433.06683 -.0107 .1338 -.1356 -.0152 .1841 -.0861 .0393 -.0603 .0366 .1398 .0332 .0301 -.1029 .0288 R 2 = .0508 R 2 = .0129 R 2 = .0199 R 2 = .0621 Fit Indices: Ch-Square = 34.1123, p = .0178, df = 19. RMSEA Estimate = .0588 Bentlers CFI = .7533 McDonalds Centrality = .9678 R 2 = .0113
103 Table 30 MODEL2 Unstandardized Parameter Estimates and Standard Errors Parameter Description Unstandardized Parameter Estimate Standard Error Path Coefficients BMI from Physical Activity .0583 .0961 BMI from Depression .0216 .0383 BMI from Bingeing .0775 .1663 BMI from Low SES 1.6255 .7461 BMI from In ESE 1.4257 .8691 BMI from African American 3.3150 1.6065 BMI from Latino-Hispanic -.2045 1.2749 Physical Activity from Male 1.3712 .4851 Physical Activity from African American -2.2644 1.0754 Physical Activity from Latino Hispanic .1957 .8415 Depression from Physical Activity -.2160 .1670 Depression from Male -1.1263 1.2432 Dieting from Male -.2483 .1593 Bingeing from Depression .0320 .0150 Bingeing from Dieting .0035 .1157 Variances of Exogenous Variables Low SES .24766 .02309 In ESE .17278 .01611 Male .24860 .02248 African American .04946 .00461 Latino Hispanic .08300 .00753
104 Table 30 (continued) MODEL2 Unstandardized Parameter Estimates and Standard Errors Parameter Description Unstandardized Parameter Estimate Standard Error Covariances Among Exogenous Variables Low SES with In ESE .03687 .01384 In ESE with Male .00887 .01331 Low SES with African American .0074 .00731 In ESE with African American .00152 .00609 Low SES with Latino Hispanic .03360 .00959 In ESE with Latino Hispanic -.01146 .00792 African American with Latino Hispanic -.00474 .00418 The path analysis also produced standardized path coefficients for each direct effect of a variable on another variable. Pa th coefficients quantify the amount of change in a dependent variable that corresponds with a one-unit change in the independent variable. The standardized pa th coefficients in Figure 5 a ppear above or next to the arrows between variables. The path coe fficients range in magnitude from .0113 to .23433. However, given the poor fit of the mode l to the data, the path coefficients have limited meaning. Overall, the results of the path analysis suggested that the model in Figure 5 was a poor fit for the data. When it was determined that the model fit th e data poorly, an additional path analysis was conducted to determine whether a modified path model might have a better fit with the data. Rather than utilizing statistical me thods to identify potential causal paths to add or drop, two relationships were added to the model based on theory and research. Figure
105 6 shows the results of the final path analys is, with two relationships added to reflect socio-economic status directly influencing physical activity and depressive symptoms. Table 31 shows the unstandardized parameter es timates and standard error for each path coefficient, as well as the va riance and standard error of ea ch exogenous variable, and the covariances among each pair of exogenous variab les that had a relationship in MODEL3. Once again, the chi-square statistic (chi-square(17) =32.0018, p=.015, alpha = .05) indicated that the null hypothesis should be re jected and that the model did not fit the data. The RMSEA Estimate (.0619), Bentler s CFI (.7551), and McDonalds Centrality (.9681) also provided little s upport for goodness of fit, leading to a similar conclusion that the model in Figure 6 was a poor fit for the data. Question 2: For the populations of obese and non-obese adolescents, a) what is the rate of binge-eating symptoms, and b) what is th e strength of the relationship between bingeeating and depression? Table 32 provides one perspective on th e first component of this research question. In order to show a definitive rate of binge eating for obese and non-obese students, the Yes/No responses to the que stion, Have you ever experienced a binge eating episode? were compiled using the SAS software. The bulimia subscale of the EDI-2 was not used to determine the rate of binge-eating because no cut-off score defines whether the respondent has bi nged. As shown in Table 32, 27% of obese participants replied that they had experienced at least one binge-eating episode, compared to 8% of non-obese participants. A chi-square anal ysis indicated that this difference was statistically significant (chi-square(1)=13.1, p=.0003).
106 Figure 6. Path Analysis Results: MODEL3. In addition, as another comparison of binge -eating behavior in obese and non-obese participants, an independent T-test was conducte d to determine if there was a statistically significant difference in EDI-2 binge-eating sc ores between the two groups. A pooled ttest was used since the variances of the bi nge-eating variable were equal for the two groups (F=1.0, p=1.0). According to results of the pooled t-test [t(248)= -0.03, p=.97], the obese students level of bingein g on the EDI-2 bingeing scale (M=1.0612, SD=2.1789) was not statistically different from that of non-obese students (M=1.0498, SD=2.1789). Depressive Symptoms Binge Eating Symptoms Physical Activity Socio-Economic Status (1=Free/Red 0=Not Free/Red) Obesity/Weight Special Education Status (1=ESE, 0=Not ESE) Dieting Ethnicity (AA) Gender (1=Male 0=Female MODEL3 Results Ethnicity (LH) 04277 17825 1076 1469 .01649 -.09571 .23433 .06683 -.0107 .1339 -.1281 .0375 .1893 -.0828 .0393 -.0616 .0366 .1398 .0332 .0301 -.1029 .0288 R 2 = .0582 R 2 = .0142 R 2 = .0199 R 2 = .0616 R 2 = .0113 -.0893 .0359 Fit Indices: Chi-Square = 32.0018, p = .015, df = 17. RMSEA Estimate = .0619 Bentlers CFI = .7551 McDonalds Centrality = .9681
107 Table 31 MODEL3 Unstandardized Parameter Estimates and Standard Errors Parameter Description Unstandardized Parameter Estimate Standard Error Path Coefficients BMI from Physical Activity .0583 .0963 BMI from Depression .0216 .0383 BMI from Bingeing .0775 .1663 BMI from Low SES 1.6255 .7485 BMI from In ESE 1.4257 .8691 BMI from African American 3.3150 1.6050 BMI from Latino-Hispanic -.2045 1.2747 Physical Activity from Low SES -.6661 .4938 Physical Activity from Male 1.4099 .4840 Physical Activity from African American -2.1392 1.0752 Physical Activity from Latino Hispanic .4835 .8649 Depression from Physical Activity -.2077 .1675 Depression from Low SES .6718 1.2300 Depression from Male -1.1500 1.2432 Dieting from Bingeing .0163 .0372 Dieting from Male -.2499 .1593 Bingeing from Depression .0321 .0150 Bingeing from Male .1424 .2804 Variances of Exogenous Variables Low SES .24766 .02309 In ESE .17278 .01611 Male .2486 .02248 African American .04946 .00461 Latino Hispanic .08300 .00753
108 Table 31 (continued) MODEL3 Unstandardized Parameter Estimates and Standard Errors Parameter Description Unstandardized Parameter Estimate Standard Error Covariances Among Exogenous Variables Low SES with In ESE .03687 .01384 In ESE with Male .00887 .01331 Low SES with African American .00740 .00731 In ESE with African American .00152 .00609 Low SES with Latino Hispanic .03360 .00959 In ESE with Latino Hispanic -.01146 .00792 African American with Latino Hispanic -.00474 .00418 Table 32 Binge-Eating History of Obese and Non-Obese students Obese Students (N=49) Non-Obese Students (N=200) Answered Y to Ever Binged Question (N=29) 13 (27%) 16 (8%) Answered N to Ever Binged Question (N=220) 36 (73%) 184 (92%)
109 The second component of the question wa s addressed using simple correlation coefficients (Pearsons r) between binge-e ating and depressive symptoms for four populations within the sample : obese binge-eaters, obe se non-bingers, non-obese bingers, and non-obese non-bingers. Obese gr oups were defined usi ng the cutoff scores for boys and girls for obesity from the 2000 CDC BMI-for-age growth charts. The RADS-2 Depression Total standard score was us ed to represent depressive symptoms for analysis. As seen in Table 33, confidence in tervals were calculated for all correlation coefficients. With alpha = .05, the data analysis shows a statistically significant relationship between depressive symptoms and binge-eating behavior for all participants (N=251, r=.14, p=.02) and also for non-obe se participants (N=201, r=.15, p=.04). Clinical significance of these correlations mu st be considered, however, with depressive symptoms explaining approximately 2% of th e variance in binge-eating for all students (R2=.0196) and for non-obese students (R2=.0225). These results are consistent with those obtained on the path analysis for the relationship between depression and bingeeating. Question 3: For the population of adol escents who engage in both binge-eating and dieting behavior, what is the order of preced ence of binge-eating and dieting (i.e., which occurs first, binge-eating or dieting)? The answer to the third research que stion required examining the data for participants that report a hist ory of both binge-eating and di eting behavior. Within this population, the age of onset of binge-eating was compared to the age of onset of dieting. These ages were provided on the EDI-2 and DEBQ answer sheets respectively. As illustrated in Table 26, 6% of all participan ts, or 15 students, responded that they had
110 Table 33 Correlation of Binge-Eating and Depressi ve Symptoms for Obese and Non-Obese Students N Pearson Correlation p-value 95% Confidence Intervals All Students 251 .14 p = .0228 .02 to .26 Obese Students 49 .17 p = 0.2413 -.12 to .43 Non-Obese Students 201 .15 p = 0.0391 .01 to .28 Obese Bingers 13 -.13 p = 0.6618 -0.63 to 0.45 Obese Non-Bingers 36 .29 p = 0.0859 -0.05 to .56 Non-Obese Bingers 16 .30 p = 0.2643 -.24 to .69 Non-Obese Non-Bingers 184 .10 p = 0.1682 -.04 to .24 statistically significant, alpha = .05 both dieted and binged. Two of those participants failed to answer the age when bingeing first occurred, leaving an N of 13 for analys is. Table 34 shows th at 23% of this group dieted before they binged, 54% binged before they dieted, and 23% dieted and binged for the first time at the same age. Exact 95% c onfidence intervals indicate that, for example, the true percentage of bingers and dieters in the population who dieted first lies between 19% and 75%, quite a wide range of possibility.
