Machine Learning Models for Predicting In-Hospital Mortality in Burn Patients
Samet Şahin1, Burak Yavuz2, Onur Karaca3
1Department of General Surgery, Muğla Sıtkı Koçman University, 48000 Muğla, Turkey.
Annali Italiani Di Chirurgia
|August 18, 2025
Summary
Machine learning models accurately predict in-hospital mortality in burn patients. Logistic Regression and Random Forest show strong potential for improving clinical decision-making in burn care.
Area of Science:
- Medical Informatics
- Computational Biology
- Burn Care Research
Background:
- In-hospital mortality prediction in burn patients is crucial for effective clinical management.
- Machine learning (ML) offers advanced analytical capabilities for complex health data.
Purpose of the Study:
- To develop and evaluate machine learning models for predicting in-hospital mortality in burn patients.
- To identify key predictors of mortality in this patient population.
Main Methods:
- A retrospective cohort study analyzed data from 218 burn patients (2015-2020).
- Seven ML models (Logistic Regression, Random Forest, SVM, Decision Tree, KNN, Naive Bayes, Gradient Boosting) were trained and evaluated.
- Key variables included demographics, burn characteristics, and inflammatory markers.
Main Results:
- The overall in-hospital mortality rate was 18.8%.
- Logistic Regression achieved the highest ROC-AUC (0.906), while Random Forest demonstrated the highest accuracy (90.9%) and recall (97.2%).
- K-Nearest Neighbors showed superior recall (99.0%).
Conclusions:
- Machine learning models, especially Logistic Regression and Random Forest, are effective in predicting burn patient mortality.
- These findings support the use of ML for data-driven prognosis and personalized treatment in burn care.
- Multicenter validation is recommended for broader applicability.
