Predictive Modeling of Diabetes Classification using Binomial Logistic Regression on Biomedical Indicators

Diabetes mellitus remains a significant health and economic burden globally, with early detection frequently delayed in resource-limited settings such as the Philippines. Addressing this gap, the present study aimed to develop a predictive model for classifying individuals as diabetic or non-diabetic using biomedical indicators, including Body Mass Index (BMI), Low-Density Lipoprotein (LDL), Glycated Hemoglobin (HbA1c), and Triglycerides. Guided by Roy Baumeister’s Self-Regulation Theory of Illness Behavior, the study employed a multi-method classification approach involving binomial logistic regression, K-means clustering, and decision tree analysis. A total of 947 participants aged 24 to 79 years were included. K-means clustering categorized participants into two distinct groups based on biomarker profiles, differentiating those at higher and lower risk of diabetes. Logistic regression identified BMI as the most significant predictor (χ²(1) = 104.44, p < .001), followed by HbA1c (χ²(1) = 51.80, p < .001), Triglycerides (χ²(1) = 12.44, p < .001), and LDL (χ²(1) = 9.15, p = .002). The model demonstrated excellent predictive performance, with an McFadden R² of 0.80 and a Nagelkerke R² of 0.85. Decision tree analysis confirmed BMI as the primary classifier, with HbA1c enhancing classification accuracy, thereby highlighting the combined diagnostic utility of both. These findings suggest that incorporating BMI and HbA1c thresholds as accessible, cost-effective screening tools within barangay health systems could improve early identification of individuals at risk for diabetes. Integrating predictive analytics with behavior modification programs based on self-regulation theory may empower communities to adopt preventive health measures. The study recommends prioritizing risk-based screening protocols, subsidizing access to essential biomarker testing, and integrating predictive modeling frameworks into primary healthcare. This multi-method model presents a robust, scalable tool to enhance diabetes risk prediction and support targeted health interventions in underserved Philippine communities.