Machine learning-based model of probability of mortality risk in patients with cranioencephalic trauma, Hospital Ernesto Sequeira Blanco

Abstract

The objective of this article is to present the results obtained from a project, whose purpose was to develop an automatic model that would facilitate the identification of complications and mortality risks in patients with Cranioencephalic Trauma who arrive at the Ernesto Sequeira Blanco Regional Teaching Hospital in the city of Bluefields. The SCRUM framework was used for the development of the work and machine learning techniques were used based on the CRASH-2 dataset, which has a base of 20,207 randomized records of patients who have suffered cranioencephalic trauma. Two learning models, logistic regression and decision tree, were used in combination to ensure better results. The data of the first test performed, applying the regression model, showed an accuracy of 76%, a sensitivity of 77% and a specificity of 73%. In the second test, applying the decision tree model, an accuracy of 80%, a sensitivity of 81% and a specificity of 79% were obtained. The results obtained in the application of both tests showed promising results for a more accurate prediction in the cases reviewed during the internal validations. Likewise, these results show that the model can be a useful tool in the estimation of mortality risk probabilities in patients with traumatic brain injury.