P Thunfors, A Hanlon, B Collins
adolescents, health behavior, obesity, weight perception
P Thunfors, A Hanlon, B Collins. Weight Status Misperception And The Health Behaviors Of Obese Adolescents. The Internet Journal of Pediatrics and Neonatology. 2010 Volume 13 Number 1.
The rate of pediatric obesity has risen significantly in the United States1 to the point at which 17% of children ages 2-19 were obese and an additional 15% were overweight in 2007-20082. This epidemic has become a major public health concern, as childhood obesity has been linked to many adverse health consequences, including juvenile hypertension, elevated cholesterol levels, orthopedic abnormalities, type II diabetes mellitus, gallstones, asthma, and sleep disorders, to name a few3. Prevention programs for children and adolescents have so far had limited success4,5. Thus, research is necessary to examine individual difference characteristics that contribute to the development of pediatric obesity and barriers to its treatment. This study intends to determine whether weight misperception may be an important factor in understanding the high prevalence of adolescent obesity.
Weight underestimation among a multiethnic sample of overweight U.S. adolescents (BMI at or above the 85 th percentile) was found to be common, with 61% of males and 40% of females underestimating their actual weight status6. This finding is indicative of an ongoing trend among many Americans with overweight who have inaccurate self-perceptions of their weights7-10. Furthermore, many parents have been found to underestimate the weight status of both their child and adolescent offspring11-13, which thereby suggests a potential mechanism for transmitting an inaccurate gauge for healthy weight status to their children.
In adults, the tendency to misperceive one’s weight varies substantially across demographic factors, with men, those with lower socioeconomic status, Hispanics, and African-Americans more likely to underestimate their overweight status10,14,15. Similar to adults, there is some evidence that gender, race, and family factors play an important role in adolescent weight perception16,17. However, Yan et al6 found only gender differences and no race or socioeconomic status associations with weight status misperception in their multiethnic sample of adolescents.
This study seeks to add to the current literature by determining associations between weight misperception and specific adolescent health behaviors, such as physical activity and nutrition, among individuals with obesity. Across healthy and unhealthy adolescent populations, weight loss efforts are strongly related to adolescents’ perceptions of themselves as overweight15,18. Therefore, underestimation of overweight status might reduce the likelihood that adolescents with overweight would participate in healthy behaviors that could promote weight loss. This hypothesis suggests a direct perception-behavior relationship.
Alternatively, evidence that self-efficacy predicts adolescent participation in a range of health behaviors, including exercise, smoking, AIDS prevention, and diet19-22, points to self-efficacy as a factor that may mediate the perception-behavior link. Thus, an opposing hypothesis exists such that accurate perception of obesity status could undermine self-efficacy to engage in health behaviors such as physical activity, thereby reducing the likelihood of participation in healthy activities. That is, they may view their overweight status as an impediment to their participation in health behaviors. In either case, an opportunity exists to gain a better understanding of relations between weight perception, self-efficacy, and health behaviors among obese adolescents.
The primary goal of this study was to test two hypotheses that examine associations between weight self-perception and participation in health behaviors among obese adolescents: 1) a direct weight perception-behavior relationship, with weight underestimation expected to decrease the likelihood of participation in health behaviors linked to weight management; and 2) a mediating effect of self-efficacy, such that accurate recognition of one’s obese status is expected to correlate with lower self-efficacy in one’s ability to participate in health behaviors, thereby reducing the likelihood of actual participation in these behaviors. To test these hypotheses, we measured self-reported readiness to participate in four areas of health behavior (i.e., physical activity, healthy diet, weight management, and overall healthy lifestyle) among adolescents who are obese. We also explored effects of hypothesized moderating (i.e., gender and grade level) variables in the relationship between weight perception and health behaviors.
Data were gathered from a large sample of obese adolescents (defined as BMI’s at or above the 95 th percentile2) who were enrolled in the 7 th and 11 th grades in two mostly rural Pennsylvania school districts from academic years 2001-2002 to 2005-2006.
