Development and Validation of a Machine Learning Model to Predict the Risk of Medical Decision-Making Delay in Acute Myocardial Infarction Patients From Multicenter Tertiary Hospitals in China

Purpose: Timely reperfusion is critical for improving outcomes in patients with acute myocardial infarction (AMI), as every delay raises the incidence of complications and mortality. Therefore, we aimed to develop a machine-learning model that quantifies the risk of pre-hospital decision-making delay and visualizes how individual determinants modulate this risk. Patients and Methods: This retrospective study included 594 AMI patients admitted to hospitals in Hainan from January to August 2023. Data were collected via medical systems and surveys. We used the Elastic Net and Boruta algorithms for feature selection and hyperparameter optimization with grid search and 10-fold cross-validation. Six machine learning models were developed: logistic regression, random forest, support vector machine, XGBoost, decision tree, and naive Bayes. The primary metric was the Area Under the Curve (AUC), and SHapley Additive exPlanations (SHAP) were used to assess feature importance. Results: The medical decision-making delay rate was 61.78%, with a median decision time of 3.98 hours. All models showed good predictive performance, with the random forest model excelling, achieving an AUC of 0.91, accuracy of 0.92, recall of 0.98, F1 score of 0.93, and specificity of 0.81. SHAP analysis revealed that pain severity, disease type, and history of myocardial infarction were the most significant predictors of delay. Pain severity had a nonlinear relationship with delay risk, while disease type and prior infarction history showed complex interactions. Conclusion: Machine learning models, especially random forest, accurately predict the risk of delayed medical decision-making in AMI patients and reliably delineate the key drivers of such delay, thereby informing targeted clinical interventions.