Objectives. Adolescent obesity is becoming an increasing public health problem. This study determines: 1) differences in teen and parental report of obesity, 2) amount of misclassification using body mass index (BMI) from self-reported versus measured height and weight as an indicator of obesity, and 3) whether misclassification varies by gender and socioeconomic status.
Design. Weighted data from 15 483 baseline (T1) youth and parental interviews from the National Longitudinal Study of Adolescent Health were used. Seventy-four percent of teens were reinterviewed 1 year later (T2). Parents reported socioeconomic status indicators and whether their teen was obese. Teens reported height, weight, and weight perception. BMI was calculated from both self-reported height and weight at T1 and T2 and from measured height and weight at T2. Those with a BMI ≥95% corrected for age and gender were considered obese.
Results. At T1, nearly one half of teens (47%) reporting they were very overweight were not obese by BMI. For teens obese by BMI, 19.6% were reported to be obese by both parent and teen, 6.4% by teen only, 29.9% by parent only, and 44.2% by neither teen nor parent. For those with persistent obesity, teen and/or parental report failed to identify more than one third (34%) as obese; 23.4% were identified by both teen and parent report, 5.4% by teen report only, and 37.2% by parent only. At T2, the correlation between BMI calculated from self-reported versus measured height and weight for the overall population was very strong (r = .92). Specificity of obesity status based on self-reported BMI, compared with obesity status based on measured BMI was .996; sensitivity, .722; positive predictive value, .860; and negative predictive value, .978. Overall, 3.8% of teens were misclassified using self-report measures. Girls were no more likely than boys to be misclassified as obese using BMI from self-reported height and weight.
Conclusions. Parental report is a better indicator of obesity than teen report of weight status, but parental and teen reports are both poor predictors of adolescent obesity. Using BMI based on self-reported height and weight correctly classified 96% as to obesity status. Thus, studies can use self-reported height and weight to understand teen obesity and its correlates/sequelae.
- BMI =
- body mass index •
- Add Health =
- National Longitudinal Study of Adolescent Health •
- T1 =
- baseline •
- T2 =
- 1 year later
Although we are a society preoccupied with thinness, obesity is now recognized as an increasing public health problem throughout the life course. Obese children are more likely to become obese adolescents, and obese adolescents are more likely to become obese adults.1 Obesity has numerous health-related sequelae, as well as social consequences, such as lower wages, less likelihood of marriage, less education, and stigma.2 ,3 In addition, social forces, such as gender and socioeconomic status, have been associated with an increased likelihood of obesity among youth.4–6
Despite the vast literature on obesity, its definition has not been clearly established. Body mass index (BMI) is one of the most commonly used measures to define obesity. Recently, an expert consensus panel convened by the Maternal and Child Health Bureau suggested that BMI ≥95% for age and gender should define obesity.7 However, even with a consensus on the definition of obesity, studies vary with regard to use of measured versus self-reported height and weight to calculate BMI.
Studies of adults have demonstrated a high correlation between self-reported and measured height and weight,8 ,9 although use of BMI based on self-report indexes to define obesity as a categorical variable is considered unreliable in adults.8Few studies have assessed the validity of adolescent reports of height and weight or categorical definitions of obesity based on BMI cutoff points, and none have done so for a nationally representative population.10–15 Although some studies suggest that adolescent reports of height and weight are valid,11others have raised concern about the accuracy of adolescent reports of both height and weight.10 12–15 Previous studies of adolescents looking at BMI cutoff points have used the 85th percentile, which overestimates the proportion of teens who would meet criteria for obesity as defined by the Maternal and Child Health Bureau Expert Consensus Panel.7 Thus, the accuracy of adolescent reports of height and weight remains in question, as does the accuracy of use of BMI from self-reported measures as an indicator of obesity. Whether inaccuracies in reporting of height and weight are caused by misperception or conscious intent is also not clear. Indeed, the relationships among adolescents' perceptions of their weight status, their parents' perception of the teen's weight status, and their actual obesity status as defined by medical criteria have not been studied. The objective of the current study was to determine, in a nationally representative sample, differences in teen and parental report of obesity, amount of misclassification using BMI from self-report versus measured height and weight as an indicator of obesity, and whether misclassification varies by gender and socioeconomic status.