111 Reliability Estimates on Measurement Instruments Since reliability of questionnaires may change from one sample to another, the data analysis also included internal cons istency reliability estimates, in the form of Cronbachs alpha parameters, for the measuremen t instruments that have multiple items: the RADS-2 depression screener, the EDI-2 eating disorders inventory, and the DEBQ dieting scale. As shown in Table 35, the depr ession and dieting scales had the strongest Table 34 Precedence of Dieting and Bingeing N Percentage95% Confidence Intervals Dieted First 3 23% 19% 75% Binged First 7 54% 46% 95% Dieted and Binged at Same Age 3 23% 19% 75% Total Bingers and Dieters 13 100% 100% Table 35 Reliability Estimates for Measurement Instruments Measurement Instrument Standardized Reliability Estimate: Cronbach Coefficient Alpha *EDI-2 Bingeing 0.69 DEBQ Dieting 0.93 RADS-2 Depression 0.92 *note: Purging item 6 excluded from data analysis
112 reliability with standardized cronbachs al pha coefficients of .92 and .93. The EDI-2 bulimia scale had a coefficient alpha of .69, considered somewhat low. This scale had fewer items than the DEBQ or RADS-2, a nd only six out of the seven items were included in the calculation because the purgi ng item was excluded from data analysis. The DEBQ dieting measure had 10 items and the RADS-2 depression screener had 30 items, and their larger number of items most likely influenced their higher reliability.
113 Chapter V: Discussion The purpose of this study was to investig ate the relationships between obesity and other variables found to be associated with obesi ty in the literature. This chapter includes a discussion of each research question and the results of the data analysis. The reader also will find the implications of the data a nd the contribution of this study to the obesity literature. Following the implications disc ussion, assumptions and limitations of the study are reviewed. Finally this chapter provides suggestions rela ted to future research. Question 1: To what degree do the data suppor t the proposed model of the correlates of weight in adolescents? Results of the path analysis indicated a poor fit of the data to the proposed path model. Results for the first research que stion began with descriptive data for the variables in the path model including BMI, age, physical activity, depressive symptoms, dieting, binge-eating symptoms, ag e at first diet and first bi nge, and the questions Have you ever been on a diet , and, Have you ever experienced a binge-eating episode.... These descriptive data will be discussed first, followed by the results of the path analysis. The BMI data reflected an overall obesity rate of 19% for all participants, higher than the 15% obesity rate for adolescents published in the 1999-2000 NHANES data. If this rate can be generalized to the U.S. population, it would s uggest that the obesity rate has continued to rise for adolescents. Males in the study had a higher rate of obesity than females, at 20% vs. 17%. The schools had sim ilar rates of obesity to one another, 18% for High School 1 vs. 19% for High School 2, despite their different demographic
114 makeup: approximately 25.5% of High School 2 participants were non-white, vs. 10.6% of High School 1 participants. Given the ty pically higher rate of obesity in AfricanAmerican and Hispanic/Latino adolescents, it might have been expected that High School 2 would have a higher rate of obesity than High School 1. Data on the ages of participants i ndicated a fairly normal distribution, characterized by fewer students at the olde st and youngest ages. Because binge-eating disorder typically has onset in the late t eens (American Psychiat ric Association, 2000), and there were fewer participants in their late teens, the prevalence of binge-eating symptoms may have been underestimated in th is study. Data for the next variable, physical activity, tended to show falling levels of physical activity as the age of the participant increased. Data analysis showed a statistically significant but weak negative correlation between age and physical activ ity (r=-0.13, p = .04). This finding could imply that students exercise less as they get ol der, and is consistent with research that indicates physical activity declines during adolescence (Steinbeck, 2001). The depression scores for participants in both schools resulted in referrals of 10% of students to a school-based mental health worker. Mental health screening tools are intended to over-identify indivi duals actually meeting diagnos tic criteria for a mental health disorder; however it is not known what percentage of referred students met the DSM-IV-TR criteria for depressi on. The 10% rate of stude nts scoring above the cutoff for depression risk was only slightly highe r than the estimated 8.5% of adolescents nationally who suffer from depression (Nati onal Institute of Mental Health, 2000). The referral rate also was lower than the 14% rate found for the normative sample for the RADS-2. Given the reliability estimate in the study of the RADS-2 (r = .92), it appears
115 that the rate of depression in the sample may have been lower than that found in the general population. The sample was demogr aphically representative of the school populations; however it is possibl e that the students who volunteered for the study tended to have less depressive symptoms than thei r school peers who did not volunteer for the study. According to participant responses on th eir history of dieting and bingeing, one out of three or 33% of student s in the study had been on a di et, and one out of eight or 12% had experienced a binge-eating episode in the past. Comparing the 12% bingeing prevalence in the present study to other rates in the research is diffi cult due to the various definitions of binge-eating that have been m easured in other studies (Bulik et al., 2002; Rosenvinge et al., 1999; Ross & Ivis, 1999). The most useful prevalence rates to compare would be those for adolescents mee ting the criteria for binge-eating disorder, rather than sub-clinical levels of binge-ea ting. Prevalence estimates of adolescent bingeeating based on sub-clinical criteria have ranged as high as 18.9% (Ross & Ivis, 1999), while one study that used a stricter set of cr iteria for adolescent binge-eating disorder found a prevalence of 1% (Rosenvinge et al., 1 999). It should also be emphasized that binge-eating disorder onset usually occurs in late adolescence or early 20s, and the sample in this study had fewer participants in their late teens, th erefore the prevalence rate may be an underestimate. The correlation matrices for BMI, physic al activity, depression, dieting, and binge-eating showed four statistically significan t correlations for the aggregated data set, none for the High School 1 data set, and one for the High School 2 data set. However, the correlations were weak and of questionable clinical value, with the highest correlation
116 explaining only 6% of the variance in the va riables. These weak correlations were consistent with the output of the path analysis, results of which indicated that the data failed to fit the revised path model (MODEL2 ), or the final path model (MODEL3). While the non-normality of the dieting and bingeing variables compromised the validity of the chi-squared results and increased the li kelihood of rejecting a tr ue model, the weak correlations in the correlation matrices substantiate the j udgment that the model did not fit the data. Several factors may have negatively impact ed the fit of the model to the data. Although the literature supports the individual relationshi ps between pairs of variables (Berkowitz & Stunkard, 2002; Decaluwe et al., 2002; Goodman & Whitaker, 2002; Price, 2002), there is no existing analys is that looks at the relatio nships among a c onstellation of variables around obesity, and it is possible that the model is incorrect. Alternatively, prior to abandoning the current model, it would seem prudent to investigate the influence of sample size on the outcomes. In additi on, it is possible that a modified sample selection procedure, targeting participants for which the literature suggests the model relationships hold especially tr ue (i.e., extremely obese or depressed individuals), may yield more promising results. For example, one study focused on adolescents who were diagnosed with depression and found that depression was predictive of binge-eating disorder (Zaider et al., 2002). In addition, obese bingers have been found to have more severe obesity, earlier onset and more fr equent dieting, and higher levels of psychopathology including depression (Stunkard, 2002). It should also be noted that one large study on obesity and depression controlle d for many of the variables in the present
117 model and still failed to find a correlation at baseline, although it did find that depression predicted obesity at follow up (Goodman & Whitaker, 2002). Two other factors related to sample sele ction may have impacted the goodness of fit of the model. The sample for the curre nt study included student athletes whose BMI indicated they were obese but who most likely had a lower level of body fat than indicated by BMI. The study did not exclude students who may have had a biologicallybased cause of obesity, such as a thyroid disord er or diabetes. In s hort, the small size of the sample in the present study and the us e of a convenience sample which included primarily non-obese, non-depressed particip ants without bingeeating problems, may have made it difficult for the path model to perform well in the data analysis. When interpreting the poor fit of the model to the data, it is important to recollect that the model excluded heritability, the va riable long known to have the strongest influence on obesity. The biologically-based factor of heritability has been shown to explain approximately two thirds of the vari ance in weight (Grilo & Pogue-Geile, 1991; Price, 2002). The path model in the current study excluded heritability due to the extent of the literature already documenting this va riable as the most significant factor influencing obesity. Assuming that 66% of the influence on weight has already been accounted for by heredity, it would be reasonable to expect the model to account for no more than a total of about 30% of the variance, so that weight would show R2 less than or equal to .30. However, the current path model accounts only for 6% (R2 = .06) of the variance in weight. Combined with the 66% estimated influence of heredity, the variables would explain approxi mately 72% of the variance in weight in the sample.
