Design and Implementation of a Fuzzy Intelligent System for Predicting Mortality in Trauma Patients in the Intensive Care Unit

Abstract

Introduction: The intensive care unit is one of the most costly parts of the national health sector. These costs are largely attributable to the length of stay in the intensive care unit. For this reason, there are significant benefits in predicting patients' length of stay and the percentage of deaths in intensive care units. Therefore, in this study, a fuzzy logic based intelligent system was designed to predict the percentage of deaths in trauma patients in the intensive care unit. Method: Data needed to design the system were collected from patient files from 2010 to 2012. Then, the system was run using data collected from each file and the system diagnosis was compared with the final diagnosis recorded in the patient file. The proposed neuro-fuzzy model was compared with five other intelligent models. This comparison was calculated and evaluated based on sensitivity, accuracy, specificity, and the area under the ROC curve. Results: The accuracy of these six models was approximately 83%, 81%, 80%, 75%, 82% and 81%, respectively. Conclusion: The neuro-fuzzy model was evaluated as the best model and had the highest accuracy. This model also had the highest area under the ROC curve. Therefore, it is recommended to use neuro-fuzzy model to diagnose and predict the percentage of deaths in trauma patients in the intensive care unit. This is important in health-related research particularly in allocating therapeutic resources to people at risk.

Keywords

• trauma patients
• intensive care unit
• classification
• prediction
• intelligent models
• anfis naïve bayes
• trees random forest 1nn
• adaboost
• svm