Study Design and Sample
This article uses weighted data from wave 1 and wave 2 of the in-home survey of the National Longitudinal Study of Adolescent Health (Add Health), a nationally representative, school-based prospective study begun in 1994 that focuses on the health and social contexts of American teenagers in grades 7 through 12.16 ,17 Wave 1 contained over 20 000 in-home interviews. Of these participants, 18 922 (91%) were assigned a sample weight to adjust for the probability of being chosen. Of these participants, 82% (n = 15 483) had a biological parent, stepparent, foster parent, or adoptive parent who was the parental respondent for a parental in-home interview. Information from these 15 483 linked parent and teen interviews was used as baseline (T1) data for this study. Seventy-four percent (n = 11 495) of youth interviewed at T1 completed a follow-up interview 1 year later (T2). A description of the baseline and follow-up study populations can be found in Table 1.
Age was derived from the date of interview minus the date of birth. Subjects indicated racial/ethnic identities. Responses were condensed to final categories of white non-Hispanic, black non-Hispanic, Hispanic, Asian/Pacific Islander, and other.
Self-Reported Height and Weight
Adolescents were asked to report their height in feet and inches and their weight in pounds at T1 and T2.
Measured Height and Weight
During the T2 in-home interview, teens were weighed with clothes but not shoes on a spring scale brought by the interviewer. The interviewer also measured the teen with a seamstress-type tape measure to determine height in feet and inches. Interviewers were trained in the methods of obtaining both height and weight according to the protocol developed by Add Health staff.
Calculation of BMI was possible from self-reported height and weight at T1 and T2, hereafter referred to as T1 self-reported BMI and T2 self-reported BMI, and from measured height and weight at T2 only, hereafter referred to as measured BMI. BMI was calculated as the weight in kilograms divided by the height in meters squared.
Measures of Obesity
Obesity was defined as BMI ≥95% for age and gender. Obesity at T1 and T2 based on self-reported BMI is referred to as obese by self-report BMI. Obesity status defined by measured BMI (available only at T2) is hereafter referred to as obese by measured BMI. The 95% was chosen as the cutoff point, because this criterion is believed to provide better specificity for obesity than the lower percentile cutoffs used among adults and is currently recommended by an expert panel for studying obesity among children and youth.6 ,7 ,18Cutoff points for the 95% were determined from National Health and Nutrition Examination Survey I data for those 18 year of age and older19 and from a study combining data from >60 000 healthy children and youth from 9 different studies for those 17 years of age and younger.20 Cutoff points for the 95% were specific for age and gender. Persistently obese teens are adolescents who were obese by self-reported BMI at T1 and T2 and by measured BMI at T2.
Adolescent Self-Report of Obesity
A single item in the adolescent interview assessed weight perception on a 5-point Likert-type scale ranging from very underweight to very overweight. Those endorsing the very overweight category were considered obese by self-report. These data were available at T1 and T2.
Parental Report of Adolescent Obesity
The parental respondent was asked in 6 separate dichotomous questions to report whether the adolescent had a particular health problem. Obesity was the first health problem assessed. If the parental respondent answered yes, the teen was considered obese per parent report. These data were only available at T1.
Misclassification was defined as a discordance between obesity status determined by measured versus self-reported BMI at T2. Teens who were not classified as obese by self-reported BMI but were obese by measured BMI, or those who classified themselves as obese when they did not meet obesity criteria by measured BMI were considered to be misclassified.
