Reliability and Validity of Birth Certificate Prepregnancy Weight and Height Among Women Enrolled in Prenatal WIC Program: Florida, 2005

To investigate the reliability and validity of weight, height, and body mass index (BMI) from birth certificates with directly measured values from the Women, Infants, and Children (WIC) Program. Florida birth certificate data were linked and compared with first trimester WIC data for women with a live birth during the last quarter of calendar year 2005 (n = 23,314 women). Mean differences for weight, height, and BMI were calculated by subtracting birth certificate values from WIC values. Reliability was estimated by Pearson's correlation. Validity was measured by sensitivity and specificity using WIC data as the reference. Overall mean differences plus or minus standard error (SE) were 1.93 +/- A 0.04 kg for weight, -1.03 +/- A 0.03 cm for height, and 1.07 +/- A 0.02 kg/m(2) for BMI. Pearson's correlation ranged from 0.83 to 0.95, which indicates a strong positive association. Compared with other categories, women in the second weight group (56.7-65.8 kg), the highest height group (a parts per thousand yen167.6 cm), or BMI < 18.5 had the greatest mean differences for weight (2.2 +/- A 0.08 kg), height (-2.4 +/- A 0.05 cm), and BMI (1.5 +/- A 0.06), respectively. Mean differences by maternal characteristics were similar, but statistically significant, likely in part from the large sample size. The sensitivity for birth certificate data was 77.3% (+/- 1.42) for underweight (BMI < 18.5) and 76.4% (+/- 0.51) for obesity (BMI a parts per thousand yen 30). Specificity was 96.8% (+/- 0.12) for underweight and 97.5% (+/- 0.12) for obesity. Birth certificate data had higher underweight prevalence (6 vs. 4%) and lower obesity prevalence (24 vs. 29%), compared with WIC data. Although birth certificate data overestimated underweight and underestimated obesity prevalence, the difference was minimal and has limited impact on the reliability and validity for population-based surveillance and research purposes related to recall or reporting bias.