Copyright 2000 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2000
The study objectives were to assess (1) the prevalence of dieting and disordered eating among adolescents; (2) the sociodemographic, anthropometric, psychosocial, and behavioral correlates of dieting and disordered eating; and (3) whether adolescents report having discussed weight-related issues with their health care providers.
Cross-sectional school-based survey.
A nationally representative sample of 6728 adolescents in grades 5 to 12 who completed the Commonwealth Fund surveys of the health of adolescent girls and boys.
Main Outcome Measures
Dieting and disordered eating (binge-purge cycling).
Approximately 24% of the population was overweight. Almost half of the girls (45%) reported that they had at some point been on a diet, compared with 20% of the boys. Disordered eating was reported by 13% of the girls and 7% of the boys. Strong correlates of these behaviors included overweight status, low self-esteem, depression, suicidal ideation, and substance use. Almost half of the adolescents (38%-53%) reported that a health care provider had at some point discussed nutrition or weight with them. Discussions on eating disorders were reported by lower percentages of girls (24%) and boys (15%).
The high prevalence of weight-related concerns suggests that all youth should be reached with appropriate interventions. Special attention needs to be directed toward youth at greatest risk for disordered eating behaviors, such as overweight youth, youth engaging in substance use behaviors, and youth with psychological concerns such as low self-esteem and depressive symptoms.
WEIGHT-RELATED behaviors and conditions include eating disorders such as anorexia and bulimia nervosa, disordered eating behaviors such as self-induced vomiting and binge eating, dieting behaviors, and obesity. Weight-related behaviors and conditions are prevalent among adolescents, and in particular among adolescent girls.1- 8 Weight-related conditions are of serious public health concern in light of their high prevalence and their potentially adverse effects on growth, psychosocial development, and physical health outcomes.9- 13 Unhealthy dieting and disordered eating behaviors among youth are of concern in that they have the potential to adversely affect nutrient intake, mental health status, and long-term health outcomes. Dieting behaviors have been found to be associated with inadequate intakes of essential nutrients such as calcium,12,14 and with mental health symptoms such as fatigue, anxiety, depression, and mental sluggishness.1 Dieting and disordered eating behaviors may be indicative of increased risk for the later development of eating disorders; Patton et al15 found that the relative risk for dieters to develop an eating disorder was 8 times higher than that for nondieters after a 1-year period.
To plan interventions aimed at the prevention of weight-related behaviors and conditions, we need to know the scope of the problems, ie, their prevalence among adolescents in the general population. In this study, we assess the prevalence of a range of weight-related concerns and behaviors among a nationally representative sample of adolescent girls and boys. We further examine the sociodemographic, anthropometric, psychosocial, and behavioral correlates of dieting behaviors and more severe disordered eating behaviors (ie, binge-purge cycling) to gain insight into the causes of these behaviors and to identify at-risk adolescents. In light of the potentially adverse health implications of these behaviors, it is important for health care providers to discuss weight-related issues with adolescents. Therefore, we also assess adolescent perceptions as to whether health care providers discuss weight-related issues with them. The major research questions that are addressed include the following: (1) How does the prevalence of dieting and disordered eating differ across body mass index (BMI) (calculated as weight in kilograms divided by the square of height in meters) and sociodemographic factors among adolescents? (2) What are the psychosocial and behavioral correlates of dieting and disordered eating among adolescents in this population? and (3) Do adolescents report having discussed weight-related issues with their health care providers?
The study population included a nationally representative sample of 6728 adolescents in grades 5 to 12 who completed the Commonwealth Fund surveys,16,17 which were conducted by Louis Harris and Associates for the Commonwealth Fund. Race/ethnicity was reported as follows: white (non-Hispanic), 53.7%; black (non-Hispanic), 14.1%; Hispanic, 9.2%; Asian, 4.2%; other, 2.7%; and unknown/missing, 16.1%.
