- Research article
- Open Access
Self-estimated BMI, but not self-perceived body size, accurately identifies unhealthy weight in US adults
BMC Public Health volume 21, Article number: 253 (2021)
Self-perceptions of health and disease can be a major driver of health behaviors. Improving accuracy of self-ascertainment of obesity may prompt uptake of weight-control behaviors in those with obesity.
We assess performance of self-perceived body size (‘too small’, ‘about right’ or ‘too large’), self-estimated BMI in kg/m2, and sociodemographics in detecting measured BMI category (under-, normal-, overweight and obese; BMI cutpoints 18.5, 25 and 30) in first bivariate and then multivariable models.
Of 37,281 adults in the US from NHANES, 2, 34, 33 and 32% were under-, normal-, overweight and obese. Respectively 56, 73, 60 and 91% self-perceived as ‘too small’, ‘about right’, ‘too large’ and ‘too large.’ Of those who self-perceived as ‘too small’, 22% were underweight and 10% were overweight or obese. 99.7% of obese participants self-estimated a BMI in the overweight/obese range, including many who did not self-perceive as ‘too large’.
Among obese participants, self-perception as either ‘about right’ or ‘too small’ was more likely for those who were younger (OR for perception as ‘too large’ 1.01 per year, 95% confidence interval 1.00–1.01) male (OR 0.33, (0.28–0.39)) nonwhite (ORs 0.36–0.79 for different ethnicities), low-income (ORs 0.61 and 1.8 for the lowest and highest of six categories, vs. the third) or measured recently (OR 0.98 (0.96–1.0) per year since 1999). Misperception was less common, but still existed, for participants with moderate or severe obesity (ORs 2.9 (2.3–3.5) and 7.9 (5.4–12), vs. ‘mild.’) (all p < 0.01.)
A tenth of adults in the US with obesity, especially those from overweight peer groups, self-perceive as normal or underweight and thus may not be motivated to control their weight. However, virtually all self-estimate an overweight or obese BMI. If measured BMI is not available, self-estimates are sufficiently accurate that interventions may rely on it to identify obesity.
Caloric imbalance leading to excess body weight is one of the largest causes of death in the US and a major contributor to six of the top ten causes of death worldwide:  it is also perhaps the only behavioral risk factor which can affect individuals without their knowledge. Mortality and morbidity caused by obesity-related conditions are as severe as those caused by cigarette smoking : yet unlike smokers, people with obesity may be unaware that their condition is a health risk and thus may not consider stopping the behaviors that cause it. Identifying and correcting weight misperception may increase the uptake of weight-control behaviors [3, 4] and help to mitigate the obesity epidemic.
Previous estimates of weight misperception suggest that up to half of overweight people self-perceive as healthy-weight: however, these studies tend to be based on small and/or strongly selected samples (e.g. university students,  physicians ) and thus their estimates are difficult to generalize outside these groups. However, it is known that self-perceived body size is largely based on comparison to peers [3, 7,8,9] and that visual adaptation to extreme body types can lead to body-size and -shape misperception even in controlled laboratory settings . Thus it is not surprising that overweight individuals are less likely to self-perceive as such if they are comparable to those around them [7, 8, 11] and that in majority-overweight communities, self-perceived body size often fails to identify overweight. The same is true in the same community over time: as prevalence of overweight increases, detection of it declines . However, corrected models are rare and thus it is not clear whether these effects are independent of each other: for example, the ethnic groups that are at elevated risk for misperception are also at elevated risk for overweight [8, 13]. Thus neither the prevalence nor the correlates of body-size misperception have been well established.
Furthermore, most research into size misperception does not distinguish overweight from obesity. While nonzero, the health risks of non-obese overweight are comparatively mild compared to those of obesity,  and these risks increase further with degree of obesity [15, 16]. Failure to ascertain obesity, especially if moderate or severe, is therefore a greater risk to health than is failure to ascertain non-obese overweight. However, neither prevalence nor correlates of misperception are well known in individuals with obesity.
