Global variation in cardiometabolic risk structures: A 48-country comparative Bayesian network analysis in 146,000 participants using WHO STEPS data

Background Cardiometabolic-based chronic disease (CMBCD) at an individual level results from complex interactions among a multi-tiered network of sociodemographic, behavioral, and metabolic factors. Though a consensus set of risk factors drives CMBCD, population context influences risk factor effects and interactions. To better understand this phenomenon, we investigated the multi-tiered networking of cardiometabolic variables across diverse populations using a comparative modelling approach. Methods and Findings Utilizing nationally representative cross-sectional data from 48 countries participating in the World Health Organization "STEPwise approach to noncommunicable disease risk factor surveillance" survey, we learned country-specific Bayesian networks including sociodemographic, behavioral, and cardiometabolic variables (adiposity, diabetes, hypertension, hyperlipidemia, and cardiovascular disease). By computing the structural Hamming distance between pairs of networks, we compared differences in network structures across regions and country income levels. We then used the learned networks to assess individual risk factor influences and interactions on cardiometabolic outcomes. Country-specific Bayesian networks varied in terms of the risk factors directly and indirectly associated with the cardiometabolic outcomes. Network structures differed significantly across regions (p = 0.023) but not across income levels (p = 0.91). These results were robust to an alternative learning algorithm, network comparison metric, and data imputation approach. Older age (60+ vs. 30-44 years old) was associated with a greater increase in probability of obesity in Europe and Central Asia (+80%) compared to other regions. Higher education was associated with increased probability of obesity (+53%), diabetes (+18%), and hypertension (+2%) in South Asia but decreased probability of obesity (-10%), diabetes (-32%), hypertension (-16%), and hyperlipidemia (-25%) in Middle East and North Africa. The interaction between age and sex in predicting obesity was significant in the highest proportion of countries in Europe and Central Asia compared to other regions. While this dataset provided standardized data across multiple countries to define cardiometabolic risk factors and drivers, there was limited data on certain health outcomes and uneven availability of data across regions. Conclusions These results revealed specific regional patterns of multi-tiered cardiometabolic risk structures, emphasizing the need for regionally tailored public health strategies rather than applying generalized consensus evidence-based models. Future research should explore the structural drivers of regional differences in inter-relationships of cardiometabolic risk factors, drivers, and disease.