118 As previously suggested, hypotheses for th e failure of the model to explain a higher percentage of the varian ce in weight fall into two cat egories: a) the model is correct but the sample size was too small with selection criteria not sufficiently focused on significantly obese and depressed adolescent s, and b) the model is incorrect. Future research should continue to explore the cu rrent model. Subse quently, if the model ultimately is rejected then future studie s might improve on the model by incorporating additional variables such as parent weight, consumption of sugar-sweetened beverages (i.e., soft drinks), or a combined variable for gender and ethnicity. These and other suggestions for future research ar e discussed in a later section. Question 2: For the populations of obese and non-obese adolescents, a) what is the rate of binge-eating symptoms, and b) what is th e strength of the relationship between bingeeating and depression? The hypotheses for research question 2 pr oposed: (a) similar levels of bingeeating in obese vs. non-obese groups, and (b ) a positive correlation between binge-eating and depression. The data provided limited support for hypothesis (a), and no support for hypothesis (b). With regard to the levels of binge-eating for obese and non-obese groups, this part of the question was addressed in two ways: comparing the responses to the Yes/No question on bingeing (Have you ever experienced a binge -eating episode?), and comparing the EDI-2 scores for the tw o groups. The first comparison failed to support the hypothesis that the groups did not differ in bingeing behavior, while the second comparison did support the hypothesis. In the first analysis, there was a statistically significant difference between the obese and non-obese groups in responses to the Ever-Binged question, with obese par ticipants (27%) more than three times as
119 likely as non-obese participants (8%) to repor t a previous bingeing experience. These findings are consistent with pr evious research which has plac ed the rate of binge-eating behavior at 36.5% for obese a dolescents (Decaluwe et al., 20 02), although a lower rate of 6.1% was found for obese adolescents meeting the criteria for binge-eating disorder. These rates compare to previously estimat ed rates of binge-eating in the overall population ranging from 1% for adolescents m eeting criteria for binge-eating disorder (Rosenvinge et al., 1999) to 18.9% for adolescen ts who engaged in binge-eating at least once in the past year (Ross & Ivis, 1999). In the next assessment related to hypot hesis (a), which involved the comparison of EDI-2 bingeing responses between obese and non-obese groups, no significant difference in EDI-2 responses was detected between the two groups. This analysis supported the hypothesis that sim ilar levels of binge-eating would be found in obese vs. non-obese groups. The data appear to cont radict each other given the statistically significant difference found between the groups on the Ever-Binged question and the lack of a significant difference between the groups on the EDI-2 bingeing questions. This discrepancy could be explaine d by the explicit and implicit time frames of the items: the Yes/No question pertained to whether a bingei ng episode had occurred at any time in the students past, however most of the EDI-2 items utilize present-tense wording (i.e., I stuff myself with food.). In summary, the results for research questi on 2(a) could be inte rpreted as follows: obese students in the sample were more likely than non-obese students to have experienced bingeing in the past, but no more likely than non-obese students to be currently experiencing binge-ea ting symptoms. It is importa nt to note that neither the
120 Ever-Binged question or the EDI-2 scale is su fficient to make a diagnosis of binge-eating disorder, criteria for which in clude at least two bingeing epis odes a week for six months. Accordingly, the rate of obese participants who met the criteria for binge-eating disorder would be lower than the rate who simply confir ms at least one prior bingeing episode. It would be expected that the group of adolescents who has e ither binged in the past or engages in bingeing currently, will have different needs than non-bingeing obese adolescents in terms of recomm ended obesity interventions. The hypothesis for Research question 2(a) was based upon prior research findings that the majority of bingeeaters are not obese, such as one study that found 12% of women with binge-eating disorder were obese (K inzl et al., 1997). In the present study, a much higher proportion of bingeeaters was obese: 13 students or approximately 45% of the 29 participants who answered Yes to the Ever-Binged question were obese. In addition, it is important to understand and conc eptually differentiate the history of bingeeating experience for obese individuals, 27% in this study, from the proportion of all binge-eaters made up by obese indivi duals, or 45% in this study. Results pertaining to hypothe sis (b), the relationship between binge-eating and depression, suggested a small level of statistic al significance for the overall sample and for the non-obese group, with little clinical significance for either group, despite being the groups with the largest N. No other groups had statistical signifi cance. It is possible that the relationship between binge-eating and depression is one that is better detected over time in a longitudinal study, similar in de sign to previous research which has found depression to predict binge-eating (Zaider et al., 2002). For example, in the present study, participants reported current sympto ms for both binge-eat ing and depression.
121 Consequently if a student pr eviously experienced symptoms of depression which then faded and were replaced by current binge-ea ting symptoms, that st udents data would appear low in depression a nd high in binge-eating. Simila rly, if a student with a high level of depressive symptoms had participated in the study, and then later the depressive symptoms disappeared and binge-eating beha vior increased, the data also would not reflect this scenario. Question 3: For the population of adol escents who engage in both binge-eating and dieting behavior, what is the order of preced ence of binge-eating and dieting (i.e., which occurs first, binge-eating or dieting)? It was expected that a higher percenta ge of students who had both dieted and binged would report having dieted before they had their first binge ing experience. The bingeing and dieting data did not support this hypothesis. The research is somewhat mixed on the order of precedence for bingeing an d dieting, and whether or not dieting is a risk factor for binge-eating disorder (Sant onastaso et al., 1999; Stunkard, 2002). Dieting is accepted as a risk factor for bulimia ner vosa and anorexia nervosa, and this provided the basis for the hypothesis that dieting woul d precede bingeing behavior in the majority of cases. However, dieting has not been firm ly established as a risk factor for bingeeating disorder (Stunkard, 2002). Possible r easons for this mixed view could be an insufficient amount of research to show the di rectionality of the relationship, or that there is no true rule that determines which behavior usually occurs first. It is conceivable that there is no rule that determines which behavior is a risk factor for the other, and that the order of precedence varies with the individual, or that perh aps there is another, stronger risk factor that predisposes an adolescent to binge-eating behavior rath er than dieting.