Although body-size misperception is common, few interventions specifically target it; and those that do often focus on failure to ascertain underweight [17,18,19] rather than obesity. The assumption may be that all obese patients know their weight is outside the healthy range, or the topic may simply be an uncomfortable one: either way, many obese Americans do not know it and thus are unlikely to benefit from available tools for weight control. Physician counseling is inadequate to correct this misperception: not only are less than a quarter of Americans regularly examined by a physician  but physicians are often reluctant to bring a patient’s unhealthy weight to his attention . Since uptake of weight-control interventions depends on initial self-perception as overweight, there is a need for an unbiased indicator of body size which can be applied by individuals to themselves.
One potential such indicator is self-estimated body mass index (BMI) which can then be categorized objectively as ‘underweight’, ‘normal weight’, or ‘overweight / obese.’ Research suggests that while self-estimation is somewhat imprecise, it is relatively accurate: one small study found an average error of 0.8 kg/m2  while a larger one found even less . Thus self-estimated BMI may be adequate to inform self-classification of weight status for most individuals. Interventions to correct misperception thus could rely on this scalable indicator, without the need for either an objective measurement or a formal physician diagnosis.
In the current study we establish the sensitivity and specificity of self-perceived body size, and self-estimated BMI, for identifying weight category and/or unhealthy weight (overweight/obesity) in a large and representative sample of adults in the US. We then establish the sociodemographic correlates of accurate self-perception as obese. In doing so we show which groups are at greatest risk for body-size misperception, in both crude and corrected models; and then evaluate whether self-estimated BMI is sufficiently accurate to serve as an alternative metric.
Data collection and handling
Data were collected from the National Health and Nutrition Examination (NHANES) 1999–2013, which are publicly available cross-sectional samples of adults in the US. For details on data collection and handling see the website (23)and the documentation for the relevant dataset(s).
We compared weight perception and weight estimation to objectively measured body mass index (BMI, indicated as kg/m2). These were quantified as follows:
Self-estimated BMI: based on self-estimated height and weight;
Self-perceived body size: reported as underweight, about right, or overweight .
Objective and estimated BMI were categorized as underweight, normal weight, overweight or obese. In adults (age 20 and up) we used the BMI cutpoints published for adults by the US CDC: 18.5, 25, and 30 kg/m2 . In adolescents (age 16–19) we used age- and sex-specific 5th, 85th, and 95th BMI percentiles of a reference population of Americans  as recommended by the CDC .
For maximum comparability between self-perception and self-estimation, some analyses combined self-estimated overweight and obesity into a single category corresponding with self-perceived ‘overweight.’
In some analyses, obesity was subset into class 1, class 2, or class 3 based on BMI cutpoints of 35 and 40 as recommended by the CDC .
As variables associated with obesity and/or weight misperception we considered age;
sex; year of followup; ethnicity; objectively measured BMI as percent of self-estimated; and household income.
Statistical methods and modeling
All analyses were performed using SAS University Edition. Statistical significance was set at p = 0.05 unless otherwise stated. Analyses were limited to participants with complete data on measured BMI, on estimated BMI, and on perceived body size. Because perceived body size was collected only for participants over 15 years of age, all analyses were limited to this age group.
All group comparisons were done using survey-weighted logistic regression. Sample weights, including stratification and clustering, for descriptive and analytic models were taken from NHANES. Sample weights were those for participants who underwent a physical exam at the Mobile Examination Center (WTMEC2YR.)
We first established prevalence and correlates of the four weight categories (underweight, normal weight, overweight, obese) in our population and established sensitivity and specificity of self-perception and self-estimation for ascertaining objectively-measured BMI category (the gold standard). Because the provided measure of self-perceived body size did not distinguish overweight from obesity, we repeated the previous analyses treating self-estimated overweight or obesity (equivalent to adult BMI over 25) as a single category. This maximizes comparability of self-perception with self-estimation. Using logistic regression, we then modeled associations of each risk factor with accuracy of weight perception in obese participants. In these models, the outcome was self-perception as ‘too large.’ Predictors that were bivariate statistically significant then entered mutually corrected models and were removed stepwise in order of significance.
Baseline prevalences and correlates
Prevalences of underweight, normal weight, overweight and obesity were 1.92, 33.5%, 32.5 and 32.1% respectively. (Table 1.) Of obese participants, 59% had class 1 obesity; 25% had class 2; and 17% had class 3. Most group comparisons were statistically significant at p < 0.01, so we state only p-values that were less extreme than this.