The survey was completed anonymously between December 1996 and June 1997. A 2-stage sampling strategy with oversampling of some ethnic groups was used to obtain a representative sample. Youth were selected from a nationally representative cross section of 265 public, private, and parochial schools and an oversampling of 32 urban schools. Also included were 218 youth who had dropped out of high school. Louis Harris and Associates collected data from students in compliance with participating schools' internal review processes regarding informed consent. For further details on the study design, the reader is referred to Schoen et al.16,17
Respondents self-reported their weights and heights, and BMI values were calculated. Using sex-specific and age-specific cutoff points based on reference data from the first National Health and Nutrition Examination (NHANES I)18,19 and different age and sex groups, respondents were classified by BMI as severely overweight (≥95th percentile); moderately overweight (85th to <95th percentile); nonoverweight (15th to <85th percentile); and underweight (<15th percentile).20 Youth with extremely low BMI values (<10th percentile) or extremely high BMI values (>50th percentile) were excluded (n=15) from analyses with BMI, because these values suggest that reported heights and/or weights were invalid. (BMI values were missing for 842 girls and 839 boys who did not report their heights and/or weights.)
Dieting was assessed with the question "Have you ever been on a diet?" Reasons for dieting were assessed with the question "Why were you dieting?" Possible responses included "for health reasons"; "to lose weight because you think you would look better"; "because your doctor or nurse said you should diet"; "because a coach or sports instructor said you should diet"; and "don't know." In this study, disordered eating behaviors refers to self-reported binge-purge cycling. Specifically, respondents were asked "Have you ever binged and purged (which is when you eat a lot of food and then make yourself throw up, vomit, or take something that makes you have diarrhea) or not?" Frequency of disordered eating was assessed for those reporting that they engaged in bingeing and purging behaviors with the question "How often did/do you binge and purge?" Perception of weight status was assessed with the question "Right now, how would you describe yourself?" Possible responses included "about the right weight," "overweight," "underweight," and "don't know." Importance of "not being overweight" was assessed on a 4-point Likert scale ranging from
"very important" to "not at all important."
Self-esteem was assessed with the Rosenberg 10-item self-esteem scale.21 On a 4-point Likert scale, respondents indicated how strongly they agreed with statements such as "I feel I do not have much to be proud of" or "On the whole I am satisfied with myself." Scores were trichotomized into low (<25), moderate (25-34), and high (>34). Depression was assessed with a 14-item modified version of the Children's Depression Inventory.22,23 For each of the 14 items, respondents were asked to choose 1 sentence of 3 that best described them in the past 2 weeks (eg, "I am sad once in awhile"; "I am sad many times";
"I am sad all of the time"). Scores were categorized as low (none or fewer than 8 depressive symptoms), moderate (9-12 depressive symptoms), and high (13 or more depressive symptoms). Suicidal ideation was based on 1 of the items in the depression scale for which the 3 options were "I do not think about killing myself"; "I think about killing myself but I would not do it"; and "I want to kill myself." Stress was assessed with a 10-item stressful life-events scale. Adolescents were asked whether 10 potentially stressful situations had happened to them in the past year (eg, moving to a new home, death of a family member, changing to a new school, and any other stressful event). Scores were categorized as low (0 or 1 stressful event), moderate (2-5 stressful events), and high (6 or more stressful events). Substance use included questions on tobacco, alcohol, and other illegal drug use. For tobacco use, adolescents were categorized into 3 groups: never (never smoked, tried 1-2 cigarettes to see what they were like, or quit), occasional (I smoke a cigarette sometimes), or heavy (at least several cigarettes in the past week to a pack or more in the past week). Similar categories were used for alcohol use: never (never, 1-2 sips), occasional (every once in awhile), or heavy (at least once a month, at least once a week). Other illegal drug use was assessed with the question "Have you used any illegal drugs in the past month?" Level of physical activity was based on the question "How often do you do things that require a lot of physical activity and exercise, like playing sports, jogging, swimming, dancing, or other things?"
A list of health-related topics was presented and respondents were asked to mark off all of the topics that "any doctor or health professional has discussed with you." Topics included in this analysis were eating disorders such as anorexia, bulimia, or compulsive overeating; good eating habits; exercise; and the ideal weight for size and age.
