Why carry out this study?

Sepsis is a common cause for hospitalization associated with a high mortality rate and morbidity.

Early detection of septic patients with potential for acute deterioration has been proven effective in improving clinical outcomes. However, relatively few constructed models have been applied for practical use due to the black box in machine learning.

This study aimed to develop and validate an interpretable machine-learning model based on clinical features for early predicting in-hospital mortality in critically ill patients with sepsis by using the SHapley Additive exPlanations (SHAP) method.

What was learned from the study?

We demonstrated the potential of machine-learning approaches for early predicting prognosis in patients with sepsis.

The SHAP method could improve the interpretability of machine-learning models and help clinicians better understand the reasoning behind the outcome.