Improved Risk Modeling for Concurrent Diabetes and Hypertension: A Biresponse Nonparametric Logistic Regression Approach

Abstract

One of the key strategic priorities in Indonesia’s health development, as outlined in the Sustainable Development Goals (SDGs) agenda, is to reduce premature mortality from Non-Communicable Diseases (NCDs) by one-third. Diabetes and hypertension are two closely related NCDs that often coexist. This study aims to develop a risk model for the simultaneous incidence of diabetes and hypertension using a biresponse approach. Data were collected from 211 patients at the Internal Medicine Polyclinic of Airlangga University Hospital Surabaya. A Chi-square dependency test revealed a significant association between the incidence of diabetes and hypertension. Additionally, the relationship between each predictor variable and the observed logit of diabetes and hypertension demonstrated a non-linear pattern, suggesting that the impact of predictor variables on the risk of both diseases is not linear. A comparison of the biresponse logistic regression model with both parametric and nonparametric approaches indicated that the Biresponse Nonparametric Logistic Regression model outperformed the parametric approach in terms of performance and stability. The model’s accuracy improved significantly from 0.436 to 0.626, and the Area Under the Curve (AUC) increased from 0.62 to 0.83.