Student data were eligible for inclusion if the participants were obese and had complete survey data (
Participants’ BMI was measured by certified school nurses. No information was provided in the dataset about methods for obtaining heights and weights (e.g., with shoes, with clothing, etc.). Obesity was defined as a BMI at or above the 95 th percentile according to the gender-specific growth charts from the Centers for Disease Control and Prevention2,26. Obese students were selected for inclusion, rather than overweight students (BMI’s at or above the 85 th percentile to the 95 th percentile), to enable a greater degree of weight misperception among the sample that was more appropriate for the study hypotheses.
Nursing and research staff from Wayne Memorial Hospital in Wayne County, Pennsylvania collected questionnaire data, height, weight, and blood pressure as part of the hospital and school district’s annual “Together for Health” program, which was designed to provide students with resources and practical assistance to make healthier lifestyle choices. Every student was invited to participate in the program and each participating student received a health profile detailing the results of their health assessments. Parental consent for the program was implicitly obtained through notification via a letter that was sent home with students and non-refusal. Parents were provided with the hospital’s Notice of Privacy Practices, which informed them of their right to refuse permission for their medical information to be used for the purposes of research. Data collection was also approved by direct student assent. There were no data on rates of refusal. This data collection procedure was approved by the school district because data was collected anonymously, was not tracked at the student level over time, and was only reported in aggregate as it pertained to general patterns in health behaviors across the student population. All identifying information was removed from the dataset used for this study prior to data sharing. Temple University Health Sciences Institutional Review Board approval was obtained for analyses of the resulting anonymous, secondary dataset.
BMI was compared by weight perception using a two-sample t-test. Chi-square tests of independence were used to evaluate the direct effects of weight perception (dichotomous: underestimators vs. accurate reporters) on gender, grade, and readiness to participate in the four dichotomous health behavior indicators, as well as on the four indicators of confidence in health behavior participation (three levels each). Effect sizes were quantified continuously using
The mediational effects of self-efficacy (four measures of self-efficacy in health related measures) on the relationship of weight perception as a predictor of health behavior were assessed using the four steps defined by Baron and Kenny27: (1) demonstration of a relationship between weight perception and health behaviors; (2) demonstration of a relationship between weight perception and the mediational variable of interest; (3) demonstration of a relationship between the potential mediator and health behaviors; and (4) demonstration of a null, or significantly reduced, effect of weight perception on health behaviors, in
Among the 1,210 obese students, ¬18% (
The distributions of self-reported rates of readiness to participate in health behavior are presented in Table 1.
The underestimators reported higher rates of physical activity, eating healthy food, healthy weight management, and an overall healthier lifestyle than those with accurate report.
As predicted, the underestimators reported greater self-efficacy in their ability to engage in each of the four health behaviors than those with accurate self-report (see Table 2).
Consistent with the differences in their average BMI’s, more than twice the proportion of underestimators compared to accurate self-reporters expressed strong self-efficacy (“I can do it, no problem”) with healthy weight management. Self-efficacy was also found to be significantly associated with self-reported participation in each of the corresponding health behaviors. The results of this series of χ2 tests are presented in Table 3.
Logistic Regression Models
In the fourth and final step of the mediational analysis, a series of four logistic regression analyses were run to determine the influence of self-efficacy on the associations between weight perception and self-reported health behavior participation. A null effect for weight perception on physical activity and healthy diet, respectively, was found when self-efficacy was entered into these logistic regression equations, indicating that self-efficacy mediates the relationship between weight perception and these two outcomes (see Figure 1).
Figure 1. Observed Models in which Self-efficacy Significantly Mediated the Relationship Between Weight Perception and Health Behaviors.
In contrast, although a null effect was not found for weight management and overall healthy lifestyle when self-efficacy was entered into those two equations, the shift in parameter estimates of exp(β) towards unity indicates partial mediational effects27. The β odds are presented in Table 4.
