Development of a Data Mining-Based Risk Screening Model for Non-Communicable Diseases among Older Adults to Support Community Health Surveillance

Abstract

In 2025, Thailand entered an Aged Society, with adults aged 60 years and older comprising over 20% of the total population, resulting in a growing burden of non-communicable diseases (NCDs), particularly diabetes mellitus and hypertension, which have become major public health concerns, especially in community settings with limited resources and data-driven decision-support tools. This study aimed to develop a risk screening model for non-communicable diseases among older adults using data mining and machine learning techniques applied to community health data. The study analyzed health screening records from 3,419 older adults in Laplae District, Uttaradit Province, Thailand, collected between 2022 and 2025. Data were processed following the CRISP-DM framework. Random Forest and Extreme Gradient Boosting (XGBoost) models were developed and evaluated using Accuracy, Precision, Recall, F1-score, and Confusion Matrix.
The results demonstrated that XGBoost was more appropriate for public health screening purposes, achieving a Sensitivity of 25.0% for diabetes risk detection and 16.0% for hypertension, compared to 17.3% and 4.0% for Random Forest, respectively. Although Random Forest yielded higher overall accuracy (91.3% and 91.7%), XGBoost produced fewer false negatives, which is critically important in preventive health screening contexts. Key risk-associated factors included body mass index, waist circumference, blood glucose level, blood pressure, and age. The findings suggest that integrating machine learning with community-level health data can effectively support preliminary risk screening, proactive health surveillance, and local healthcare planning for older adults. The model is intended as a decision-support tool only, and is not a substitute for clinical diagnosis.