Development and validation of an explainable artificial intelligence model to predict in-hospital adverse events for patients with acute myocardial infarction
- 1School of Biomedical Engineering, Capital Medical University, Beijing, China
- 2Beijing Key Laboratory of Clinical Engineering Solutions for Mental Health, Capital Medical University, Beijing, China
- 3Department of Cardiology, Xuanwu Hospital, Capital Medical University, Beijing, China
Abstract
In-hospital adverse events after acute myocardial infarction (AMI) can occur abruptly and adversely affect short-term prognosis. We developed and externally validated explainable machine learning models to predict a composite of in-hospital atrial fibrillation (AF), ventricular fibrillation (VF), cardiac arrest (CA), or death using routinely available admission data. We retrospectively analyzed 7,496 patients from the MIMIC-IV database as the development cohort and 2,776 patients from Xuanwu Hospital as the external validation cohort. Predictors included demographics, medical history, and admission laboratory tests. Logistic regression, random forest, extreme gradient boosting (XGBoost), multilayer perceptron, and TabNet models were evaluated using five-fold cross-validation. Performance was assessed using discrimination, classification, calibration, and decision curve analysis. Latent class analysis identified three outcome clusters with distinct event-probability profiles. XGBoost achieved the best performance for the composite outcome in the development cohort (AUROC, 0.798). Predictive performance was moderate to good for AF and in-hospital death, with AUROCs of up to 0.834 and 0.870 in the development cohort and 0.803 and 0.849 in the external validation cohort, respectively. Discrimination was lower for VF and CA. SHapley Additive exPlanations highlighted neutrophil and leukocyte counts, N-terminal pro-B-type natriuretic peptide, renal function, glucose, glycated hemoglobin A1c, and prior AF as important predictors. These explainable models may support early risk stratification and help guide monitoring intensity and timely intervention in patients with AMI. Author summary: Acute myocardial infarction (AMI) is a life-threatening cardiovascular emergency associated with a high risk of severe in-hospital adverse events, including atrial fibrillation, ventricular fibrillation, cardiac arrest, and death. Early and accurate risk stratification using routinely available admission data remains a major clinical challenge. In this two-center retrospective study, we developed and externally validated explainable artificial intelligence models for predicting multiple in-hospital adverse events in patients with AMI. Latent class analysis was applied to identify heterogeneous phenotypic clusters of adverse events, while SHapley Additive exPlanations were used to enhance model interpretability. Our results demonstrated that tree-based models, particularly XGBoost and random forest, achieved favorable predictive performance for both composite and individual adverse events across development and external validation cohorts. Key predictors included inflammatory markers, cardiac stress indicators, renal function parameters, and metabolic biomarkers. These explainable models may support early admission-stage risk stratification, facilitate individualized monitoring strategies, and assist timely clinical intervention for hospitalized AMI patients. Overall, this study provides a clinically applicable, data-driven framework for improving inpatient risk management and optimizing resource allocation in high-risk cardiovascular care.
Related
- Physics-Informed and Explainable Artificial Intelligence for Nanomaterial-Based Biosensors: From Sensor Design and Signal Processing to Clinical Translation
- Explainable artificial intelligence reveals spatially divergent effects of global change on mammals
- Explainable Artificial Intelligence (XAI) and Molecular Modeling Techniques to Discover Putative HER2 Inhibitors
- Beyond Algorithmic Accuracy: Field Readiness, Ecological Intelligence and Real-World Validation of Artificial Intelligence for Insect Pest Management.
- Artificial Intelligence for Predictive Microbiome Engineering: Integrating Community Design and Environmental Validation
BibTeX
@article{tian2026development,
title = {{Development and validation of an explainable artificial intelligence model to predict in-hospital adverse events for patients with acute myocardial infarction}},
author = {Xican Tian and Mengyuan Ji and Tao Wang and Yingliang Yang and Pixian Yuan and Yang Zhang and Congmin Zhu and Jinggang Xia and Yu Gu and Honglei Liu},
journal = {PLOS Digital Health},
year = {2026},
month = nov,
doi = {10.1371/journal.pdig.0001812},
url = {https://journals.plos.org/digitalhealth/article?id=10.1371/journal.pdig.0001812},
note = {CC BY},
}
RIS
TY - JOUR
TI - Development and validation of an explainable artificial intelligence model to predict in-hospital adverse events for patients with acute myocardial infarction
AU - Xican Tian
AU - Mengyuan Ji
AU - Tao Wang
AU - Yingliang Yang
AU - Pixian Yuan
AU - Yang Zhang
AU - Congmin Zhu
AU - Jinggang Xia
AU - Yu Gu
AU - Honglei Liu
PY - 2026
DA - 2026/11/01
JO - PLOS Digital Health
DO - 10.1371/journal.pdig.0001812
UR - https://journals.plos.org/digitalhealth/article?id=10.1371/journal.pdig.0001812
AB - In-hospital adverse events after acute myocardial infarction (AMI) can occur abruptly and adversely affect short-term prognosis. We developed and externally validated explainable machine learning models to predict a composite of in-hospital atrial fibrillation (AF), ventricular fibrillation (VF), cardiac arrest (CA), or death using routinely available admission data. We retrospectively analyzed 7,496 patients from the MIMIC-IV database as the development cohort and 2,776 patients from Xuanwu Hospital as the external validation cohort. Predictors included demographics, medical history, and admission laboratory tests. Logistic regression, random forest, extreme gradient boosting (XGBoost), multilayer perceptron, and TabNet models were evaluated using five-fold cross-validation. Performance was assessed using discrimination, classification, calibration, and decision curve analysis. Latent class analysis identified three outcome clusters with distinct event-probability profiles. XGBoost achieved the best performance for the composite outcome in the development cohort (AUROC, 0.798). Predictive performance was moderate to good for AF and in-hospital death, with AUROCs of up to 0.834 and 0.870 in the development cohort and 0.803 and 0.849 in the external validation cohort, respectively. Discrimination was lower for VF and CA. SHapley Additive exPlanations highlighted neutrophil and leukocyte counts, N-terminal pro-B-type natriuretic peptide, renal function, glucose, glycated hemoglobin A1c, and prior AF as important predictors. These explainable models may support early risk stratification and help guide monitoring intensity and timely intervention in patients with AMI. Author summary: Acute myocardial infarction (AMI) is a life-threatening cardiovascular emergency associated with a high risk of severe in-hospital adverse events, including atrial fibrillation, ventricular fibrillation, cardiac arrest, and death. Early and accurate risk stratification using routinely available admission data remains a major clinical challenge. In this two-center retrospective study, we developed and externally validated explainable artificial intelligence models for predicting multiple in-hospital adverse events in patients with AMI. Latent class analysis was applied to identify heterogeneous phenotypic clusters of adverse events, while SHapley Additive exPlanations were used to enhance model interpretability. Our results demonstrated that tree-based models, particularly XGBoost and random forest, achieved favorable predictive performance for both composite and individual adverse events across development and external validation cohorts. Key predictors included inflammatory markers, cardiac stress indicators, renal function parameters, and metabolic biomarkers. These explainable models may support early admission-stage risk stratification, facilitate individualized monitoring strategies, and assist timely clinical intervention for hospitalized AMI patients. Overall, this study provides a clinically applicable, data-driven framework for improving inpatient risk management and optimizing resource allocation in high-risk cardiovascular care.
SN - 2767-3170
ER -