Obesity was more common in more recent years (mean years since 1999 were 8.00 for obese participants, vs. 7.63 and 6.99 for overweight and normal weight). Overweight participants were older, and more likely to be male, than either obese or normal-weight participants (46.7 years, vs. 46.0 and 39.9; 58% vs. 44 and 47%.)
Self-perceived body size matched objectively measured BMI category for 56% of participants with underweight, 73% of those with normal weight, 60% of those with overweight, and 91% of those with obesity. Objectively-measured BMI averaged 97, 100, 102 and 104% of self-estimated for each group, and 90% of each group estimated their BMI within about 5% of the group mean on either side. 86% of participants with overweight estimated a BMI within the overweight/obese range, as did 99.7% of those with obesity.
Errors in self-perception were usually, but not always, between adjacent categories. One percent of participants with measured overweight, and 0.54% of those with measured obesity, self-perceived as ‘underweight.’ Of those with measured underweight, 1.3% self-perceived as “too large”; this corresponds to about 24 per 100,000 population.
Sensitivity and specificity of perceived body size
Self-perception as ‘too large’ had a sensitivity of 60% in detecting measured overweight and 91% in detecting measured obesity. (Table 2.) Its specificity was 90% in detecting combined overweight and obesity.
Self-perception as ‘underweight’ had a sensitivity of 56% in detecting measured underweight, and a specificity of 22%.
Sensitivity and specificity of estimated BMI
Estimated BMI had a sensitivity of 81% for detecting overweight, and 84% for detecting obesity. Its specificity was 78 and 95%, respectively. When overweight and obesity were combined in a single group, that group was detected by estimated BMI with a sensitivity and specificity of 93 and 96%.
Estimated BMI had a sensitivity of 67% in detecting measured underweight, and a specificity of 76%.
Detection of obesity
In corrected models predicting accurate self-perception of obesity (Table 3) most sociodemographic predictors remained statistically significant after correction for each other. Female gender was a significant predictor of accurate perception, as was high income and European-American ethnicity. All p-values were < 0.001 unless stated otherwise.
Obese participants were less likely to self-perceive as ‘too large’ if they were followed up more recently (OR for accurate perception 0.979 per year) or if they were older (OR 1.01 per year) or male (OR 0.326.) There was also a clear trend of increasing accuracy in weight perception across income categories: obese people earning below the poverty line had an OR of 0.608 for detection, while the next categories had ORs of 0.734, 1 (reference), 1.31, 2.21, and 1.83 (all pairwise p-values ≤0.05.)
All ethnic groups were at higher risk of misperception than was the European-American reference group: odds ratios were 0.356, 0.730, and 0.791 (pairwise p > 0.10) for African-American, Hispanic, and ‘other race’ respectively.
Lastly, obesity class had an effect: 83, 14, and 3.5% of obese participants who self-perceived as ‘about the right size’ were class 1, 2 and 3 respectively, compared with 56, 26 and 18% of those who self-perceived as ‘too large’ Corrected odds ratios for accurate perception were 2.86 for class 2 and 7.92 for class 3, compared to class 1 (reference.)
We find that even in the presence of body-size misperception, self-estimated BMI can accurately identify obesity. 9% of participants with obesity, including some with moderate or severe obesity, self-perceived themselves as being either the right weight or underweight. However, over 99% estimated a BMI within the overweight/obese range, which suggests that self-estimated BMI can adequately identify obesity.
Our study confirms some previously-found correlates of weight misperception among overweight/obese individuals,  such as Black and Hispanic ethnicity; and identifies some new ones, such as male gender. The increased risk of misperception among those followed up more recently supports previous findings of a generational effect in overweight perception : we suggest this may be due to the tendency of individuals to base their eating habits  and perceptions of body size [8, 27] on the people around them rather than any objective standard. This is further supported by our observations, and those of others,  that obesity misperception was often more pronounced in groups where obesity was more prevalent. To this observation we add that the effects of group membership were almost independent: the elevated risk of misperception among ethnic minorities remained even after correction for income.
