To the Editor: The global obesity epidemic has posed a major threat to population health.1 Given that weight change and obesity have been linked with health outcomes, physical functioning, and quality of life,2,3 accurate population surveillance data on the trend of weight change and obesity has remained important for public health efforts to combat the obesity epidemic. In large-scale population-based studies, self-reported height and weight have been used to derive body mass index (BMI) for classifying obesity categories.2–5 Despite the cost savings and the general high correlation between self-reported and measured height and weight, important differences may have continued to exist in the estimated obesity prevalence, but the most recent study on the reporting accuracy of height, weight, and BMI measures used data from 1988 to 1994 from the National Health and Nutrition Examination Survey.6 Therefore, the concordance between self-reported versus measured height, weight, and BMI was assessed using the 2006 data from a nationally representative sample of U.S. adults.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Select a PDF to begin download
The data collected from 6,799 participants of the Health and Retirement Study (HRS), a biennial longitudinal survey of U.S. adults aged 51 and older were examined.7 Information on self-reported height and weight was collected during interviews at each wave from 1992 to 2006. In 2006, the HRS also collected physical measures of height and weight.8 Height was measured without shoes and recorded in inches to the nearest quarter inch. Weight was measured using a Healthometer 830 kiloliter scale (without shoes and with light clothing; Jarden Corporation, Rye, NY) and recorded to the nearest half pound.8 Trained interviewers took all measurements. Mean differences between the self-reported and measured height and weight were evaluated. In addition, sensitivity and specificity of obesity classification based on self-reported data were assessed.
On average, respondents (men and women) overreported their height by 1% to 2% and underreported their weight by 1% to 3%, resulting in an underestimation of BMI of approximately 4%. Mean reporting errors of height tended to increase slightly with age for men and women (P<.001), whereas underreporting of weight decreased with age for both sexes (P<.001). As a result, the underestimation of BMI increased slightly with age (Table 1).
|n||Height, cm||Weight, kg||BMI, kg/m2|
|Mean||Difference ± SE||%||Mean||Difference ± SE||%||Mean||Difference ± SE||%|
|Full sample||6,799||167.0||1.9 ± 0.1||1||81.3||−1.6 ± 0.1||−2||29.1||−1.2 ± 0.1||−4|
|50–59||673||176.0||1.6 ± 0.2||1||91.6||−1.6 ± 0.2||−2||29.6||−1.0 ± 0.1||−3|
|60–69||971||175.1||2.2 ± 0.1||1||89.4||−1.5 ± 0.2||−2||29.1||−1.2 ± 0.1||−4|
|70–79||851||173.9||2.7 ± 0.1||2||88.1||−1.3 ± 0.1||−1||29.1||−1.3 ± 0.1||−4|
|≥80||394||172.0||3.0 ± 0.3||2||79.9||−0.8 ± 0.3||−1||27.0||−1.1 ± 0.1||−4|
|50–59||924||162.0||0.8 ± 0.1||1||78.1||−2.1 ± 0.1||−3||29.8||−1.1 ± 0.1||−4|
|60–69||1,296||160.7||1.6 ± 0.1||1||77.2||−1.9 ± 0.1||−2||29.9||−1.3 ± 0.1||−4|
|70–79||1,097||159.3||2.0 ± 0.1||1||72.0||−1.5 ± 0.1||−2||28.4||−1.3 ± 0.1||−4|
|≥80||593||156.9||3.2 ± 0.2||2||65.1||−0.8 ± 0.1||−1||26.4||−1.4 ± 0.1||−5|
The prevalence of obesity was underestimated by 6% to 12% for all age and race and ethnicity groups and both sexes. The prevalence of overweight was overreported by 1% to 5% for all but women aged 80 and older. Obesity classification based on self-reported measures demonstrated high specificity (97–99%) across all age, sex, and racial and ethnic groups, but sensitivity was low and tended to decrease with age. Sensitivity decreased from 76% for men aged 50 to 59 to 46% for those aged 80 and older and from 83% to 57% in women.
As expected, older age was associated with decreased concordance between self-reported and measured height, whereas reporting accuracy of weight increased as age increased. The overreporting of height may have occurred because of the onset of diseases (such as osteoporosis) during the study period.9 In addition, the pattern of overreported height and underreported weight might indicate that social desirability affects reporting accuracy of height and weight.10 Although the average absolute reporting errors for height and weight were small, population classification based on self-reported measures consistently underestimated the prevalence of obesity. With respect to the effect of race and ethnicity on the concordance between self-reported and measured height and weight, the current analysis did not indicate a significant difference in reporting bias between racial and ethnical subgroups.
These findings suggest that self-reported height and weight measures would be better used as continuous variables than BMI categories to classify overweight or obesity status in middle-aged and older U.S. adults because of the identified patterns of reporting errors. The effects of age and sex on concordance should be taken into consideration when interpreting overweight or obesity status and related clinical outcomes. Practitioners and researchers should collect direct measures of height and weight to obtain more-accurate measures for individuals, as well as to correctly estimate obesity prevalence in clinical and public health practice. Collecting measured height and weight is especially important in longitudinal studies to correctly capture changes in these measures. Methods to adjust existing self-reported height and weight data maybe needed to correctly reflect longitudinal trend in obesity in U.S. middle-aged and older adults.
Conflict of Interest: The editor in chief has reviewed the conflict of interest checklist provided by the authors and has determined that the authors have no financial or any other kind of personal conflicts with this paper.
Dr. He was supported by a mini-grant from the College of Urban Affairs at Cleveland State University, and a 2009 to 2011 Pfizer Scholars Grant in Health Policy (0262022510HEXIA04).
Author Contributions: Hongdao Meng conceptualized the research question, performed the analysis, and assisted with letter writing. Xiaoxing He assisted with conceptualization, analysis, and data interpretation and wrote the letter. Denise Dixon assisted in data interpretation and letter writing.
Sponsor's Role: None.