A Comparative Study of Univariate and Bivariate Logistic Regression on Hypertension and Obesity

Hypertension and obesity frequently co-occur and share common pathophysiological mechanisms. However, most existing studies model these conditions independently using separate logistic regression models, which ignore their dependency structure and may yield inefficient estimates. This study compares univariate and bivariate binary logistic regression approaches in modelling hypertension and obesity simultaneously, using individual-level data from the Indonesian Family Life Survey wave 5 (IFLS5), comprising 8,100 respondents. Age, sex, waist circumference, handgrip strength, lung capacity, and pulse rate were included as predictors. The bivariate model estimated a significant dependence parameter of θ = 1.216 (p = 0.010), confirming positive co-occurrence between the two conditions. Waist circumference emerged as the dominant predictor for both outcomes. Model comparison based on AIC demonstrated that the bivariate approach outperformed the combined univariate models, indicating superior fit when accounting for the dependence structure.