Failure to ascertain obesity was more common in males, those with low income, and those from ethnic minority groups. These groups are at relatively low risk for disorders associated with the failure to ascertain underweight, such as anorexia nervosa  but at high risk for low health literacy [28, 29] and its consequences. This suggests that accuracy of ascertainment of obesity may be an aspect of health literacy rather than a symptom of disordered eating per se. This supports the findings of Bullivant et al., who showed that although obesity has similar correlates and symptoms to eating disorders (EDs) it is considered to be a separate condition by both health practitioners  and members of the public . Given the association between low health literacy and obesity-related conditions such as heart failure,  it is plausible that interventions shown to improve other aspects of health literacy could be adapted to encourage accurate self-perception of body size. Such interventions may include the use of self-estimated BMI over 25 as a scalable, sensitive and specific size indicator.
Inaccuracies in estimation of BMI contributed to inaccurate classification mostly in those participants whose BMI was relatively close to the cutoff. These included many participants with underweight, normal weight and non-obese overweight, but almost none with obesity: almost all participants with obesity self-estimated their BMI as being over 25. Thus while there are systematic errors in self-estimated BMI, these errors have only minimal effects on BMI-based ascertainment of obesity.
Lastly, while this paper focused on self-perception in individuals with obesity it also informs practice pertaining to underweight. Over three-fourths of participants who perceived themselves as underweight were incorrect, and about 10% of them were in fact overweight or obese. In the absence of objective measures, almost all perceived underweight is incorrect: treatment should not begin until the perception is confirmed by objective measures [17, 19].
This study is limited, first of all, by its cross-sectional observational nature. While research does support self-perception of body size as a major driver of the decision to lose weight, it is not the only driver.
We are also limited by the lack of data on Asian origin for most participants . The BMI cutpoints we used to ascertain weight categories were chosen for European-origin populations in spite of the known between Europeans and different populations of Asians ; thus we likely misclassified some individuals of Asian origin whose BMI was near the cutoff. Consequently, our findings for individuals of ‘other ethnicity’ (which includes all Asians) may be somewhat less reliable than those for the other ethnicities [32, 33].
These findings suggest that screening and treatment decisions for abnormal body weight (both over- and under-) should rely on BMI rather than subjective perception. Objective measures are best, but self-estimated BMI over 25 is also an adequate screening tool for obesity. Clinicians should consciously inform their subjective perception of body size to match the BMI cutoffs. Ethnic minorities, low-income people, individuals from majority-overweight communities, and men are least likely to be aware of their own obesity and thus, if properly counseled, may be more willing to lose weight.
Availability of data and materials
Publicly available at https://wwwn.cdc.gov/nchs/nhanes/search/default.aspx
Body mass index, kg/m2
US Centers for Disease Control
Ratio of household income to poverty limit
WHO. World Health Organization Fact Sheet 310: The top 10 causes of death. 2018.
Prospective Studies Collaboration. Body-mass index and cause-specific mortality in 900,000 adults: collaborative analyses of 57 prospective studies. Lancet. 2009;373:1083–96.
Duncan DT, Wolin KY, Scharoun-Lee M, Ding EL, Warner ET, Bennett GG. Does perception equal reality? Weight misperception in relation to weight-related attitudes and behaviors among overweight and obese US adults. Int J Behav Nutr Phys Act. 2011;8.
Singh S, Somers VK, Clark MM, Vickers K, Hensrud DD, Korenfeld Y, Lopez-Jimenez M. Physician diagnosis of overweight status predicts attempted and successful weight loss in patients with cardiovascular disease and central obesity. Am Heart J. 2010;160(5):934–42.
Peltzer PK, Pengpid S. Underestimation of weight and its associated factors in overweight and obese university students from 21 low, middle and emerging economy countries. Obes Res Clin Pract. 2015;9:234–42.
Perrin EM, Flower KB, Ammerman AS. Pediatricians’ own weight: self-perception, misclassification, and ease of counseling. Obesity. 2012;13:326–32.
Ali MM, Amialchuk A, Renna F. Social network and weight misperception among adolescents. South Econ J. 2011;77:827–42.
Kernper KA, Sargent RG, Drane JW, Valois RE, Hussey JR. Black and white females' perceptions of ideal body size and social norms. Obesity. 1994;2:117–26.
Lim H, Wang Y. Body weight misperception patterns and their association with health-related factors among adolescents in South Korea. Obesity (Silver Spring). 2013;21.
Brooks KR, Mond J, Mitchison D, Stevenson RJ, Challinor KL, Stephen ID. Looking at the figures: visual adaptation as a mechanism for body-size and -shape misperception. Perspect Psychol Sci. 2020;15:133–49.