Grade in school was dichotomized into grades 5 to 8 and grades 9 to 12 for most analyses. Ethnicity/race was categorized as follows: white non-Hispanic, black non-Hispanic, Hispanic, and other. Socioeconomic status (SES) was calculated using responses to questions on parental level of education (highest level of either parent) and on the economic situation of the family (ie, whether the family had enough money for food, clothing, and basic living costs). Classification generally went in accordance with educational level: low (no education beyond high school), moderate (some post–high school education, or college graduation without being "well off" financially), and high (postgraduate education, or college graduation and being "well off" financially). When data on educational level of both parents were missing, data on the family economic situation were used.
The nationwide survey of students conducted by Louis Harris and Associates for the Commonwealth Fund used a 2-stage sampling strategy with oversampling of some units to ensure a representative sample. Two statistical consequences of the sampling plan were incorporated into the analysis: (1) responses were weighted inversely to the probability of selection; and (2) classes were included in the analyses as random effects to allow for correlation of student responses within classes. All analyses were performed using the Statistical Analysis System.24
Bivariate relationships were examined using cross-tabulations. Row or column percents were examined to discern patterns of association. Independence was tested by the χ2 statistic deflated to reflect the extra variance coming from classes. The deflation consists of dividing a χ2 value by the estimated design effect,25 resulting in a less significant P value.
These associations were then examined in weighted multivariate logistic regression analyses that included class as a source of extra variation. We examined the associations of BMI and sociodemographic variables (grade, ethnicity, and SES) with dieting and disordered eating behaviors. In the analysis of variance tables, a significant F test for a variable legitimated examination of individual comparisons between categories of that variable. Individual comparisons were tested for significance either by the P value attached to the coefficient or by the noninclusion of 1 in the 95% confidence interval of the particular odds ratio. Further, we investigated the association of a range of psychosocial and behavioral variables (eg, self-esteem and depression) with dieting and disordered eating, adjusted for BMI and sociodemographic variables. For predictor variables, the category of lowest risk (eg, few depressive symptoms) was usually taken as the reference group.
Between 15% to 16% of the population was mildly overweight and between 8% to 9% of the population was severely overweight (Table 1). Although sex differences in the prevalence of being overweight were small, girls were more likely to perceive themselves as overweight than boys. About two thirds of the girls and one half of the boys responded that it was "very important" to them to not be overweight.
Almost half of the girls (45.4%) reported that they had at some point been on a diet, compared with 20.2% of the boys (Table 1). Respondents who reported that they had been on a diet were asked why they had dieted. Reasons cited for dieting were as follows: to improve health (girls, 17.6%; boys, 27.7%); to look better (girls, 88.5%; boys, 62.2%); because of a physician's or nurse's suggestion to diet (girls, 7.1%; boys, 14.2%); because of a coach's or sports instructor's suggestion to diet (girls, 3.9%; boys, 22.3%); and because of a parent's suggestion to diet (girls, 14.5%; boys, 13.6%). For both girls and boys, the most frequently listed reason for dieting was to look better.
Disordered eating (binge-purge cycling) was reported by 13.4% of the girls and 7.1% of the boys. The frequencies with which these behaviors were reported were as follows: less than a few times a week (girls, 2.4%; boys, 1.7%); a few times a week (girls, 2.7%; boys, 1.1%); once a day (girls, 2.2%; boys, 1.5%); and several times a day (girls, 6.7%; boys, 2.6%). Therefore, 8.9% of the girls and 4.1% of the boys reported engaging in binge-purge behaviors at least once a day.
Among girls, there were strong direct associations between BMI and both dieting and disordered eating behaviors, and these associations remained significant after adjusting for sociodemographic characteristics (Table 2 and Table 3). Among boys there was also a direct association between BMI and dieting behaviors, and this association remained strong in multivariate analyses. With regard to disordered eating, the pattern was less clear-cut among the boys; prevalences were highest among boys with the highest BMI values, followed by boys with the lowest BMI values. Severely overweight boys were at increased risk for disordered eating compared with nonoverweight boys after controlling for sociodemographic characteristics.
