Self-reported Versus Measured BMI And Mortality Among Older Americans: The Health And Retirement Study

Large, nationally representative studies often rely on self-reported BMI (rBMI) so it is important to examine whether rBMI differs from measured BMI (mBMI) when examining its relationship with mortality. PURPOSE: To examine the differences between rBMI and mBMI and the impact on associations with 10-year mortality among middle-age and older adults. METHODS: The sample (n = 13,042) was from the 8th and 9th waves recruited for the Health and Retirement Study, a nationally representative longitudinal study, where both rBMI and mBMI were reported. Those aged 51 years or over (68.7 ± 9.8 years) with both rBMI and mBMI were included. Ten-year mortality was determined using data from waves 13 and 14. Paired t-test, regression, and Bland Altman plots were used to assess agreement between rBMI and mBMI. Multiple regression was performed to assess factors (age, gender, education, race, marriage status, health condition, and smoking) that were associated with the difference between rBMI and mBMI. Chi-square and ROC analyses were used to assess whether rBMI was accurate in classifying obesity. Alpha was set at p < 0.05. RESULTS: The mean rBMI (M = 27.8 kg/m2, SD = 5.37) and mBMI (M = 29.1 kg/m2, SD =5.86, p < 0.05) were significantly different, but were also significantly related (R2 = 85.1%, SEE = 2.08). Bland-Altman analysis showed a significant relationship between mBMI and the differences between these two methods (R2 = 16.1%, p < 0.05). Older adults and those with higher mBMIs had significantly larger differences between rBMI and mBMI (R2 = 17.1%, p < 0.01). Obesity classification (BMI > 30 kg/m2) based on rBMI showed high specificity (97.7-98.7%) among all age and gender groups, but low sensitivity which decreased with age (50-59 years old: 81.1%; 60-69 years old: 78.8%; 70-79 years old: 73.1%; over 80 years old: 57.7%). For those over 80 years, sensitivity was 53.3% in males, and 60.8% in females. Overall, rBMI misclassified 24%, 8%, 23% and 25% of people in the Underweight, Normal, Overweight and Obese categories, respectively. Mortality increased from 28.39% to 33.61% (p < 0.05) as the difference between methods became greater. CONCLUSIONS: The accuracy of rBMI becomes lower with age and higher BMI which results in misclassification of obesity status. rBMI should be used with caution when determining factors associated with mortality.