122 If the results of the present study coul d be generalized to the population of adolescents who have both binged and purge d, they would support the contention that bingeing is more often a risk factor for dieting than the other way around. However, given the small number of participants who reported both dieting and bingeing experiences, it is difficult to generalize th ese results to the po pulation. The wide confidence intervals shown in Table 34 undersco re this problem, suggesting, for example, a 95% likelihood that the true percentage of dieters/bingers who binged first ranges anywhere from 46% to 95%. Even at the low end of the confidence in terval, these results could suggest that some adolescents may be gin bingeing behavior and then initiate dieting as a compensatory behavior, while others begin dieting first and then begin bingeing as a reaction to the re strictiveness of dieting. Implications Perhaps the most critical implication that may be drawn from this investigation is that, if the results of this study can be generalized to the popu lation of American adolescents, the rate of obesity in adolescents continues to ri se. In this study, 19% of all participants were classified as obese and 20% as overwei ght, with a tota l of 39% of adolescents either overweight or obese. These rates are based upon cutoff scores published in 2000, which originally by defin ition identified only the highest 5% of adolescent BMIs as obese. Given the contin ual increase in the ra te of obesity and the many negative health and psycho-social ou tcomes of adolescen t and adult obesity, prevention and treatment should continue to be a priority in ou r society. Similarly, research to enlighten obesity prevention and treatment should continue to be paramount. Another implication of this study concerns the finding th at 12% of adolescents had
123 experienced a binge-eating episode, 33% had been on a diet, and some children started dieting and binge-eating as young as 9 and 10 y ears of age. Results of this study also suggest that a sizeable portion of students who report bingeing beha vior are not obese. The results of the path analysis indicate that there is much still to be understood about the non-hereditary factors that contribute to obesity. It remain s likely that the rise in obesity in the United States and elsewhere stems from some form of gene-environment interaction which has taken shap e in the last several decades of history, with culturally determined lifestyle changes playing a key ro le. Many causal theories and intervention programs are discussed in the popular press a nd in the research literature, however the rates of child and adolescent obes ity continue to rise unabated. It is understandable that institutions and individuals urgently desire to implement solutions to the obesity epidemic, given the serious medical and psyc ho-social outcomes of obesity. However, the results of the current st udy highlight the importance of understanding the forces at work behind the obesity puzzle before implementing effective, research-based interventions. In fact, one could argue that the lack of understandi ng of precisely what has caused the rise in obesity explains why there has been so little success in controlling or reversing it. In this co ntext of urgency, the value of action research which could be conducted in the context of in tervention, might be paramount. School psychologists and others involved in the education of children must be mindful that obesity is not an individual prob lem but a societal one, and an obese child or adolescent should not be assigned the blame fo r his or her obesity or be the subject of negative stereotypes or discrimination. Even in the absence of cl ear data on the causal factors of obesity, the results of this study point to the importan ce of research-based
124 interventions that teach positive messages about nutrition, health, and body image to children at home and at school. These are me ssages that can be conveyed to children by parents, teachers, guidance counselors, sc hool psychologists, and school nurses. For example, mental health professionals can wo rk with health professionals to design and implement primary, secondary, and tertiary prevention and intervention programs for obesity. Again, action research on effectiven ess of these programs can be conducted to contribute to the knowledge base on obesity. Action research programs could address motivating children and adolescents to initia te lifestyle changes in areas that have individually been found to influence obesit y, such as physical activity and time spent on sedentary activities. Assumptions and Limitations Several limitations related to reliability and validity are evident for the present study. For example, the test-rete st reliability of physical act ivity self-report items on the Demographic Physical Activity Questionnaire (DPAQ), as estimated by Brener et al. (1999), was quite low at .55. Howe ver, it is unclear what time interval was used in this assessment, and it would be expected that re liability would decrease as the time interval increases due to the participants changing ha bits and behavior. In addition, the short length of the bulimia scale of the EDI-2, with only five items, would appear to contribute to lower reliability for the binge-eating vari able. Although previous research estimated the reliability for the five bingeing items at .82, the internal consistency reliability calculated in the current st udy was lower at .69. In addition, the EDI-2 test manual cautions that although studies have used the bu limia scale in isolati on, validity can not be assumed if a subscale is used in isolation as the bulimia scale was used in the present
125 study. Another concern relates to the binge ing and dieting history questions, Have you ever been on a diet ? and, Have you ever experienced a binge eating episode ? No validity or reliability data exist on these questions which require Yes/No answers. Another limitation related to validity of BMI should be noted. No secondary measurement of body fat was used for student athletes whose BMI fell within the obese range. Consequently, some students with a high level of muscle mass may have been identified in error as obese. Ideally, a nurse should provide a backup or secondary measurement of body fat, such as skin fold thickness, to avoid wrongly identifying students who engage in rigor ous exercise as obese. Th is limitation of BMI as a measurement tool is common in obesity resear ch and is considered to be outweighed by its convenience. One final limitation regarding reliability concerns the socio-economic status variable and the indicator for free or reduced lu nch. This study used the indicator for free or reduced lunch eligibility to determine socio-economic st atus. Although students who participate in the free and re duced lunch plan must appl y and qualify for the program, there is no requirement that students do th is. Thus, it is conc eivable that some participants may be of low SES but not be taking advantage of the free and reduced lunch program, resulting in an underestimate of t hose from low SES backgrounds. However, 40% of High School 1 student s and 49% of High School 2 students receive free or reduced lunch, suggesting that many student s are participating in the program. Other limitations of the study relate to th e sample. First, the recruitment methods at the two schools differed due to requiremen ts of the school administration, suggesting that different types of students may have been recruited at the two schools. This concern
126 was addressed by statistically comparing the two school sample s, as discussed previously in Chapter 4, and while the groups were not st atistically identical, re sults suggested minor differences in the groups. Another limitation of the study concerns sample size. The number of participants in the study was 251. A sample size of 200 is considered by some to be relatively small for a structural e quation modeling study (Chou & Bentler, 1995). A smaller sample size means that power will be lower and it will therefore be more difficult to reject the null hypothesis. However, offse tting this decreased power to some extent is the reliability and validity of the measurem ent tools for depression (RADS-2) and dieting (DEBQ). Another limitation with re gard to the sample is the f act that all participants are from a single school district and state, which limits external validity of any results and the ability to generalize results to the larger population of Ameri can adolescents. In addition, the demographic makeup of both high schools is le ss diverse than for the state as a whole, limiting external validity of the findings. Another limitation related to the sample relates to the opinion by some researchers that individuals with eating disorders are underrepresented in groups that participate in eating disorder st udies (Wilson et al., 1996). If students with eating disorders chose not to participate in the present study, the sample would be biased and consequently th e data would be less than representative of the high school population. Finally, the samp le by definition did not include students who had dropped out of school. To combat th ese limitations, students were recruited in a variety of ways and incentives were provi ded for participation. In addition, some participants were recruited from the perf ormance-based diploma program and were not enrolled in the regular high school program.