Salvy SJ, Rocco DJ, Numrah P, Pliner P. Effects of social influence on eating in couples, friends and strangers. Appetite. 2007;49:92–9.
Burke MA, Heiland FW, Nadler CM. From “overweight” to “about right”: evidence of a generational shift in body weight norms. Obesity. 2010;18:1226–34.
Powell TM, de Lemos JA, Banks K, Ayers CR, Rohatgi A, Khera A, McGuire DK, Berry JD, Albert MA, Vega GL, Grundy SM, Das SR. Body size misperception: a novel determinant in the obesity epidemic. Arch Intern Med. 2010;170:1695–7.
WHO. Overweight and obesity. Global Health Observatory (GHO) data. 2016.
CDC. Adult Obesity Causes & Consequences. In: Division of Nutrition PA, and Obesity, editor. Atlanta, GA: National Center for Chronic Disease Prevention and Health Promotion; 2018.
Mond JM, Baune BT. Overweight, medical comorbidity and health-related quality of life in a community sample of women and men. Obesity. 2009;17:1627–34.
National Eating Disorders Association . Eating disorders screening tool. 2017.
Insel T. Post by former NIMH director Thomas Insel: spotlight on eating disorders. National Institutes of Mental Health; 2012.
Morgan JF, Reid F, Lacey HJ. The SCOFF questionnaire a new screening tool for eating disorders. West J Med. 2000;172:164–5.
Mehrotra A, Zaslavsky AM, Ayanian JZ. Preventive health examinations and preventive gynecological examinations in the United States. Arch Intern Med. 2007;167:1876–83.
Huber LRB. Validity of self-reported height and weight in women of reproductive age. Matern Child Health J. 2007;11:137–44.
Luo J, Thomson CA, Hendryx M, Tinker LF, Manson JE, Li Y, Nelson DA, Vitolins MA, Seguin R, Eaton CB, Wactawski-Wende J, Margolis KL. Accuracy of self-reported weight in the Women’s health initiative. Public Health Nutr. 2019;22:1019–28.
National Health and Nutrition Examination Survey 2005–2006 Questionnaires, Datasets, and Related Documentation [database on the Internet]. National Center for Health Statistics. 2008. Available from: https://wwwn.cdc.gov/nchs/nhanes/continuousnhanes/default.aspx.
US CDC. Defining Adult Overweight and Obesity; 2017.
Data Table of BMI-for-Age Charts. US Centers for Disease Control and Prevention, National Center for Health Statistics. 2001.
CDC. Recommended BMI-for-age Cutoffs.
Brooks KR, Mond JM, Stevenson RJ, Stephen ID. Body image distortion and exposure to extreme body types: contingent adaptation and cross adaptation for self and other. Front Neurosci. 2016;10:334.
Cajita MI, Cajita TR, Han HR. Health literacy and heart failure: a systematic review. J Cardiovasc Nurs. 2016;31:121–30.
Lanpher MG, Askew S, Bennett GG. Health literacy and weight change in a digital health intervention for women: a randomized controlled trial in primary care practice. J Health Commun. 2016;21:34–42.
Bullivant B, Denham AR, Stephens C, Olson RE, Mitchison D, Gill T, Maguire S, Latner JD, Hay P, Rodgers B, Stevenson RJ, Touyz S, Mond JM. Elucidating knowledge and beliefs about obesity and eating disorders among key stakeholders: paving the way for an integrated approach to health promotion. BMC Public Health. 2019;19.
Bullivant B, Mitchison D, Skromanis S, Mond J. Elucidating individuals' beliefs about the severity of eating disorders and obesity: implications for public health programs. Eat Weight Disord. 2020;25:929–38.
CDC. National Health and Nutrition Examination Survey: 2005–2006 Data Documentation, Codebook, and Frequencies.
WHO Expert Consultation. Appropriate body-mass index for Asian populations and its implications for policy and intervention strategies. Lancet. 2004;363:157–63.
Ethics approval and consent to participate
Consent for publication
All authors consent to publication.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
About this article
Cite this article
Smith, M.P. Self-estimated BMI, but not self-perceived body size, accurately identifies unhealthy weight in US adults. BMC Public Health 21, 253 (2021). https://doi.org/10.1186/s12889-021-10316-8
- Behavior change