Younger girls (grades 5-8) were significantly less likely to engage in dieting and disordered eating than older girls (grades 9-12), and this association remained statistically significant in the multivariate analyses (Table 2 and Table 3). Percentages of girls who reported dieting and disordered eating were also examined within the specific grades. Dieting was reported by 31.1% of the 5th-grade girls and increased monotonically to 62.1% among 12th-grade girls. The largest jump was found between 8th graders (39.5%) and 9th graders (52.7%). Disordered eating also increased with grade level, but the rise was not monotonic, suggesting that the younger grades may have had some difficulties in understanding the question. Percentages of girls reporting disordered eating by grade level in school were as follows: 5th grade, 11.3%; 6th grade, 9.2%; 7th grade, 8.8%; 8th grade, 11.5%; 9th grade, 16.9%; 10th grade, 17.6%; 11th grade, 16.7%; and 12th grade, 17.0%. Among the boys, associations between grade and dieting and disordered eating were weak and inconsistent in their directions and were therefore not examined by specific grades. Results from multivariate analyses suggested that younger boys were at greater risk than older boys for disordered eating.
The prevalence of dieting was highest among white non-Hispanic girls and lowest among black non-Hispanic girls (Table 2). The prevalence of disordered eating was highest among Hispanic girls and lowest among black non-Hispanic girls. In multivariate analyses, black non-Hispanic girls were found to be significantly less likely to diet than white non-Hispanic girls. Other associations between ethnicity and dieting and disordered eating were not statistically significant among the girls (Table 3). Among the boys, there were no associations between ethnicity and dieting, but nonwhite boys (black non-Hispanic, Hispanic, and other) were at significantly greater risk for disordered eating behaviors than white boys.
Among girls, large differences in prevalences of dieting and disordered eating across socioeconomic levels were not noted, although rates tended to be lower among girls with high SES. Associations with SES were not statistically significant in multivariate analyses for the girls. Among boys, disordered eating was highest among those with low SES (Table 2 and Table 3).
In univariate (Table 4) and multivariate (Table 5) analyses for both girls and boys, strong and statistically significant associations were found between dieting and disordered eating and low self-esteem, high levels of depression, suicidal ideation, and high levels of stress.
Alcohol and drug use were directly and significantly associated with both dieting and disordered eating (Table 4) among girls and boys, and these associations remained statistically significant after controlling for BMI and sociodemographic characteristics (Table 5). As measured by the odds ratios, associations between substance use behaviors and disordered eating seem to be stronger than associations between substance use and dieting. Tobacco use was directly and significantly associated with both dieting and disordered eating among girls, but not among boys.
Finally, associations were examined between frequency of physical activity and both dieting and disordered eating behaviors. In unadjusted analyses, statistically significant inverse associations were found; girls engaged in more frequent physical activity were less likely to report dieting or disordered eating (Table 4). However, these associations were weakened and not consistently statistically significant after controlling for BMI and sociodemographic characteristics (Table 5). Among boys, associations with physical activity were not apparent in either univariate or multivariate analyses.
Almost half of the adolescents (girls: 1420 [53.1%]; boys: 1118 [43.7%]) reported that a physician or other health professional had at some point discussed nutrition (good eating habits) with them. Discussions on weight-related issues (the ideal weight for someone your size and age) were reported by 47.4% of the girls and by 37.7% of the boys. Discussions on physical activity (the importance of exercise) were reported by 41.1% of the girls and 40.4% of the boys. Discussions with a health care provider on eating disorders were reported by lower percentages of girls (23.8%) and boys (15.0%).
The high prevalence of obesity among children and adolescents is a major public health issue. In this study, almost one quarter of the respondents were overweight, based on self-reported heights and weights, which tend toward underestimations of BMI values.26- 28 The findings are similar to those reported by Troiano et al,29 in which 22% of a national sample of adolescents was determined to be overweight (≥85th percentile) using actual weight and height measurements. Although sex differences in prevalence rates of obesity were small, girls were much more likely than boys to perceive themselves as overweight, engage in dieting and disordered eating, and indicate that "not being overweight" was very important. Field et al30 have similarly found that girls are more likely than boys to perceive themselves as overweight. The challenge facing health professionals is how to reach youth with interventions aimed at preventing obesity without leading to excessive weight preoccupation and unhealthy weight control practices.