Measures of Socioeconomic Status
Household income was assessed in the parent interview. Parental respondents reported, in thousands of dollars, 1994 total household income before taxes from all sources. A reported income of zero was considered missing data. Income was then classified into 5 ordered levels based on 1994 federal poverty thresholds adjusted for household size. Federal poverty levels were taken from household income tables available from the US Census Bureau. For a more conservative estimate of poverty, the lowest income category was defined as below 1.5 times the adjusted poverty threshold.21 Income categories 2 through 5 were defined as follows: 1.5 to <2.5 times the poverty threshold; 2.5 to 4 times the poverty threshold; >4 times the poverty threshold but not in top 5% of households; and top 5% of household incomes.
Information about parental education was also derived from the parental interview. Items assessing the parental respondent's education level and gender, and education level of the parental respondent's partner were combined to create variables corresponding to educational level of the highest educated parent. This variable, like the income variable, had 5 ordered categories. Categories were: less than a high school degree; high school degree, graduate equivalency diploma, or vocational training instead of high school; vocational training after high school or some college; college graduate; and professional training beyond college.
Add Health sample weights, derived to compensate for differences in selection probabilities, were used in analyses.17 ,22 To decrease the likelihood of a type I error attributable to the large weighted sample size, these weights were recalculated for the entire wave 1 dataset, such that the total weighted n equaled the sample size.23 Descriptive frequencies and means were generated with SPSS for Windows, Version 9.0 (SPSS, Chicago, IL). SUDAAN (Research Triangle Institute, Research Triangle Park, NC) was used for all bivariate and multivariate analyses to control for the complex cluster design of Add Health. χ2tests were used to determine significant differences in proportions among categorical variables, Student's t tests were used for determining grouped differences for continuous variables (differences in self-reported vs measured height, weight, and BMI). Linear polynomial testing was performed to look for linear trends in the effects of education and income on dependent variables. Logistic regression modeling was performed for variables found to be significantly related to misclassification. Unadjusted and adjusted odds ratios with 95% confidence intervals are presented for logistic regression analyses. Correlational analyses using Pearson correlation coefficients were performed with SPSS for Windows, Version 9.0, because SUDAAN does not perform this type of analyses. Although a significance level of P < .05 was chosen for analyses with SUDAAN, to decrease the likelihood of a type I error, P < .001 was chosen as a significance level for analyses with SPSS, because this statistical software package does not account for the complex cluster design of Add Health.
Correlations between measured and self-reported anthropomorphic indices were very strong. Correlation between measured and self-reported height was .94, between measured and self-reported weight was .95, and between measured and self-reported BMI was .92 (P < .0005 for all). Race, parental education, and household income were not significantly associated with differences in self-reported versus measured anthropomorphic indices. Although there was no gender difference in self-reported versus measured height, there were significant gender differences for both weight and BMI. Although adolescent boys and girls both tended to underreport weight, adolescent girls underreported weight to a greater degree than adolescent boys (mean difference in measured vs reported weight = 1.02 kg for girls vs .19 kg for boys; P < .00005). Although BMI from self-reported measures underestimated BMI from measured height and weight for girls, BMI based on self-reported measures overestimated BMI for adolescent boys, compared with measured BMI (mean difference in measured vs reported BMI = .27 for girls vs −.03 for boys; P < .00005).
At T2, 9.7% were obese by measured BMI, 8.1% by self-reported BMI, and 4.7% were persistently obese. Overall, 3.8% (unweightedn = 341) of teens were misclassified using BMI from self-reported height and weight as opposed to BMI from measured height and weight at T2. A description of this population can be found inTable 1. More than two thirds of those who were misclassified (70.4%) were teens who were obese by measured but not self-reported BMI at T2. None of the teens with persistent obesity were misclassified. Test characteristics of the various self-reported measures of obesity (specificity, sensitivity, positive predictive value, and negative predictive value) are reported in Table 2. Gender-specific values for specificity, sensitivity, positive predictive value, and negative predictive value of the various indicators are presented in Table 3. Specificity was high for all measures of adolescent obesity. However, sensitivity varied greatly. Adolescent self-report of obesity status had the lowest sensitivity (.235 at T1). Gender-specific analyses suggested that girls are better informants than are boys. However, neither gender nor race was significantly associated with misclassification assessed as a single category or with type of misclassification, compared with controls (those who correctly classified themselves). Logistic regression modeling revealed that both lower parental education and lower household income were significantly associated with misclassification (Table 4). However, in a composite model including both indicators of socioeconomic status, neither remained significant (Table 4). Logistic modeling by type of misclassification suggested that socioeconomic indicators were not related to misclassification as nonobese, but that those of lower socioeconomic status were more likely to misclassify themselves as obese than those of higher socioeconomic status. However, in the composite model including both indicators of socioeconomic status, this relationship did not remain significant.
