Enhancing prognostic accuracy in sepsis-induced cardiomyopathy: a machine learning approach.

Area of Science:

• Critical Care Medicine
• Cardiology
• Data Science in Healthcare

Background:

• Sepsis-induced cardiomyopathy (SIMD) is a serious complication of sepsis, leading to myocardial dysfunction and high mortality.
• Existing prognostic tools inadequately address SIMD's complex pathophysiology, necessitating advanced predictive models.

Purpose of the Study:

• To develop and validate a robust prognostic model for predicting short-term mortality in adult patients with SIMD.
• To identify key predictors of mortality in SIMD patients.

Main Methods:

• Retrospective analysis of 1068 adult SIMD patients from the MIMIC-IV database.
• Feature selection using Boruta, LASSO, and RFECV, followed by logistic regression model development.
• Model performance evaluated using ROC curves, calibration curves, decision curve analysis, and SHAP for interpretability.

Main Results:

• Seven predictors identified: APS III, age, CCI, cerebrovascular disease, ALP, lactate, and CK-MB.
• The logistic regression model demonstrated strong discrimination (AUC up to 0.88) and calibration (K-S > 0.45).
• Key predictors of mortality included acute illness severity (APS III), age, and hepatic dysfunction (ALP).

Conclusions:

• A validated logistic regression model effectively predicts short-term mortality in SIMD patients.
• The model highlights acute illness severity, aging, and hepatic dysfunction as primary determinants of mortality.
• This tool aids early risk stratification in critical care settings for SIMD patients.