Using Artificial Intelligence to Develop a Multivariate Model with a Machine Learning Model to Predict Complications...

Area of Science:

• Diabetes Mellitus Research
• Machine Learning in Healthcare
• Chronic Disease Prediction

Background:

• Diabetes mellitus is a chronic disease with severe complications, necessitating predictive models for risk identification.
• Limited data exists on chronic complication risk factors for diabetic patients.
• This study addresses the need for better prediction of diabetes-related chronic complications.

Purpose of the Study:

• To develop a machine learning model for identifying risk factors of chronic complications in diabetes mellitus patients.
• To predict complications such as amputations, myocardial infarction, stroke, nephropathy, and retinopathy.
• To analyze key predictors using SHAP values for enhanced understanding.

Main Methods:

• A national nested case-control study involving 63,776 patients and 215 predictors over four years.
• Utilized an XGBoost machine learning model for predicting chronic complications.
• Employed SHAP values to determine the significance of various risk factors.

Main Results:

• The XGBoost model achieved an AUC of 84% in predicting chronic complications.
• Crucial risk factors identified include continued management, metformin use, advanced age (68-104 years), nutrition consultation, and treatment adherence.
• Elevated diastolic (>70 mmHg) and systolic (>120 mmHg) blood pressure significantly increased risk, while a BMI > 32 showed a protective effect.

Conclusions:

• Artificial intelligence, specifically machine learning, is a powerful tool for analyzing diabetes complications.
• The study identified key modifiable and non-modifiable risk factors for chronic complications in diabetes.
• Further research is recommended to validate and expand upon these findings.