127 An additional limitation of the study concer ns the self-report measures used to assess depression, dieting, binge ing, and physical activity. The most accurate assessment of both depression and binge-ea ting would be through individu al diagnostic interview, rather than through a questionnaire. As in a ny study utilizing self-report measures rather than direct assessment of the va riable, there is an inherent ri sk that partic ipants are not truthful or results will not be valid or reliable. These concerns were mitigated by the procedures documented and explained to par ticipants, the presence of the investigator during completion of questionnaires, and the or dering of the questionnaires within each packet. In another limitation of the study, the key a ssumption in the path analysis is the fact that heredity was excluded from the mode l. As a result, as much as 66% of the variance in weight was unexplai ned by the model at the outset. One additional caution on interpreting the resu lts of the path analysis should be mentioned. Although it is clear that the relationships in the path model are dire ctional and significance may have been found in the data analysis, we must keep in mind that the predictor variables were not manipulated and the study was not an experi mental one. Therefore, had there been significant results of the path analysis, we w ould be careful with statements of causality and limit any interpretations to correlational discussions. Directions for Future Research An opportunity remains for future resear ch to better understand the relationships among the variables that influence weight and ob esity. Ideally future research will begin with further investigat ion of the viability of the current path model by using a modified sample selection process and a larger sample. Many of the re search studies that provided
128 the basis for the model focused upon participan ts who were obese (Decaluwe et al, 2002; Stunkard, 2002), depressed (Zaider et al ., 2002), or had binge-eating disorder (Santonastaso et al, 1999). Similarly, future investigation into the current path model should focus on participants who are significan tly obese, have clini cal depression, and/or have binge-eating disorder. Such a sa mple population may be easier and more appropriate to recruit through pe diatricians or clinics than th rough the school setting. The sample selection process would also exclude participants who had a medical cause of obesity. Future path analysis research utilizing a much larger sample size would provide greater power and opportunity for statistical significance. A larger sample size would allow the inclusion of the gender/ethnicity va riable in the model rather than breaking them into two variables. In addition, future research should attempt to mirror the general population in ethnic diversity, to provide the mo st useful data and allow generalization to the population. In order to avoid having athl etes incorrectly identified as obese, an alternate, direct method of body fat analysis should be employed such as skin fold thickness, whereby individuals with a high BMI would be directed to the secondary measurement station. Alternatively, the direct method of body fat measurement could be used for all participants. In addition, future researcher s could consider other means of measuring the physical activity and bingeing behaviors. Actu al physical activity might be tracked by a participant on a daily log, or with a pedomet er worn each day. Binge eating could be assessed directly by interview as a followup to the bulimia scale of the EDI-2 or other screening tool.
129 In the event that future research cannot va lidate the current path model even with the use of a larger sample and modified sample selection process, changes to the path model should be considered. For example, a path model that includes parent weight as one of the variables might be found to have si gnificant results, give n the known impact of heredity on weight and obesity (Grilo & Pogue -Guile, 1991; Price, 2002). Heredity also bears influence upon several other variables in the current model, therefore its inclusion may produce a model with a better fit. Pare nt depression and pare nt binge-eating might be considered as additional variables since heredity is considered a risk factor for depression (Smith et al., 2003) and binge-eati ng (Bulik et al., 2002). Future research might consider inclusion of soft drink consum ption as a variable in the path model. Authors of a recent review of 30 experime ntal and epidemiological studies, conducted between 1966 and 2005, concluded that greater intake of sugar-sweetened beverages, particularly carbonated soft dri nks, is associated with weight gain and obesity in both children and adults (Malik, Schulze, & Hu, 2006). Time spent watching television and playing video games represents another variab le that might be considered for a future path analysis. These sedentary activities can be viewed as types of physical inactivity, and they have become integral to American lifestyle over the last several decades. Many researchers have linked them in theory to the rise in child and adolescent obesity (Stunkard, 2002), and have promoted reductions of child sedentary activities (Steinbeck, 2001). Obesity research in general, and path an alysis research in particular, remain important and could shed light on the interrelationships of variables contributing to the continuing upsurge in obesity in adolescents Genetic research should continue, along
130 with recently emerging lines of research in possible biochemical and viral causes of obesity. Heredity alone cannot explain the increase in obesity in the last 20 years, and further research can help to uncover the comb ination of influences that has led to the obesity epidemic. Such unders tanding is critical in orde r to properly design effective programs for prevention and treatment. The current lack of understanding of the causes of obesity, and the tendency to blame each individual, have allowed the obesity problem to progress to now epidemic proportions. Conclusion In conclusion, this study investigated the relationships betw een variables that have been found to influence weight and obes ity. Results indicate th at the proposed path model and its revisions were a poor fit for th e data. The level of binge eating reported by groups of obese and non-obese students was si milar, however a higher percentage of obese students than non-obese students reported a previous binge ea ting experience. For participants in the study who reported a history of both dieting and binge eating, onset of bingeing preceded onset of dieting more often than dieti ng preceded bingeing. However, the wide confidence intervals and the small number of participants with a history of both dieting and bingeing prevent ge neralizing this result to the U.S. adolescent population. In the pursuit of effective prev ention and treatment interventions, future academic research and action research in the schools should continue to investigate the causes and the relationships between the vari ables that work together to produce the countrys epidemic of obesity.
131 References Ackard, D. M., Neumark-Sztainer, D., Stor y, M., & Perry, C. (2003). Overeating among adolescents: prevalence and associations with weight-related characteristics and psychological health. Pediatrics, 111(1), 67-74. American Psychiatric Association. (1994). Diagnostic and statistic al manual of mental disorders (4th ed.). Washington, DC: Author. American Psychiatric Association. (2000). Diagnostic and statistic al manual of mental disorders (4th ed., text rev.). Washington, DC: Author. Ash, P. (1994). Review of the Eating Disorder Inventory-2. Mental Measurements Yearbook, 12 Berkowitz, R. I., & Stunkard, A. J. (2002). Development of childhood obesity. In T. A. Wadden, A. J. Stunkard (Eds.), Handbook of obesity treatment (pp. 515-531). New York: Guilford Press. Birch, L. L., & Fisher, J. O. (1998). Devel opment of eating behavi or among children and adolescents. Pediatrics, 101 (Suppl.), 539-549. Brener, N. D., Kann, L., McManus T., Kinchen, S. A., Sundberg, E. C., & Ross, J. G. (2002). Reliability of the 1999 Youth Ri sk Behavior Survey questionnaire. Journal of Adolescent Health, 31 336-342. Brosse, A. L., Sheets, E. S., Lett, H. S., & Blumenthal, J. A. (2002). Exercise and the treatment of clinical depression in adults: recent findings and future directions. Sports Medicine, 32 (12), 741-60. Bulik, C. M., Sullivan, P. F., & Kendler, K. S. (2002). Genetic and environmental contributions to obesity and binge eating. International Journal of Eating Disorders, 33 (3), 293-298. Carter, J. C., Stewart, D. A., & Fairburn, C. G. (2001). Eating disorder examination questionnaire: Norms for young adolescen t girls. Behavi our Research and Therapy, 39(5), 625-632. Centers for Disease Control. (n.d.). Using the BMI-for-age growth charts Retrieved April 19, 2003, from
132 http://www.cdc.gov/nccdphp/dnpa/growthcha rts/training/modules/module1/text/in tro.htm Chou, C., & Bentler, P. M. (1995). Estimates and tests in structural equation modeling. In R. H. Hoyle (Ed.), Structural Equation Modeling: Concepts, Issues, and Applications (pp. 37-55). Thousand Oaks, CA: Sage Publications. Decaluwe, V., Braet, C. & Fairburn, C. G. (2002). Binge eating in obese children and adolescents. International Journal of Eating Disorders, 33 (1), 78-84. Ebbeling, C. B., Pawlak, D. B., & Ludwi g, D. S. (2002). Childhood obesity: publichealth crisis, common sense cure. Lancet, 360, 473-482. Fairburn, C. G., Doll, H. A., Welch, S. L., Ha y, P. J., Davies, B. A., & OConnor, M. E. (1998). Risk factors for binge eating di sorder: A community based, case-control study. Archives of General Psychiatry, 55 425-432. Field, A. E., Barnoya, J., & Colditz, G. A. (2002). Epidemiology and health and economic consequences of obesity. In T. A. Wadden, A. J. Stunkard (Eds.), Handbook of obesity treatment (pp. 3-94). New York: Guilford Press. Fisher, M., Schneider, M., Burns, J., Symons, H., & Mandel, F. S. (2001). Differences between adolescents and young adults at presentation to an eating disorders program. Journal of Adolescent Health, 28 222-227. Florida Department of Education (2005). Florida School Indicators Report. Retrieved September 24, 2006, from http://data.fldoe.org/fsir/default.cfm Florida Department of Education (2006a). Florida School Grades Retrieved September 24, 2006, from http://schoolgrades.fldoe.org/default.asp Florida Department of Education (2006b). NCLB School Public Accountability Report. Retrieved September 24, 2006, from http://doewebprd.doe.state.fl.us/eds/nclbspar/main0506.cfm Garner, D. M. (1991). Eating Disorder Inventory-2 Lutz, FL: Psychological Assessment Resources, Inc. Goodman, E. (1999). The role of socio economic status gradients in explaining differences in US adolescents health. American Journal of Public Health, 89 (10), 1522-1528. Goodman, E., & Whitaker, R. C. (2002). A prosp ective study of the role of depression in the development and persistence of adolescent obesity. Pediatrics, 109 (3), 497503.