Subjective Obesity and Informant Differences in Obesity
Although 9.7% of these youth were obese by measured BMI at T2, only 3.4% believed they were very overweight. Adolescent girls were more likely to report that they were very overweight than adolescent boys (2.0% boys vs 4.9% girls; P < .00005). Of those who were persistently obese, 60.6% were identified by parent report and 28.8% by adolescent report. Over one third (34%) were not identified by teen or parent report.
At T1, 7.2% were obese by self-reported BMI, 6.5% were obese by parental report, and 3.5% reported subjective obesity. Nearly one half of those who considered themselves very overweight (47.4%) were not obese by self-reported BMI and nearly three quarters of those who were obese by self-reported BMI (74.0%) did not consider themselves very overweight. Although obesity by self-reported BMI was more common among boys at T1 (8.2% boys vs 5.6% girls; P < .00005), adolescent girls were more likely to report that they were very overweight (2.0% boys vs 5.0% girls; P < .00005) and to have a parent report that they had a problem with obesity (5.9% boys vs 7.4% girls; P = .012). Over one half of those who were obese by BMI at T1 (51.5%) had an obese parent per the parental respondent. Teens from families with obese parents were more likely to be identified as obese by a parent than were teens from families without an obese parent (15.4% vs 4.5%; P < .0005). Data on differences in adolescent versus parental reports of adolescent obesity are presented in Table 5. Among the total T1 population, 8.2% of teens were identified as obese by parent or teen report. Among those who were obese by self-reported BMI at T1, nearly one half (49.5%) were identified by parent report and slightly more than one quarter (26.3%) by adolescent self-report. However, nearly one half of those who were obese by self-reported BMI (44.2%) were not identified by teen or parental report.
This study answers an important question for those studying adolescent obesity—self-reported height and weight can be used to calculate BMI and provide a categorical measure of obesity among teens using the 95% cutoff points. Using this method, over 96% of youth in this large, nationally representative sample were correctly classified. Sensitivity, specificity, positive predictive value, and negative predictive value of obesity by self-reported BMI were all high. This study also showed that adolescent girls were no more likely than boys to be misclassified as obese using self-reported data. No independent correlates of misclassification were identified in the final logistic regression models.
In addition to verifying that BMI from self-reported height and weight is a valid indicator of obesity, these data also provide new important insights into the duality of adolescent obesity—it is both a subjective and objective phenomenon. How this duality impacts on the understanding of obesity by both health care professionals and those impacted by the disease are important areas for discussion and future research. The reasons that adolescents seek treatment for weight concerns are varied. In contrast to adults, referral of adolescents to health care providers is often initiated by parents or by other concerned adults.24 These same adults are also often critical participants in the assessment, diagnosis, and treatment of youth. Thus, although giving adolescents their own voice is a crucial part of caring for youth, understanding and integrating information from multiple informants across diverse environments is an equally important issue in their care.
