Joint Modeling of Hypertension, Diabetes, and Cardiovascular Disease in Uganda: A Copula-Based Approach

Non-communicable diseases (NCDs), including hypertension, diabetes mellitus, and cardiovascular disease (CVD), are increasingly responsible for morbidity and mortality in low- and middle-income countries (LMICs). In Uganda, these conditions are becoming more prevalent due to rapid urbanization, lifestyle transitions, and demographic shifts. Despite their frequent co-occurrence, existing research typically investigates each condition in isolation, failing to capture their shared etiology and potential dependence. This study analyzes nationally representative data from the 2014 Uganda WHO STEPS survey to assess both individual-level risk factors and the joint distribution of hypertension, diabetes, and CVD. A total of 3,987 adults aged 18-69 years were included. Survey-weighted logistic regression models were fitted for each binary outcome, adjusting for age, sex, region, marital status, body mass index (BMI), education, income, physical activity, cholesterol, tobacco, and alcohol use. Missing data were addressed using multiple imputation by chained equations. Probability-scale residuals from the marginal models were used to fit both Gaussian and Clayton tri-variate copulas, enabling flexible modeling of the dependence structure. Hypertension affected 62.3% of participants, diabetes 65.8%, and CVD 9.3%. Obesity was strongly associated with hypertension (OR = 2.00; 95% CI: 1.02-3.92), while overweight and obesity were inversely associated with diabetes. Age was positively associated with both hypertension and CVD. Women had lower odds of hypertension but higher odds of CVD. Participants in the Eastern region had lower CVD risk, while those in the Northern region showed elevated hypertension risk. The Gaussian copula showed a better fit than the Clayton copula (log-likelihood: 184.04 vs. 87.33; AIC: -362.08 vs. - 172.65), indicating moderate symmetric dependence between the outcomes. The Clayton copula revealed minimal lower-tail dependence (Kendalls {tau} = 0.0365), suggesting weak clustering of low- outcome probabilities. These findings highlight the need for integrated, person-centered approaches to NCD care in Uganda. Joint modeling using copula-based frameworks enhances understanding of multi-morbidity patterns and enables better targeting of prevention and intervention strategies. The Gaussian copulas superior performance underscores the potential of symmetric dependence models in capturing the clustering of NCDs, particularly the strong correlation observed between hypertension and CVD. These insights are essential for informing health policy and optimizing resource allocation in settings undergoing epidemiological transition.