The high prevalence of reported dieting among the girls is disturbing; almost half of the girls in the study population reported dieting. Other population-based studies have also found high rates of dieting among adolescent girls.6 Although overweight girls were significantly more likely to report dieting than nonoverweight girls, dieting was also prevalent among nonoverweight girls. Previous research has shown that dieting is associated with several negative outcomes and has been identified as a risk factor for the later development of eating disorders.1,15 However, it is important to note that the question on dieting in the Commonwealth Fund surveys was very general. Dieting may actually be viewed as a desirable behavior if it refers to healthful behaviors such as moderate fat reduction or increased consumption of fruits and vegetables. In a previous study in which adolescents participating in focus groups were asked about their interpretations of the terms dieting and binge eating, a range of different responses were given that were suggestive of both healthful and unhealthful practices.31
The relatively high prevalence of reported disordered eating (binge-purge cycling) among adolescents in the study population is of concern. It is interesting to note that rates tended to be highest among overweight youth. This may be due to the nature of the question, in that it jointly assessed bingeing and purging and did not allow for the assessment of purging behaviors without bingeing. However, in previous studies we also found that overweight girls were at increased risk for disordered eating behaviors such as self-induced vomiting (without bingeing).8 Therefore, screening questions on disordered eating behaviors should not be limited to youth who are underweight, but should be directed at all youth. Furthermore, in working with overweight youth, it is important to be aware of their high risk for engaging in disordered eating behaviors and to focus on the long-term adoption of healthy eating and physical activity behaviors rather than on dieting, which tends to be a short-term behavior.
It is important to examine prevalence rates of dieting and disordered eating across sociodemographic characteristics such as ethnicity and SES to increase our understanding of the role of sociocultural factors in the onset of these behaviors, and to increase our ability to identify subgroups of the population at increased risk who need to be targeted for intervention. As in other studies,4,8 prevalence rates of dieting behaviors were lowest among black girls. The findings suggest that black girls may experience lower levels of body dissatisfaction than white girls. However, this should not be interpreted as meaning that overweight black girls are not concerned with their weight. In a previous study in which we conducted in-depth interviews with overweight black and white girls, we found weight concerns among both groups.32 The higher risk for disordered eating among nonwhite boys is puzzling and raises questions of validity, yet it has been documented in other large population-based studies.4,8 The weak associations between SES and both dieting and disordered eating behaviors among girls and the increased risk for disordered eating among boys with low SES are noteworthy. Although these behaviors have typically been viewed as being more common among youth from high socioeconomic backgrounds than among youth from low socioeconomic backgrounds, data from population-based studies have revealed a different picture. Using data from the Minnesota Adolescent Health Survey, Story et al4 found that girls with high SES dieted more than girls with low SES, but were less likely to report vomiting, use of laxatives, and binge eating. Using data from the Connecticut Youth Survey, we did not find statistically significant associations between SES and dieting, but did find that youth with low SES were at increased risk for disordered eating behaviors including self-induced vomiting, diet pills, laxatives, and diuretics.8
Disordered eating behaviors and, to a lesser extent, dieting behaviors were strongly associated with a range of psychosocial concerns (eg, low self-esteem, high depression, and suicidal ideation) and substance use behaviors. This pattern suggests that these behaviors may be symptoms of a greater underlying psychopathology or stressful living situation. This is not to say that all youth who are dieting or engaging in disordered eating behaviors have underlying serious problems. It does, however, suggest that reasons for engaging in these behaviors be explored by a health professional. It also suggests that for some youth, reasons for engaging in these behaviors extend beyond sociocultural pressures to be thin. The interaction of sociocultural norms (ie, regarding thinness and expectations for women), personal characteristics (eg, personality and body size), and familial dynamics (eg, overall relationships and food- and or weight-related attitudes) play a role in the onset of dieting and disordered eating behaviors, and the relative contribution of each factor probably differs among youth.
In this study, correlations were found between substance use behaviors and both dieting and disordered eating. In a previous study, using data from the Youth Risk Behavior Survey, we found associations between substance use and disordered eating (vomiting and dieting pills) of a similar magnitude to those in this study, but did not find consistent associations between substance use behaviors and "other weight loss methods."33 Patterns of covariation of behaviors warrant further study because they provide insight into the causes of the behavior, increase our understanding of the differences and similarities between less severe and more severe weight control behaviors, and assist in the identification of youth to be targeted for intervention.
