The XGBoost machine learning model effectively predicted in-hospital mortality risk in patients with acute myocardial infarction and diabetes mellitus, achieving an AUC of 0.845 in the external validation set.
Observational (n=1,446)
Sí
Can a machine learning-based model accurately predict in-hospital mortality risk in patients with acute myocardial infarction complicated by diabetes mellitus?
An XGBoost-based machine learning model using seven routine clinical variables can predict in-hospital mortality in patients with AMI and diabetes, providing an interpretable tool for early risk stratification.
Estimación del efecto: AUC 0.845 (95% CI 0.807-0.884)
Objective: To develop and validate a machine learning-based (ML) predictive model for in-hospital mortality risk in patients with acute myocardial infarction (AMI) complicated by diabetes mellitus (DM). Methods: This retrospective study enrolled AMI patients with DM from the Affiliated Hospital of North Sichuan Medical College and the MIMIC-IV database. Common variables identified by both LASSO regression and the Boruta algorithm were selected as the final feature set. Utilizing the MIMIC-IV database, predictive models were constructed incorporating seven machine learning algorithms based on these common variables. The comprehensive performance of these models was evaluated through extensive metrics in both internal and external validation sets to identify the preferred model. Finally, the SHapley Additive exPlanations (SHAP) method was employed to quantitatively analyze and visually display the feature contributions of the preferred model. Results: Seven predictors were identified in this study through variable selection using two distinct methods, including heart rate, neutrophil count, monocyte count, neutrophil-to-lymphocyte ratio (NLR), serum albumin, total bilirubin, and urea nitrogen. Seven different ML models were built based on these predictors. Comprehensive performance evaluation across multiple metrics in both internal and external validation sets has shown that the XGBoost-based model achieved the numerically highest AUC and was selected as the preferred model. By employing the SHAP method for visual interpretation of this model, the interpretability and clinical credibility of the in-hospital mortality prediction model were significantly enhanced. This model can provide valuable auxiliary support in identifying high-risk patients and implementing early intervention measures. Conclusion: Interpretable machine learning models have been developed to predict in-hospital mortality risk in patients with AMI complicated by DM, providing insights into the influence of various features on the prediction outcome. Therefore, this model can serve as an exploratory and auxiliary risk stratification tool limited to clinical settings similar to our study cohorts, and it is not intended for generalized broad clinical application.
Zeng et al. (Mon,) conducted a observational in Acute myocardial infarction complicated by diabetes mellitus (n=1,446). XGBoost machine learning predictive model vs. Other machine learning models (LR, KNN, SVM, DT, LightGBM, AdaBoost) was evaluated on In-hospital all-cause mortality (AUC 0.845, 95% CI 0.807-0.884). The XGBoost machine learning model effectively predicted in-hospital mortality risk in patients with acute myocardial infarction and diabetes mellitus, achieving an AUC of 0.845 in the external validation set.