Gender And Grade Level
Gender and grade level were included as covariates in the logistic regression analyses generated in step 4. Gender did not emerge an independent predictor in the final step for any of the dependent health behavior variables. Grade remained an independent predictor of only one of the four outcome measures, healthy diet, in step 4 of the comprehensive analysis. Among 11 th graders, a null effect for weight perception on healthy diet was found when self-efficacy was entered into the logistic regression model, indicating that self-efficacy mediates the relationship between weight perception and outcome. The parameter estimates of exp(β) for weight perception were 1.70 (95% CI 1.07 to 2.70) and 1.16 (95% CI 0.70 to 1.94) in models 1 and 3, respectively. In contrast, although a null effect was not found among the 7 th graders for healthy diet when self-efficacy was entered into the model, the shift in parameter estimates associated with the odds ratio towards unity (i.e., an attenuation of effect), indicates partial mediational effects27. The parameter estimates of exp(β) for weight perception were 2.24 (95% CI 1.37 to 3.66) and 1.82 (95% CI 1.07 to 3.08) in models 1 and 3, respectively.
This study intended to determine the association between weight perception and health behavior participation among obese adolescents. Two competing hypotheses were tested: (1) underestimation is associated with
Consistent with the second hypothesis, self-efficacy was shown to be positively associated with both weight underestimation (
Although self-efficacy appears to be a key mediating factor between weight underestimation and self-reported health behavior participation, other mediators not accounted for in this dataset could be involved. One speculative hypothesis comes from the finding that intrinsic motivation (e.g., fitness, recreation, etc.) predicts exercise and general quality of life, whereas extrinsic motivation (e.g., weight loss, attractiveness) is negatively associated with these outcomes17. Without an accurate recognition of their obese status, it would seem that underestimators would be more likely to participate in health behaviors for intrinsic reasons other than weight loss. On the other hand, those who recognize that they are obese may tend to be interested in health behaviors for the specific purpose of trying to lose weight, which is an extrinsic source of motivation that negatively predicts actual exercise participation.
Body dissatisfaction could be another mediating factor in the relationship between weight underestimation and health behavior participation. Presumably, those who identified themselves as overweight would be more at-risk for lower body satisfaction. Although body dissatisfaction has been shown to be associated with unhealthy weight loss efforts, such as binge eating and smoking, it predicts lower levels of healthy weight loss behaviors, such as fruit and vegetable intake and physical activity28.
Consistent with previous adolescent research6, the males who were obese were more likely to underestimate their weight status than the females. Also, the 7 th graders were more prone to underestimate their obese status than the 11 th graders, perhaps suggesting that weight self-awareness is more accurate among older than younger adolescents. Yet in our multivariate analyses, grade level appeared to be a more influential moderator of health behavior than gender (see Table 4).
The most significant limitation of this study was the lack of available psychometric data for the self-report measure. Although the PWP items that are intended to measure health behavior participation, self-efficacy, and interest have strong content validity, they are not standardized measurements of these constructs. The authors emphasize that the use of this self-report measure enabled an initial investigation into variables of interest that have potential to inform and improve future obesity research and prevention efforts.
The use of self-report in itself is another limitation of the study. Although self-report is commonly used to determine levels of health behavior participation, it is susceptible to the influence of a number of cognitive and situational factors that could reduce reliability and validity of this key variable29. In this study, there is a possibility of a spurious, bidirectional relationship with self-reported weight. That is, self-reported weight status may have been influenced by reporting biases other than actual weight perception, such as an optimistic view of reality or concern with impression management (i.e., the desire to look “good’). From the perspective of cognitive dissonance theory30, for example, participants may have strived to maintain cognitive consistency by reporting health behavior levels that were correspondent to their self-images. Also, those who underestimated their weight status may just have been less accurate overall in their self-report, which could have also led to skewed self-reported participation rates.
The use of nurse-obtained BMI was a strength of this study because adolescents, particularly those who are obese or overweight, tend to under-report their body weight, which can lead to an underestimation of the prevalence of overweight and obesity31,32. On the other hand, although BMI is considered a valid indicator of adiposity in adolescents33,34, a flaw of the measure is that it does not take into account body frame and muscularity. Therefore, the use of BMI to assess adiposity could result in the erroneous categorization of some muscular and broad-shouldered athletic students as obese.
Both the underestimators and accurate reporters had average BMI’s within the obese range. However, consistent with the underestimators’ assertion of healthier weight status, they had lower BMI’s than the accurate reporters. It makes intuitive sense that the students with lower levels of obesity who are somewhat closer to the cut-off for normal weight status would be more prone to weight underestimation those with a greater degree of obesity. It is also possible that a portion of the heightened participation in health behaviors among underestimators may have been due to the limitations of BMI as a measure of obesity in adolescents, rather than a direct perception-behavior relationship. In addition to improved assessment of health behavior participation, future study in this area will benefit from including multiple fitness indicators.
Other limitations included the correlational nature of the analyses, thereby precluding any conclusions about the causal relationships among variables. The sample was obtained from a predominantly rural school district and therefore cannot be assumed to be representative of the U.S. adolescent population. Furthermore, due to the school districts’ concerns with privacy issues, there were no racial or socioeconomic data available to further determine generalizability and the potential influence of these important variables. Interpretation of these findings is also limited to 7 th and 11 th graders and the results cannot be assumed to be representative of the entire adolescent developmental spectrum.
Despite this study’s limitations, results supported a hypothesis about the role of weight perception in a sample of obese adolescents. The associations between accurate self-perception and lower rates of health behaviors and lower self-efficacy in one’s ability to participate in these behaviors suggest a potential mechanism that may create barriers to weight loss efforts among some obese adolescents. The proposed mechanism may be similar to the stereotype threat that has been found to lead to the underperformance of minority students on academic skills tests when their minority status is made salient. Similar observations have been observed among female students who have demonstrated underperformance on math tests when their self-awareness of their gender status is raised35. The strong associations between self-efficacy and participation in healthy weight loss behaviors suggest that this factor may deserve special focus in prevention and intervention efforts. As further indication of the importance of this variable, self-efficacy was found to at least partially mediate the relationship between weight perception and the health behaviors.
We speculate that interventions that foster self-efficacy in regard to weight loss behaviors may be a relatively more effective strategy than focusing on increasing self-awareness of obesity and the associated risks. Programs that provide opportunities for supervised instruction and coaching while supporting and encouraging student participation may be particularly beneficial for adolescents who are self-aware of their obese status. Research on stereotype threat offers encouraging support for the notion that relatively straightforward steps can be taken to overcome negative self-perceptions36,37, which suggests that it would be possible to reduce the potential deleterious influence of self-perception on health behavior participation. Also, given that self-recognition of one’s obesity status does not render adolescents more likely to participate in health behaviors, interventions should seek to tap into and to further develop the intrinsic motivations, such as enjoyment, competition, and opportunity for socialization, that may be more important in facilitating participation. Nonetheless, the direct effects of both weight perception and self-efficacy should be considered when attempting to understand and predict which obese adolescents are likely to engage in health behaviors.
This study has provided preliminary insights into the relationship between obesity and weight perception among adolescents, but the implications of the findings should remain tentative at this point due to the limitations of this study design. It is uncertain whether weight perception has played a meaningful role in contributing to the current adolescent obesity epidemic. Future studies are necessary to further determine and the most effective ways for intervention efforts to address the issue of misperception.
The authors want to thank Donna Decker, Jack Dennis, Elizabeth Hanna, and Joseph Glutting for their assistance with this manuscript. This project was completed to meet the research requirement for Dr. Thunfors’ clinical psychology internship at Temple University’s College of Health Professions.