Related Experiment Videos
Interpretative machine learning for predicting 60-day mortality in burn patients with suspected infection
11Department of Vascular Surgery, the Second Affiliated Hospital of Zhejiang University School of Medicine, Hangzhou 310009, China.
World Journal of Emergency Medicine
|May 27, 2026
Summary
A machine learning model using the APACHE IV score accurately predicts mortality in infected burn patients. This interpretable model offers improved clinical decision-making for burn care and patient outcomes.
Area of Science:
- Medical Informatics
- Computational Biology
- Clinical Medicine
Background:
- Traditional burn severity scores lack accuracy in predicting mortality for infected burn patients.
- Infection significantly complicates outcomes and mortality prediction in burn cases.
Purpose of the Study:
- To develop an interpretable machine learning (ML) model for predicting 60-day mortality in burn patients with suspected infection.
- To enhance the accuracy of mortality prediction beyond existing traditional scoring systems.
Main Methods:
- Utilized data from 1,391 adult burn patients with suspected infections from the Dryad database.
- Employed Boruta and LASSO for feature selection, developing 12 ML models, with CatBoost identified as optimal.
- Assessed model performance using AUC and employed SHAP for interpretability, comparing against ABSI and rBaux scores.
Main Results:
- The CatBoost ML model demonstrated superior performance in predicting 60-day mortality compared to ABSI and rBaux.
- Key predictors identified included %TBSA, APACHE IV score, and age.
- The ML model achieved a well-balanced performance, outperforming traditional scores.
Conclusions:
- An ML model incorporating the APACHE IV score significantly improves 60-day mortality prediction in infected burn patients.
- The model's high interpretability is expected to facilitate its clinical adoption.
- This approach offers a promising tool for optimizing burn patient management and care.