Research on informant effects comes almost entirely from the literature on emotional and behavioral problems, rather than on health problems such as obesity. Studies of informant differences in the reporting of mental health problems have found that reports of informants show significant agreement, but that levels of agreement vary across different pairs of informants, and self-reports consistently show low agreement with all other informants, including peers.24 ,25Further, cross-informant agreement may be stronger among informants reporting on conduct or externalizing problems, compared with affective or internalizing problems.24–26 The literature suggests that youth are more likely to report internalizing problems such as depression,26 while parents may be likely to report externalizing problems, such as conduct problems.27
How the current findings on informant differences in obesity fit into this literature is unclear but intriguing. Youth were accurate informants with respect to objective measures of height and weight. However, despite the accuracy of teen reports of height and weight, parents and teens were poor informants with respect to subjective obesity. Parents and teens failed to identify teens that were in fact obese, and incorrectly identified teens as obese, who were not obese. These teens define 2 potential risks groups—those who may not receive needed treatment, and those who may receive inappropriate treatment or may suffer distorted body image. Thus, education of both parents and teens, which addresses both the medical and psychosocial aspects of obesity, may be a necessary prerequisite to medical treatment. Acceptance of the diagnosis and understanding of obesity as a serious health risk may be necessary before actual weight loss protocols can be effective.
Although teens and parents both failed to identify obesity correctly by subjective report, parents were more accurate than teens. This suggests that obesity is akin to externalizing problems, such as conduct problems, which parents are more likely to report and youth are more likely to deny. Indeed, obesity is an observable, external feature. However, as found in the current study, obesity is also a subjective, emotional experience of one's body and body image, and has been significantly associated with internal states, such as lower self esteem28 and experiences of social stigma,2 ,3 phenomena for which self-reports are thought to be more reliable.27 Thus, obesity does not fall neatly into objective versus subjective categories. Although it is clearly an objective, measurable phenomenon (eg, BMI), obesity is also socially constructed and has inherent psychosocial implications and psychological meaning for the individual. This subtle interplay of objective and subjective factors in the assessment and treatment of obesity is underscored by the discordance between subjective and objective reports of obesity shown in this study. Obese adolescents may be willing to acknowledge feelings of low self-esteem,28but may be unwilling or unable because of denial and/or stigma, to acknowledge the obesity behind these feelings. In contrast, parents may be better, though imperfect, at reporting subjective obesity.
A parent's ability to report a child as obese may reflect biases resulting from the parent's level of distress or psychological symptoms and/or from labeling of a given teen. For example, parent symptoms in general and maternal depression in particular have been found to be positively related to parent report of child symptoms and depression, respectively.25 ,29 Data here suggest that parental report of parent obesity is positively associated with parental report of teen obesity. Whether parental obesity is related to increased accuracy with respect to reporting adolescent obesity is unclear. The impact of parental distress on the subjective identification of teen obesity is also unclear. Because assessment and treatment of obesity often involve parents, how parental weight status, objectivity about weight status, and/or distress regarding weight status interact may be important areas of further inquiry. Parental identification of obesity may be an indicator that the teen's family is concerned about weight status and may be more amenable to intervention than families who do not acknowledge weight problems. Alternatively, parental distress over parental obesity may impact treatment negatively. In addition, discrepancies in adolescent and parental reports of weight problems are certainly a reflection of discrepant perceptions, and may be an indicator of other underlying issues in the parent–teen relationship that can exacerbate the obesity problem.
This study does have several notable limitations. First, Add Health is a comprehensive dataset, but it is school-based and does not include out-of-school youth or school-avoidant youth. Whether obese youth fall into these categories is unclear. A second limitation is that indicators of obesity were not uniformly present in both waves of Add Health. Only adolescent report of height, weight, and weight status were available in both waves of data collection. Parental report of subjective obesity was not available at T2, while measured height and weight were only obtained at this data collection. Third, although an expert panel agreed that BMI ≥95% should define obesity,7 a clear standard of 95% cutoff points has not been developed. In an attempt to use the most up-to-date cutoff points, this study used cutoff points determined by a study of standardized measurements of height and weight from 9 large epidemiologic studies.20 A consensus on the appropriate 95% cutoff points would help standardize the field and make findings across studies more comparable.
