Assessment and Prediction of Body Mass Index (BMI) Distributions among Adult Populations in Mexico, Colombia, and Peru, 1988-2014
MetadataShow full item record
Unequal distribution of adult obesity across subpopulations in low- and middle-income countries has been reported, but findings have been mainly from data of women of reproductive age. While mean body mass index (BMI), overweight and obesity prevalence are commonly used obesity indicators, incorporation of ever-changing skewed BMI distributions has been a challenge. In this context, our study aimed to assess differences in magnitude and rates of change in BMI distributions by sex, age, geographic and socioeconomic factors in Mexico, Colombia, and Peru by modeling entire BMI distributions. Furthermore, this modeling technique was applied for the prediction of future obesity indicators. Data from nationally representative health surveys conducted between 1988 and 2014 in these 3 countries were used. The analyses were conducted using the generalized additive model for location, scale, and shape (GAMLSS) in order to model BMI distributions. BMI was assumed to follow a Box-Cox Power Exponential (BCPE) distribution, and each of its 4 parameters was modeled as a function of demographic, geographic, and socioeconomic factors. Prediction models were evaluated using data before the last survey, with the predicted values compared to actual values at the time of the last survey. Whereas women had more right-shifted and wider BMI distributions than men across the countries in 2010, men generally experienced more rapid increases in BMI between 2005 and 2010. More education was negatively associated with BMI in women after covariate adjustment whereas it was somewhat positively associated in men. Higher household wealth was positively associated with BMI in men. Lower household wealth was associated with higher rates of change in BMI distributions in women. The BCPE-GAMLSS model yielded the best prediction performance among the assessed models in predicting obesity prevalence. Observed differences in BMI distributions across subpopulations suggest the necessity of tailoring relevant policies and programs to reach target populations. Increases in BMI imply increases in obesity-associated diseases, such as cardiovascular diseases and diabetes, for which preventive and preparative actions would be urgent. The BCPE-GAMLSS method worked well for estimation and prediction of BMI by modeling its distributions precisely.