• Hilmi Yazıcı, Onur Ugurlu, Yesim Aygul, Mehmet Yildirim, and Ahmet Deniz Uçar.
• Department of General Surgery, Marmara University, Pendik Training and Research Hospital, İstanbul-Türkiye.
• Ulus Travma Acil Cerrahi Derg. 2023 Oct 1; 29 (10): 113011371130-1137.

BackgroundBurns is one of the most common traumas worldwide. Severely injured burn patients have an increased risk for mortality and morbidity. This study aimed to evaluate well-known risk factors for burn mortality and comparison of six machine learn-ing (ML) Algorithms' predictive performances.MethodsThe medical records of patients who had burn injuries treated at Izmir Bozyaka Training and Research Hospital's Burn Treatment Center were examined retrospectively. Patients' demographics such as age and gender, total burned surface area (TBSA), Inhalation injury (II), full-thickness burns (FTBSA), and burn types (BT) were recorded and used as input features in ML models. Pa-tients were analyzed under two groups: Survivors and Non-Survivors. Six ML algorithms, including k-Nearest Neighbor, Decision Tree, Random Forest, Support Vector Machine, Multi-Layer Perceptron, and AdaBoost (AB), were used for predicting mortality. Several different input feature combinations were evaluated for each algorithm.ResultsThe number of eligible patients was 363. All six parameters (TBSA, Gender, FTBSA, II, Age, BT) that were included in ML algorithms showed a significant difference (p<0.001). The results show that AB algorithm using all input features had the best predic-tion performance with an accuracy of 90% and an area under the curve of 92%.ConclusionML algorithms showed strong predictive performance in burn mortality. The development of an ML algorithm with the right input features could be useful in the clinical practice. Further investigations are needed on this topic.

7 keywords...

hide…