Predicting the Survival or Death of Burn Patients Using Data Mining Techniques

Abstract

Background and Objectives: Burns are wounds brought on by contact with heat, chemicals, or electricity. Proteins are altered as a result of these wounds, and the amount of intravascular fluid is decreased. The experiments show that it is impossible to say that one strategy will work with every type of data. Additionally, the information on burn patients admitted to Imam Musa Kazem Hospital (AS) to date has not been examined. Using data mining techniques, the researchers forecast whether burn patients at Imam Musa Kazem Hospital will survive or die. Methods: The data set contains the electronic records of 8280 burn patients who were admitted to the Imam Mousa Kazem Hospital of the Isfahan University of Medical Sciences between 2012 and 2021. Statistical operations are performed using the Rapidminer program. The decision tree family of approaches was used throughout this study. Random Forest, CHAID, and ID3 are some of the methods employed. Results: The ID3 strategy outperformed the other techniques in Table 2 in terms of Accuracy. More examples that were still alive could be found using the random forest method. According to this rule (support: 324-confidence: 97), if a person has a burn code of T31.1, is hospitalized in a typical ward, and is young, they lived. Conclusion: The extracted rules take into account variables including burn percentage, burn degree, hospital stay duration, gender, ward, and age. The other dataset can be utilized in the future to forecast survival or death. Results were compared to those in this publication.