A Machine Learning Model for Predicting the Occurrence of Early Heart Failure in Patients with Acute Myocardial Infarction

Aims Heart failure (HF) remains a frequent and burdensome complication of acute myocardial infarction (AMI), posing a substantial challenge to global healthcare systems. This study aimed to develop and compare six machine learning (ML) algorithms to identify the optimal model for the early prediction of HF following AMI. Methods We retrospectively enrolled patients admitted for AMI at the Second Affiliated Hospital of Xuzhou Medical University between June 1, 2022, and December 31, 2024. Participants were categorized into HF and non-HF groups based on the occurrence of in-hospital heart failure. The cohort was randomly split into a training set (70%) and a validation set (30%) for model development and internal validation, respectively. Model performance was assessed using the receiver operating characteristic (ROC) curve, and clinical utility was evaluated via decision curve analysis (DCA). Results Among the six ML models evaluated, the extreme gradient boosting (XGBoost) algorithm demonstrated superior predictive performance. Feature importance analysis within the XGBoost model identified the top eight predictors, in descending order of contribution: high-sensitivity C reactive protein (hsCRP), age, aspartate aminotransferase (AST), left ventricular anterior–posterior diameter (LVAPD), blood urea nitrogen (BUN), albumin (ALB), glucose (GLU), and myocardial infarction type(MI). In the validation cohort, the model achieved an area under the ROC curve (AUC) of 0.818. DCA further confirmed its favourable net clinical benefit. Conclusion An XGBoost model incorporating eight readily available clinical features was developed and validated for the early prediction of HF after AMI, showing promising discriminative ability and clinical utility. This tool may assist clinicians in stratifying risk and guiding early intervention.