133 Grilo, C. M., Masheb, R. M., & Wilson, G. T. (2001). A comparison of different methods for assessing the features of eati ng disorders in patients with binge eating disorder. Journal of Consulting an d Clinical Psychology, 69 (2), 317-322. Grilo, C. M., & Pogue-Geile, M. F. (1991). Th e nature of environmental influences on weight and obesity: A beha vior genetic analysis. Pyschological Bulletin, 110 520-537. Hammen, C., & Rudolph, K. D. (1996). Chil dhood Depression. In E. J. Mash, R. A. Barkley (Eds.), Child psychopathology (pp.153-195). New York: Guilford Press. Hartley, P. (1998). Eating disorders and health education. Psychology, Health and Medicine, 3 (1), 133-140. Hatcher, L. (1996). Using SAS PROC CA LES for Path Analysis: An Introduction. Structural Equation Modeling, 3 (2), 176-192. Hu, L., & Bentler, P. M. (1995). Evalua ting Model Fit. In R. H. Hoyle (Ed.), Structural Equation Modeling (pp. 76-99). Thousand Oaks, Calif ornia: Sage Publications. Kinzl, J. F., Traweger, C., Trefalt, E., Ma ngweth, B., & Biebl, W. (1999). Binge eating disorder in females: A popul ation-based investigation. International Journal of Eating Disorders, 25 (3), 287-292. Kirkcaldy, B. D., Shephard, R. J., & Siefe n, R. G. (2002). The relationship between physical activity and self-image and pr oblem behavior among adolescents. Social psychiatry and psychiatr ic epidemiology, 37 544-550. Latner, J. D., & Stunkard, A. J. (2003). Ge tting worse: The stigmatization of obese children. Obesity Research, 11 (3), 452-456. Lissau, I., & Sorensen, T. I. (1993). Sc hool difficulties in childhood and risk of overweight and obesity in young adulthood: A ten year prospective population study. International Journal of Obesity and Related Metabolic Disorders: Journal of the International Associ ation for the Study of Obesity, 17 (3), 169-175. Malik, V. S., Schulze, M. B., & Hu, F. B. ( 2006). Intake of sugar-sweetened beverages and weight gain: A systematic review. The American Journal of Clinical Nutrition, 84 (2), 274-288. McMurray, R. G., Harrell, J. S., Deng, S., Bradley, C. B., Cox, L. M., & Bangdiwala, S. I. (2000). The influence of physical activ ity, socioeconomic status, and ethnicity on the weight status of adolescents. Obesity Research, 8 130-139.
134 Mokdad, A. H, Marks, J. S., Stroup, D. F., & Gerberding, J. L. (2004). Actual causes of death in the United States. Journal of the American Medical Association, 291 (10), 1238-45. National Center for Health Statistics (2002). Prevalence of overweight among children and adolescents: United States, 1999-2000 Retrieved April 19, 2003 from http://www.cdc.gov/nchs/products /pubs/pubd/hestats/overwght99.htm National Institute for Mental Health (2000). Depression in childre n and adolescents: A fact sheet for physicians. Retrieved March 14, 2004 from http://www.nimh.nih.gov/publ icat/depchildresfact.pdf National Research Council (2002). Minority students in spec ial and gifted education Washington, D.C.: National Academy Press. Netemeyer, S. B., & Williamson, D.A. (2001) Assessment of eating disturbance in children and adolescents with eating disord ers and obesity. In J. K. Thompson, L. Smolak (Eds.), Body image, eating di sorders, and obesity in youth: Assessment, prevention, and treatment (pp. 215-233). Washington D.C.: American Psychological Association. Paluska, S. A. (2000). Physical activity and mental health: Current concepts. Sports medicine, 29 (3), 167-180. Perez, M., & Joiner, T. E. (2003). Body di ssatisfaction and disord ered eating in black and white women. International Journal of Eating Disorders, 33 (3), 342-350. Pine, D. S., Goldstein, R. B., Wolk, S., & Weissman, M. M. (2001). The association between childhood depression a nd adulthood body mass index. Pediatrics, 107 (5), 1049-56. Prentice, A. M., & Jebb, S. A. (2001). Beyond body mass index. Obesity Reviews, 2 141-147. Price, A. (2002). Genetics and common obes ities: Background, current status, strategies, and future prospects. In T. A. Wadden, A. J. Stunkard (Eds.), Handbook of obesity treatment (pp. 73-94). New York: Guilford Press. Price, R. A., & Lee, J. H. (2001). Risk ra tios for obesity in families of obese AfricanAmerican and Caucasian women. Human Heredity, 51 (1-2), 35-40. Reynolds, W. M. (2002). Reynolds Adolescent Depression Scale, 2nd Edition Lutz, FL: Psychological Assessment Resources, Inc.
135 Rosenvinge, J. H., Borgen, J. S., & Borresen, R. (1999). The prevalence and psychological correlates of anorexia ne rvosa, bulimia nervosa and binge eating among 15-year-old students: A controlled epidemiological study. European Eating Disorders Review, 7 382-391. Ross, H. E., & Ivis, R. (1999). Binge eati ng and substance use among male and female adolescents. International Journal of Eating Disorders, 26 245-260. Santonastaso, P., Ferrara, S., & Favaro, A. (1999). Differences between binge eating disorder and nonpurging bulimia nervosa. International Journal of Eating Disorders, 25 215-218. Schwartz, M. B., & Puhl, R. (2003). Childhood obesity: A societal problem to solve. Obesity Reviews, 4 57-71. Schwimmer, J. G., Burwinkle, T. M., Varni, J. W. (2003). Health-related quality of life of severely obese children and adolescents. Journal of the American Medical Association, 289 (14), 1813-1819. Shinke, S. (1994). Review of th e Eating Disorder Inventory-2. Mental Measurements Yearbook, 12 Smith, D., Muir, W., & Blackwood, D. (2003). Genetics of earlyonset depression. The British Journal of Psychiatry, 182 363-364. Steinbeck, K. S. (2001). The importance of physical activity in the prevention of overweight and obesity in childhood : A review and an opinion. Obesity Reviews 2, 117-130. Stice, E. (1999). Clinical implications of pschosocial research on bulimia nervosa and binge-eating disorder. Journal of Clinical Psychology, 55 (6), 675-683. Stunkard, A. J. (2002). Binge-e ating disorder and the nighteating syndrome. In T. A. Wadden, A. J. Stunkard (Eds.), Handbook of obesity treatment (pp. 107-121). New York: Guilford Press. Stunkard, A. J., Faith, M. S., & Allison, K. C. (2003). Depression and obesity. Biological Psychiatry, 54 (3), 330-337. Tershakovec, A. M., Weller, S. C., & Ga llagher, P. R. (1994). Obesity, school performance and behaviour of black, urban elementary school children. International Journal of Obesity and Rela ted Metabolic Disorders: Journal of the International Association for the Study of Obesity, 18 (5), 323-327.
136 Thames Valley Test Company (2004). The Dutch Eating Behavior Questonnaire Retrieved March 6, 2004, from http://www.tvtc.com/download/ppt/1 United States Department of Health and Human Services (2004, July 15). HHS Announces Revised Medicare Obesity Coverage Policy Retrieved August 1, 2004, from http://www.hhs.gov/news /press/2004pres/20040715.html Van Strien, T. (2002). Dutch Eating Behaviour Questionnaire (DEBQ). Thames Valley Test Company Ltd: Suffolk, England. Varni, J. W., Seid, M., & Kurtin, P. S. (2001) The PedsQL 4.0: Reliability and validity of the Pediatric Quality of Life Invent ory Version 4.0 Generic Core Scales in healthy and patient populations. Medical Care, 39 (8), 812. Weissman, M. M., Wolk, S., & Goldstein, R. B. (1999). Depressed adolescents grown up. Journal of the American Medical Association, 281 1701-13. West, S. G., Finch, J. F., & Curran, P. J. (1995). Structural Equation Models with Nonnormal Variables: Problems and Re medies. In R. H. Hoyle (Ed.), Structural Equation Modeling: Concepts, Issues, and Applications (pp. 37-55). Thousand Oaks, CA: Sage Publications. Wilson, G. T., Heffernan, K., & Black, C. M. D. (1996). Eating disorders. In E. J. Mash, R. A. Barkley (Eds.), Child psychopathology (pp.541-571). New York: Guilford Press. Wilson, G. T., Nonas, C. A., & Rosenblum, G. D. (1993). Assessment of binge eating in obese patients International Journal of Eating Disorders, 13 (1), 25-33. Zaider, T. I., Johnson, J. G., & Cockell, S. J. (2002). Psychiatric disorders associated with the onset and persistence of bulimia nervosa and binge eating disorder during adolescence. Journal of Youth and Adolescence, 31 (5), 319-329.