Associations between levels of physical activity and dieting and disordered eating were explored and were found to be inconsistent and relatively weak. On one hand, adolescents who are physically active might be expected to avoid these behaviors; they may avoid dieting and try to control their weight by adopting an overall healthy lifestyle. On the other hand, these adolescents may be preoccupied with their weight and may engage in a variety of behaviors aimed at weight control including exercise, dieting, and more extreme disordered eating behaviors. Clearly, this is an area worth further study, ie, further delineating the role of physical activity in preventing the use of unhealthy weight control behaviors and examining motives for being physically active and the associations between these motives and use of dieting and disordered eating behaviors.
The Commonwealth Fund surveys provided a unique opportunity to examine youth perceptions of issues raised by their health care providers. To the best of our knowledge, this type of information has not been collected in other national surveys of youth. It was encouraging that about half of the youth reported that a health care provider had discussed nutrition and weight issues with them. While the content of these discussions remains unclear, at least the youth remembered that these issues had been discussed. It is important to explore the content of these discussions to ascertain that issues are being addressed in a manner most likely to lead to a positive body image and to appropriate eating and physical activity behaviors. Efforts are also needed to ensure that the other half of the adolescent population, who reported that a health care provider had not discussed nutrition and weight issues with them, also receive appropriate and routine guidance by health care providers. One of the Healthy People Year 2000 health objectives is to "increase to at least 75% the proportion of primary care providers who provide nutrition assessment and counseling and/or referral to qualified nutritionists or dieticians."34
A major strength of the study was the use of a nationally representative sample of youth, thus allowing for the extrapolation of findings from the study population to a larger population of youth. Another strength of this study is that the survey included robust measures to assess psychosocial concerns such as self-esteem, depression, and stress. Often in comprehensive surveys of adolescent health, complete scales are not included owing to constraints on the number of items that may be included. Nevertheless, in drawing conclusions from the study, the limited nature of the questions assessing dieting and disordered eating needs to be taken into account. The types of dieting behaviors used by respondents remain unclear; dieting behaviors may have included healthy strategies, such as increasing fruit and vegetable consumption, or unhealthy methods, such as skipping meals or using diet pills. With regard to disordered eating, the focus was only on binge-purge cycling. Information on other types of behaviors would have been useful. In examining associations between BMI and dieting and disordered eating behaviors, it should be noted that BMI values were based on self-reported heights and weights. Past studies have shown that overweight people tend to underestimate their weight.26,35- 38 Furthermore, overweight youth may be more likely not to report their heights or weights. The effect of missing or inaccurate BMI data for a nonrepresentative sector of the respondents on associations between BMI and dieting and disordered eating is unclear.
The high rates of dieting and disordered eating behaviors, coupled with the high prevalence of obesity found in this and previous studies,4,8 indicate a clear need for interventions aimed at the primary and secondary prevention of weight-related disorders. The large scope of the problem and the complexity of the issues at hand indicate that there is a need for multiple interventions at the individual and familial level (eg, within clinical practices), at the group level (eg, within school settings), and at the community or larger societal level (eg, changes in the physical and social environment). The high prevalence of weight-related concerns suggests that all youth need to be reached with appropriate interventions; however, special attention needs to be directed toward youth at greatest risk for engaging in disordered eating behaviors, such as overweight youth; youth engaging in other high-risk behaviors, such as substance use; and youth with psychological concerns, such as low self-esteem and depressive symptoms.
Accepted for publication November 12, 1999.
We thank Elizabeth Simantov, PhD, senior research analyst, from the Commonwealth Fund, for her invitation to analyze these data and for all of her assistance.
Corresponding author: Dianne Neumark-Sztainer, PhD, MPH, RD, Division of Epidemiology, School of Public Health, University of Minnesota, 1300 S Second St, Suite 300, Minneapolis, MN 55454 (e-mail: firstname.lastname@example.org).
Neumark-Sztainer D, Hannan PJ. Weight-Related Behaviors Among Adolescent Girls and BoysResults From a National Survey. Arch Pediatr Adolesc Med. 2000;154(6):569-577. doi:10.1001/archpedi.154.6.569