Classification model for prediction of death from heart failure

Abstract

Heart failure is a syndrome in which the heart is unable to pump blood at optimal levels throughout the body. In recent decades, there has been a significant increase in deaths caused by this condition. This growth is associated with changes in behavior and aging, among other factors. Given the growing challenge presented by this scenario, it is essential to develop a sophisticated classification model that can predict death and that is easily integrated into existing health systems. In light of this challenge, five classification models were created for a sample of 299 patients and their results were compared. This sample contains 12 explanatory variables, in addition to the response variable, which indicates whether the patient died during treatment. The results achieved were satisfactory and show that, although more than one model presents good results, considering technical and analytical aspects, the binary logistic model is the one that presented the best balance of metrics. Because it is a high-performance, easy-to-interpret algorithm with low cost of training, availability and execution, it is concluded that the binary logistic model created in this study can be integrated into current systems in order to help health professionals with their diagnosis and, thus, they can suggest changes in habits and/or treatment to the patient so that he or she can have a longer and healthier life.