Background and Aim: Burn injury are one of the most common traumas worldwide and the sixth leading cause of death in Iran. The challenges related to the survival rate of burn patients, as well as the associated mortality cases, have led to advancements in the identification of risk factors. Early detection and recognition of these risk factors are essential, and the provision of predictive models can be beneficial. This research was conducted with the aim of reviewing the effectiveness of artificial intelligence in predicting survival in burn patients.
Materials and Methods: This study was a systematic review. A comprehensive search of Scopus, PubMed, IEEE, and Web of Science databases was conducted from inception to July 2023 following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. Keywords and Mesh terms related to burn, artificial intelligence, survival and prediction were used in the search.
Results: Out of 3599 identified studies, only nine were included in the analysis. Based on the articles reviewed, the effective factors in predicting survival or mortality in burn patients were classified into four main categories: demographic, clinical, tests and co-morbidities. Some of the known effective factors in patient survival, which have been examined in over 40% of studies, include age, gender, total body surface area, inhalation injury, and type of burn. The results showed that in the studies reviewed, the volume of the smallest dataset used in the analyses was 92 samples. In contrast, the volume of the largest dataset used was reported to be 66,611 samples. Among these studies, 33% have indicated that artificial neural network algorithms and random forest show the best performance. The criteria used to evaluate the models in the retrieved studies are diverse.
Conclusion: The use of machine learning algorithms in predicting the survival of burn patients is promising. The results obtained from identified influential factors can assist data science researchers in the data understanding phase and can serve as a roadmap in collecting the initial dataset.