138 Appendix A Space below reserved for IRB Stamp Please leave blank Dear Parent: The following information is to help you decide whether or not you want to allow your teenager to be a part of a low risk research study. Please read this carefully. If you do not understand anything, ask the person in charge of the study: Angela T. Sheble The title of the research study is Correlates of Weight in Adolescents: A Path Analysis The study will be done at 2 high schools. Other high schools also may be included in the study if necessary. Your teenager is being asked to pa rticipate because he/she attends one of these schools. The study will have about 400 participants altogether. The purpose of this research study is to learn more about the things that may have to do with weight in adolescents, like physical activity, eating habits, depression, dieting, socio-economic status, special education, gender, and ethnicity/race. The study will take place from January through May, 2005 Your teenager will visit the clinic one time for 30-40 minutes, during one class period on a Friday, to participate. During this visit, yo ur teenager will be asked to complete 4 short questionnaires on dieting, depression, eating habits, and physical activity, and he/she will have their height, weight, and Body Mass Index (BMI) measured. Your teenager will also be entered in a free raffle to win a gift for participating. Raffle prizes may include gift certificates for items like CDs, clothing, movies, and restaurants You and your teenager will not be paid for participation in this study. However, you and/or your teenager may benefit by learning his/her Body Mass Index. Also, yo ur teenager will be entered in a raffle. There is no known risk to your teenager from being a part of this research study. You and your teenagers privacy an d research records will be kept confidential to the full extent required by law. Authorized research personnel, employees of the Department of Health and Human Services, and the USF Institutional Review Board and its staff and an y other individuals acting on behalf of USF, may inspect the records from this research project. If your teenagers depression questionnaire indicates that he or she may be depressed, you will be notified promptly by a school professional such as a counselor, social worker, or psychologist. The results of this study may be published. However, the data obtained from your teenager will be combined with data from other teen agers in the publication. The published results will not include your teenagers name or any other information that would personally identify your teenager in any way. Packet ID numbers will be used so your teenagers name will not appear on the questionnaires or in the computer data. Only the person in charge of the study will have access to the packet ID numbers and names. Allowing your teenager to participate in this resear ch study is voluntary. You are free to allow your teenager to participate in this research study or to withdraw him/her at any time. If you choose not to allow your teenager to participate or if you remove your t eenager from the study, there will be no penalty. Your decision to allow your teenager to participate (or not to allow your teenager to participate) will in no way affect his/her status at school. If you have any questions about this research study, contact Angela T. Sheble If you have questions about your rights as a person who is taking part in a research study, you may contact the Division of Research Compliance of the University of South Florida at (813) 974-5638 Thank you in advance for allowing your teenager to participate in this research study. Please sign the Parent Consent form on page 2. Your teenager will also be asked to sign (at school) the Childs Assent form on page 3. Sincerely, Angela T. Sheble, Ed.S.
139 Appendix A (Continued) Consent for Child to Take Pa rt in this Research Study I freely give my consent to let my child take part in this study. I understand that this is research. I have received a copy of this consent form. ________________________ ________________________ ___________ Signature of Parent Printed Name of Parent Date of child taking part in study Investigator Statement I certify that participants have been provided with an informed consent form that has been approved by the University of South Floridas Institutional Review Boar d and that explains the nature, demands, risks, and benefits involved in participating in this study. I further certify that a phone number has been provided in the event of additional questions. ________________________ Angela T. Sheble, Ed.S. _______________ Signature of Investigator Printed name of Investigator Date
140 Appendix A (Continued) Childs Assent Statement Angela Sheble has explained to me this research stud y called Correlates of Weight in Adolescents: A Path Analysis. I agree to take part in this study. ________________________ ________________________ ___________ Signature of Child Printed Name of Child Date taking part in study ________________________ Angela T. Sheble, Ed.S. ___________ Signature of person Printed name of person Date obtaining consent obtaining consent ________________________ ________________________ ___________ [Optional] Signature of Witness Printed Name of Witness Date If child is unable to give assent, please explain the reasons here: ________________________ Angela T. Sheble, Ed.S ___________ Signature of person Printed name of person Date obtaining consent obtaining consent ________________________ ________________________ ___________ [Optional] Signature of Witness Printed Name of Witness Date
141 Appendix A (Continued) El espacio abajo es para el cuo de la Junta de Revisin InstitucionalPor favor de dejar en blanco Estimados Padres: La siguiente informacin es para ayudarles a decidir si usted quiere o no dejar que su adolescente participe en un estudio de investigacin de bajo riesgo. Por favor lea esto cuidadosamente. Si usted no entiende nada, por favor pregntele a la persona encargada de este estudio: Angela T. Sheble El titulo de este estudio es Factores Relacionados con el Peso en los Adolescentes. El estudio ser hecho en 2 escuelas superiores. Otras escuelas superiores podrn ser incl uidas tambin si hay necesidad. Su adolescente ha sido elegido a participar porque el o ella van a esta escuela. Este estudio tendr alrededor de 400 participantes en total. El propsito de este estudio de investigacin es para saber mas acerca de las cosas que pueden tener que ver en el peso de los adolescentes como actividad fsica, hbitos alimenticios (lo que comen), estado de animo (depresin), dieta, clase socio-econmi ca, educacin especial, sexo, y raza. El estudio se llevara a cabo de Enero a Mayo del 2005 Su adolescente visitara la clnica una vez por 30-40 minutos un Viernes durante un periodo de clase para participar. Durante esta visita, su adolescente tendr que llenar 4 cuestionarios cortos sobre dieta, depresin, lo qu e comen, y actividad fsica. Tambin se obtendr la estatura, el peso, y se le medir el ndice de la masa del cuerpo Su adolescente podr participar en una rifa gratis para poder ganarse premios. Estos pueden incluir discos compactos, ropa, entradas para el cine, y restaurantes Su adolescente no recibir pago por participar en este estudio, pero se podr beneficiar sabiendo su ndice de masa y el poder participar en la rifa. Que se sepa, no hay ningn riesgo en que su adolescente participe en este estudio. La privacidad de usted y la de su adolescente adems de los datos obtenidos en este estudio sern todos confidenciales segn lo requiere la ley. El personal investigativo autorizado, empleados del Departamento de Salud y Servicios Humanos y la Junta de Revisin Institucional de La Universidad del Sur de la Florida y sus empleados u otras personas representando a la Universidad, pueden inspeccionar los datos de este estudio investigativo. Si el resultado del cuestionario de depresin de su adolescente indica que el o ella puede estar deprimido, usted ser notificado rpidamente por un profesional de la escuela como un consejero, trabajador social o sicologo. Los resultados de este estudio podrn ser publicados pero, los datos obtenidos de su adolescente sern combinados con los datos de los otros participantes. Los resultados publicados no incluirn el nombre de su adolescente o cualquier otra informacin que pueda identificar a la persona en alguna forma. Un numero de identificaci n ser usado y de esa forma, el nombre de su adolescente no aparecer en le cues tionario ni en los datos de com putadora. Solamente la persona encargada de este estudio tendr acceso al numero de identificacin y a los nombres de los participantes. Dejar que su adolescente participe en este estudio es voluntario. Usted esta libre de permitir de que su adolescente participe en este estudio o puede dejar de participar en cualquier momento. Si usted decide que su adolescente no participe o si usted quita a su adolescente de este estudio, no habr ningn tipo de penalidad. Su decisin de participar o no partic ipar no afectaran a el o ella en su escuela. Si tiene alguna pregunta acerca de este estudio, favor de ponerse en contacto con Angela T. Sheble. Si usted tiene preguntas relacionadas con sus derechos como persona quien esta tomando parte en un estudio de investigacin, puede llamar a La Divisin de Cumplimiento de Investigacin de La Universidad del Sur de la Florida a (813) 974-5638.
