Key result
Eight-variable model predicts 28-day mortality in ICU patients with AMI and COPD yielding ~0.78 AUC.
Why the study?
ICU patients with AMI complicated by COPD face an elevated risk of mortality, but validated interpretable prediction models for short-term mortality in this population remain limited.
Observational (n=662)
No
Effect estimate: AUC 0.782
High short-term mortality signals need for close monitoring in AMI-COPD ICU patients; leaves open whether ML models improve risk stratification.
Background Patients with acute myocardial infarction(AMI)complicated by chronic obstructive pulmonary disease (COPD) in the intensive care unit (ICU) face a significantly elevated risk of mortality. Therefore, timely and accurate risk stratification is critical for guiding clinical decision-making. However, validated interpretable prediction models for short-term mortality in patients with both AMI and COPD remain limited. Methods We retrospectively extracted data from the MIMIC-IV database (version 3.1) and identified ICU patients with acute myocardial infarction complicated by chronic obstructive pulmonary disease. The primary outcome was 28-day in-hospital mortality. Five machine learning models were developed and compared: eXtreme Gradient Boosting (XGBoost), logistic regression (LR), gradient boosting decision tree (GBDT), light gradient boosting machine (LightGBM), and Adaptive Boosting (AdaBoost). Model discrimination was assessed using the area under the receiver operating characteristic curve (AUC), calibration was evaluated with calibration curves, and clinical utility was examined using decision curve analysis (DCA). In addition, SHAP (SHapley Additive exPlanations) analysis was used to provide interpretable visualization of model predictions. Results A total of 662 ICU patients with acute myocardial infarction and chronic obstructive pulmonary disease were included, of whom 185 (27.9%) died within 28 days of hospital admission. Among the five machine learning models, the logistic regression (LR) model demonstrated superior discriminative performance, with an AUC of 0.782 in the validation cohort. The AUCs of the other models were 0.739 for XGBoost, 0.761 for LightGBM, 0.767 for GBDT, and 0.764 for AdaBoost. In addition, the LR model demonstrated good calibration and clinical utility. Eight variables were ultimately selected to build the prediction model, including age, heart rate (HR), respiratory rate (RR), lactate dehydrogenase (LDH), blood urea nitrogen (BUN), sepsis, β -blocker use (BB), and angiotensin-converting enzyme inhibitors/angiotensin receptor blockers (ACEI or ARB). SHAP analysis was subsequently applied to identify key predictors and enhance model interpretability. Conclusion Using the logistic regression (LR), we developed a predictive model for 28-day in-hospital all-cause mortality in ICU patients with AMI and COPD. This model may help clinicians identify high-risk patients at an early stage, enabling more informed treatment decisions and more efficient allocation of medical resources.
No takes yet. Share an insight, caveat, or question.
He et al. (2026) conducted an observational in Acute myocardial infarction complicated by chronic obstructive pulmonary disease (n=662). Logistic regression prediction model vs. Other machine learning models (XGBoost, GBDT, LightGBM, AdaBoost) was evaluated on 28-day in-hospital mortality prediction (AUC) (AUC 0.782). A logistic regression model incorporating eight clinical variables demonstrated superior discriminative performance for predicting 28-day in-hospital mortality in ICU patients with AMI and COPD, achieving an AUC of 0.782.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: