Predicting Diabetes Mellitus Using Logistic Regression On Clinical And Demographic Data

Abstract

Diabetes mellitus is a long term metabolic condition characterized by elevated blood glucose levels due to impaired insulin production, insulin action or both. The global rise in diabetes prevalence presents a major public health concern. This study utilized a dataset of 768 Diabetes Cases (No Diabetes (Type 1) = 500 Cases while Yes (Type 2) Diabetes Cases = 268 Cases) Obtained from Kaggle.com to explore the clinical and demographic predictors of diabetes mellitus using logistic regression analysis. Results revealed that glucose concentration, body mass index (BMI), diabetes pedigree function and number of pregnancies were the most significant predictors of diabetes. Elevated glucose emerged as he strongest predictor while obesity and hereditary risk substantially increased the likelihood of diabetes. The model demonstrated a good fit and moderate explanatory power, correctly classifying 78.3% of cases, though it performed better at identifying non-diabetic than diabetic individuals. Receiver Operating Characteristic (ROC) analysis confirmed glucose as the most discriminative variable followed by BMI and age whereas insulin, skin thickness and blood pressure contributed minimally. These findings reinforce the multifactorial etiology of diabetes emphasizing the combined influence of clinical, genetic and demographic factors in disease prediction. Clinically, the results suggest that regular monitoring of glucose levels, BMI and family history could enhance early detection and preventive management of diabetes in at risk populations.