A risk model to predict atrial fibrillation in diabetes using machine learning: The ACCORD study

Area of Science:

• Cardiology
• Medical Informatics
• Diabetes Research

Background:

• Atrial fibrillation (AF) is a common complication in patients with type 2 diabetes.
• The predictive value of machine learning (ML) models for AF in this population is not well-established.
• Comparison with traditional risk prediction models like Cohorts for Heart and Aging Research in Genomic Epidemiology (CHARGE-AF) is needed.

Purpose of the Study:

• To compare the predictive performance of a machine learning (ML) random forest (RF) model against the traditional CHARGE-AF model for atrial fibrillation (AF) risk in patients with type 2 diabetes.
• To identify key predictors of AF in this cohort using ML.

Main Methods:

• Utilized data from 9,307 patients with type 2 diabetes and no prior AF from the Action to Control Cardiovascular Risk in Diabetes (ACCORD) study.
• Developed and validated a random forest (RF) classifier using clinical and metabolic variables.
• Compared RF model performance with a CHARGE-AF Cox model using five-fold cross-validated area under the receiver operating curve (AUC).

Main Results:

• Over a median follow-up of 6.26 years, 175 patients developed AF.
• The RF model achieved an AUC of 0.731, comparable to the CHARGE-AF model's AUC of 0.756 (p=0.18).
• Key predictors identified by the RF model included age, waist circumference, race, total cholesterol, and estimated glomerular filtration rate.

Conclusions:

• Machine learning (ML) models demonstrate comparable performance to traditional models in predicting AF in patients with type 2 diabetes.
• ML identified distinct predictors, suggesting potential for personalized AF risk stratification and prevention strategies.
• This study supports the integration of ML into cardiovascular risk assessment for diabetic patients.