142 Appendix A (Continued) Muchas gracias por dejar que su adoles cente participe en este estudio. Por favor firme la forma de Permiso de Padres en la pgina 3. Su adolescente tend r que firmar tambin en la escuela la forma de Consentimiento de Hijo que aparece en la pagina 4. Sinceramente, Angela T. Sheble, Ed.S.
143 Appendix A (Continued) Permiso para que su hijo o hija tome pa rte en este estudio de investigacin Yo doy m consentimiento libremente para que m hijo o hija tome parte en este estudio. Yo entiendo que esto es un estudio de investigacin. Yo he recibido una copia de esta forma de permiso o consentimiento. ________________________ ________________________ ___________ Firma del padre del hijo/hija que Nombre del Padre Fecha toma parte en este estudio en letra de molde Declaracin del Investigador Yo certifico que los participantes has sido provedos con la forma de permiso la cual ha sido aprobada por Junta de Revisin Institucional de La Universidad del Sur de la Florida y explica este estudio incluyendo las demandas, riesgos, y beneficios relacionados con este estudio. Yo certifico adems que un numero de telfono ha sido dado en caso de que haya preguntas adicionales. ________________________ Angela T. Sheble, Ed.S. ____________ Firma del Investigador Nombre del Investigador Fecha
144 Appendix A (Continued) Declaracin de Consentimiento del Adolescente Participante Angela Sheble me ha explicado este estudio llamado Factores Relacionados con el Peso en los Adolescentes. Yo estoy de acuerdo en participar en este estudio. ________________________ ________________________ ___________ Firma del Adolescente que Nombre del Adolescente Fecha toma parte en este estudio en letra de molde ________________________ Angela T. Sheble, Ed.S. ___________ Firma de la persona que Nombre de la persona que Fecha obtiene el consentimiento obtiene el consentimiento ________________________ ________________________ ___________ [Opcional] Firma del Testigo Nombre del Testigo Fecha en letra de molde Si el adolescente no puede dar su consentimi ento, por favor explique las razones aqu: ________________________ Angela T. Sheble, Ed.S ___________ Firma de la persona que Nombre de la persona que Fecha obtiene el consentimiento obtiene el consentimiento ________________________ ________________________ ___________ [Opcional] Firma del Testigo Nombre del Testigo Fecha en letra de molde
145 Appendix B Demographic and Physical Ac tivity Questionnaire (DPAQ) For School Use: Survey Number: ______ BMI:______ SES:_____ ESE:_____ Demographic and Physical Activity Questionnaire (DPAQ) Part I. Student Information: Please provide the following information. 1. Sex Male Female 3. What grade are you in? 9th 10th 11th 12th 2. Ethnicity : Do you think of yourself as ? __ __ __ __ __ African American/Black Asian Caucasian/White Latino/Hispanic Native American/Alaskan Other:_________________ 4. How old are you? 13 17 14 18 15 19 16 20 Part II. Physical Activity: The following questions are about physical activity. Circle only one answer for each question. 1. On how many of the past 7 days did you exercise or participate in physical activities for at least 20 minutes that made you sweat and breathe hard, such as basketball, jogging, swimming laps, tennis, fast bicycling or similar aerobic activities? 0 days 1 day 2 days 3 days 4 days 5 days 6 days 7 days 2. On how many of the past 7 days did you participate in physical activity for at least 30 minutes that did not make you sweat or breathe hard such as fast walking, slow bicycling, skating, pushing a lawn mower, or mopping floors? 0 days 1 day 2 days 3 days 4 days 5 days 6 days 7 days 3. On how many of the past 7 days did you do exercises to strengthen or tone your muscles, such as pushups, sit-ups, or weight lifting? 0 days 1 day 2 days 3 days 4 days 5 days 6 days 7 days 4. On how many of the past 7 days did you participate in at least 20 minutes of physical activity outside the normal school day? 0 days 1 day 2 days 3 days 4 days 5 days 6 days 7 days 5. In an average week when you are in school (Monday-Friday), on how many days do you go to physical education (PE) classes? 0 days 1 day 2 days 3 days 4 days 5 days 6. During an average physical education (PE) class how many minutes do you spend actually exercising or playing sports? I do not take PE Less than 10 minutes 10 to 20 minutes 21 to 30 minutes More than 30 minutes
146 Appendix B (Continued) 7. In an average school day (Monday-Friday), how many hours do you spend doing any of the following activities: watching television/movies, playing video games, reading, homework, or on the computer/Internet? 0 (No time on these activities) Less than 1 hour per day 1 hour per day 2 hours per day 3 hours per day 4 hours per day 5 or more hours per day
About the Author Angela T. Sheble received her B.S. in Commerce in 1983 from the University of Virginia in Charlottesville, Virginia. She wo rked for 18 years in the computer software industry in various management, consulting, and technical capacities, in both the private and public sector. Ms. Sheble entered th e University of South Floridas School Psychology doctoral program in 2000 where she chose pediatric school psychology as her area of emphasis. She earned her Masters degree in 2001 and her Education Specialist degree in 2004. Ms. Sheble won the Flor ida Association of School Psychologists Graduate Studies Award in 2001, and the Un iversity Involvement Award in 2003 from the Tampa Bay Academy of Hope. As part of her clinical tr aining, Ms. Sheble worked in early intervention in 2003 and 2004. Ms. Sheb le completed her internship in 2004 and currently works as a school psychologist. Ms. Sheble has a 13-year old son with autism spectrum disorder.
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 001914245
007 cr mnu|||uuuuu
008 071019s2006 flu sbm 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0001743
Sheble, Angela T.
Correlates of weight in adolescents :
b a path analysis
h [electronic resource] /
by Angela T. Sheble.
[Tampa, Fla] :
University of South Florida,
ABSTRACT: This study examined the interrelationships between adolescent weight and several other variables thought to impact weight and obesity: physical activity, depressive symptoms, binge-eating symptoms, dieting, socio-economic status, special education status, gender, and ethnicity. The sample consisted of 251 high school students in rural Florida who completed measures of depression, binge-eating, dieting, and physical activity. Measurement instruments included the Reynolds Adolescent Depression Scale-2nd Edition (RADS-2), the bulimia scale of the Eating Disorder Inventory-2 (EDI-2), the dieting scale of the Dutch Eating Behavior Questionnaire (DEBQ), and a physical activity questionnaire derived from the Youth Risk Behavior Scale for Students (YRBSS). The study utilized path analysis, a group correlational design, to determine whether the proposed path model fit the data. Obese and non-obese students also were compared with regard to a) the levels of binge-eating symptoms,^ and b) the relationship between binge-eating and depression. Path analysis results were not statistically or clinically significant, suggesting a poor fit of the model to the data. Results indicated 19% of participants were obese and 20% were overweight. More than three times as many obese students than non-obese students reported experiencing a binge-eating experience at some time in the past. However, on the bulimia scale of the EDI-2, obese and non-obese participants did not differ statistically in their responses. Statistically but not clinically significant correlations were revealed between depression and binge-eating for the sample and also for non-obese students. For the sub-sample of 13 students who had both binged and dieted, 7 had binged first, 3 had dieted first, and 3 binged and dieted for the first time at the same age. Future research should continue to investigate the relationships of the variables related to obesity. Future directions might include a larger sampl e size and a modified sample selection process. Action research should continue in the areas of obesity prevention and intervention, and student services personnel should promote healthy lifestyle choices and a recognition of obesity as a socio-cultural problem.
Dissertation (Ph.D.)--University of South Florida, 2006.
Includes bibliographical references.
Text (Electronic dissertation) in PDF format.
System requirements: World Wide Web browser and PDF reader.
Mode of access: World Wide Web.
Title from PDF of title page.
Document formatted into pages; contains 146 pages.
Adviser: Kathy Bradley-Klug, Ph.D.
Structural equation modeling.
x Interdisciplinary Education
t USF Electronic Theses and Dissertations.