Interpretable machine learning model for predicting myocardial injury in intensive care unit patients using SHapley...

Area of Science:

• Critical Care Medicine
• Machine Learning in Healthcare
• Cardiovascular Research

Background:

• Myocardial injury identification in intensive care units (ICUs) is under-researched.
• Early detection of myocardial injury is crucial for patient outcomes in critical care settings.

Purpose of the Study:

• To develop and validate a machine-learning model for predicting myocardial injury in adult ICU patients.
• To identify key clinical variables associated with myocardial injury in the ICU setting.

Main Methods:

• Retrospective cohort study involving 7453 adult, non-cardiac surgery patients admitted to the ICU (2012-2022).
• Development and comparison of five machine-learning models: logistic regression, random forest, LASSO, support vector machine, and XGBoost.
• Utilized SHapley Additive exPlanations (SHAP) for model interpretability.

Main Results:

• The XGBoost model demonstrated the highest predictive performance with an area under the curve (AUC) of 0.779 and accuracy of 0.781.
• Key predictors identified by the XGBoost model included maximal heart rate, respiratory rate, temperature, minimal heart rate, age, and plasma transfusion.
• The model successfully differentiated between patients with and without myocardial injury (29% vs. 71% prevalence).

Conclusions:

• An XGBoost-based machine-learning model shows significant potential for predicting myocardial injury in ICU patients.
• This model can serve as a valuable tool for clinical decision-making and early detection of myocardial injury.
• Further research and clinical validation are warranted to integrate this predictive tool into routine ICU practice.