Machine Learning Models for Predicting Type 2 Diabetes Complications in Malaysia

Area of Science:

• Medical informatics
• Public health
• Machine learning in healthcare

Background:

• Type 2 diabetes (T2D) poses a significant health burden in Malaysia.
• Predicting diabetic complications is crucial for effective patient management and intervention.
• Existing predictive models may not fully capture the complexity of T2D progression.

Purpose of the Study:

• To develop and evaluate machine learning (ML) models for predicting five key diabetic complications in Malaysian T2D patients.
• To identify the most effective ML algorithm for this predictive task.
• To support the integration of ML tools into clinical practice for diabetes care.

Main Methods:

• Utilized data from the Malaysian National Diabetes Registry and Death Register (2011-2021).
• Developed predictive models for all-cause mortality, retinopathy, nephropathy, ischemic heart disease (IHD), and cerebrovascular disease (CeVD).
• Tested seven ML algorithms, including Light Gradient Boosting Machine (LGBM), and compared their performance using ROC-AUC scores.

Main Results:

• The Light Gradient Boosting Machine (LGBM) algorithm demonstrated superior performance among the tested models.
• LGBM achieved notable ROC-AUC scores: 0.84 (all-cause mortality), 0.71 (retinopathy), 0.71 (nephropathy), 0.66 (IHD), and 0.74 (CeVD).
• The study cohort included 90,933 patients with type 2 diabetes.

Conclusions:

• Machine learning, particularly LGBM, shows strong potential for predicting diabetic complications in a real-world Malaysian T2D population.
• These predictive models can aid in targeted preventive strategies and personalized diabetes management.
• Further validation and optimization are recommended for broader clinical application and diverse populations.