Data for "Development and validation of a SHAP-based machine learning model for predicting acute myocardial infarction in emergency department patients with chest pain" - Training and Validation Datasets for Predictive Model Development

After applying these criteria and removing records with substantial missing data, 232 patients were finally included in the model development cohort, of whom 110 were diagnosed with AMI and 122 were non‑AMI. An independent validation set was constructed by retrospectively including 90 consecutive chest pain patients from the same institution between January and June 2025. Thus, the training set contained 232 samples and the validation set 90 samples.

Both datasets included 17 clinical features covering demographics, laboratory parameters, and imaging results, with the outcome being AMI diagnosis. All data were fully anonymized; patient identifiers (names, hospital numbers, and contact information) were removed before sharing. The dataset is provided in XLSX format for convenient inspection.

Missing data were handled by listwise deletion, and the predictive model was built using XGBoost. A README file in the compressed package describes each variable in detail, including coding and units. The analytical code for model development and validation is available upon reasonable request to the corresponding author.