Machine learning-based return-to-work assessment system for acute myocardial infarction patients within 12 months

Area of Science:

• Cardiology
• Medical Informatics
• Machine Learning

Background:

• Returning to work is a key recovery indicator post-acute myocardial infarction (AMI).
• Identifying patients with low return-to-work (RTW) rates is crucial for timely interventions.
• Predictive models can aid in managing patient recovery and reintegration into the workforce.

Purpose of the Study:

• To develop and validate a machine learning (ML) model for predicting return-to-work in AMI patients.
• To identify key factors influencing RTW post-AMI.
• To provide a tool for healthcare professionals to assess RTW probability.

Main Methods:

• Retrospective analysis of 539 AMI patient records from the Incidence Rate of Heart Failure After Acute Myocardial Infarction With Optimal Treatment database.
• Random division of patients into training (70%) and validation (30%) cohorts.
• Evaluation of seven ML algorithms, with performance assessed using AUC, accuracy, F1 score, and Brier score.

Main Results:

• The eXtreme gradient boosting model demonstrated superior performance with an AUC of 0.821.
• The model achieved an accuracy of 0.802 and an F1 score of 0.873.
• The developed RTW score effectively stratified patients into low, medium, and high probability groups (33.3%, 60.0%, 91.7% RTW rates, respectively; P < 0.001).

Conclusions:

• A novel ML model for predicting return-to-work in AMI patients has been successfully developed.
• This model can aid in identifying patients at high risk for delayed work resumption.
• The model, accessible via a web application, serves as a valuable management tool for AMI patients.