Misclassification of Body Mass Index (BMI) using self-reported weight and height in an urban Mexican-American population

Purpose To assess the impact of differential misclassification of BMI category on resultant data analysis and interpretation when using self-reported weight and height versus measured weight and height. Methods In an ongoing cohort study of Mexicans and Mexican-Americans in Houston, Texas, we evaluated the reliability of Body Mass Index (BMI) calculated from self-reported to measured weight and height among a sample of 955 adults, aged 20–93. Spearman correlations were calculated for height, weight and BMI. We examined the mean differences between measured and self-reported values using Studentʹs t-tests. Results The mean difference, calculated as BMI from measured weight and height − BMI from self-reported weight and height, was 0.79 units for women (p < 0.0001) and 0.42 units for men (p = 0.0141). Spearman correlations were calculated for height (0.81 in men and 0.70 in women), weight (0.92 in men and 0.95 in women), and BMI (0.85 in men and 0.88 in women). We evaluated the potential for BMI misclassification based on the 1998 NHLBI categories (Underweight <18.5, Normal 18.5 – 24.9, Overweight 25.0 – 29.9, Obesity I 30.0 – 34.9, Obesity II 35.0 – 39.9, Extreme Obesity ≥ 40). We found that even small inaccuracies in self-reported weight (1–2 lbs.), and height (<1 inch), lead to 33% of participants being misclassified. Using measured values as gold standard, 11% were misclassified to a higher BMI category and 22% to a lower category. The BMI misclassification was differential by gender. BMI from self-reported data was overestimated by one or more categories by 16% of men and 12% of women and underestimated by 14% of men and 21% of women. This differential misclassification seems to be primarily due to both men and women overestimating their height and women underestimating their weight. Conclusion These findings suggest the need for caution when using self-reported data to calculate BMI among adults, and underscore the need for similar studies in other populations.