Predicting survival rates of critically ill septic patients with heart failure using interpretable machine learning models.

This article focuses on the development and validation of an interpretable machine learning model, specifically the Deep Learning Survival (DeepSurv) model, to predict the 28-day survival rate of critically ill septic patients with heart failure (HF). Using data from the MIMIC-III and MIMIC-IV intensive care unit databases, the study identified 22 key clinical features and demonstrated that the DeepSurv model outperformed traditional scoring systems and other predictive models in both internal and external validations. The model's interpretability was enhanced through Shapley Additive Explanations (SHAP), which highlighted important prognostic factors such as use of β-blockers, angiotensin-converting enzyme inhibitors/angiotensin receptor blockers (ACEI/ARB), acute kidney injury (AKI), continuous renal replacement therapy (CRRT), and ethnicity. The study suggests that integrating this model into clinical workflows could aid clinicians in early identification of high-risk septic patients with HF, potentially improving patient outcomes.