Despite these limitations, this study has important findings and implications for future research. The social construction of obesity and its impact on treatment is an important area of future inquiry. Although parents and teens functioned poorly as reporters of adolescent obesity, adolescent self-report of height and weight accurately identified >96% of obese adolescents. Findings from other studies that have used self-reported BMI should be considered valid, and future studies can use self-reported data to understand adolescent obesity, its correlates, antecedents, and sequelae.
This work was funded in part by Project MCJ-MA 259195 Award from the Maternal and Child Health Bureau (Title V, Social Security Act), Health Resources and Services Administration, Department of Health and Human Services, and a Children's Hospital Research Scholar Award, a competitive grant funded by the Trustees of the Children's Hospital, Boston.
This research is based on data from the AddHealth project, a program project designed by J. Richard Udry (principal investigator) and Peter Bearman, and funded by Grant p01-HD31921 from the National Institute of Child Health and Human Development to the Carolina Population Center, University of Carolina at Chapel Hill, with cooperative funding participation by the National Cancer Institute; the National Institute of Alcohol Abuse and Alcoholism; the National Institute on Deafness and Other Communication Disorders; the National Institute of Drug Abuse; the National Institute of General Medical Sciences; the National Institute of Mental Health; the National Institute of Nursing Research; the Office of AIDS Research, National Institutes of Health; the Office of Behavior and Social Science Research, National Institutes of Health; the Office of the Director, National Institutes of Health; the Office of Research on Women's Health, National Institutes of Health; the Office of Population Affairs, Department of Health and Human Services; the National Center for Health Statistics, Centers for Disease Control and Prevention, Department of Health and Human Services; the Office of Minority Health, Centers for Disease Control and Prevention, Department of Health and Human Services; the Office of Minority Health, Office of the Assistant Secretary of Health, Department of Health and Human Services; the Office of Assistant Secretary of Planning and Evaluation, Department of Health and Human Services; and the National Science Foundation.
We thank S. Jean Emans, MD, for her helpful comments on earlier drafts of this manuscript.
- Received May 18, 1999.
- Accepted October 11, 1999.
Reprint requests to (E.G.) Division of Adolescent Medicine, Children's Hospital Medical Center, PAV-2129, 3333 Burnet Ave, Cincinnati, OH 45229. E-mail: goodeθ@chmcc.org
These data are not available from the author. Persons interested in obtaining data files from the National Longitudinal Study of Adolescent Health should contact Francesca Florey, Carolina Population Center, 123 West Franklin St, Chapel Hill, NC 27516-3997. E-mail:
- ↵Barlow SE, Dietz WE. Obesity evaluation and treatment: expert committee recommendations. The Maternal and Child Health Bureau, Health Resources and Services Administration, and the Department of Health and Human Services. Pediatrics. 1998;102(3). URL: http://www.pediatrics.org/cgi/content/full/102/3/e29
- Popkin BM,
- Udry JR
- ↵Bearman PS, Jones J, Udry JR. The National Longitudinal Study of Adolescent Health web site. Available at: http://www.cpc.unc.edu/projects/addhealth/design.html Accessed April 24, 1997
- Must A,
- Dallal GE,
- Dietz WH
- ↵Tourangeau R, Shin H. National Longitudinal Study of Adolescent Health: Grand Sample Weight. Chapel Hill, NC: Carolina Population Center; 1998
- ↵Brogan DJ. Pitfalls of using standard statistical software packages for sample survey data. In: Armitage P, Colton T, eds. Encyclopedia of Biostatistics. New York, NY: John Wiley and Sons; 1998:4167–4174
- ↵Kazdin AE. Informant variability in the assessment of childhood depression. In: Reynolds WM, Johnston HF, eds. Handbook of Depression in Children and Adolescents—Issues in Child Clinical Psychology. New York, NY: Plenum Press; 1994:249–271
- Copyright © 2000 American Academy of Pediatrics