Predicción de Mortalidad y Complicaciones Cardiovasculares Intrahospitalarias en Pacientes con Síndrome Coronario Agudo Mediante un Modelo Machine Learning Supervisad

Abstract

Acute Coronary Syndromes (ACS) represent the leading cause of death in the world. Clinical prediction models Clinical prediction models can be useful for decision making mainly in high-risk patients who require early surveillance and more aggressive treatments. need early surveillance and more aggressive treatment. Numerous prediction models exist for types of ACS, mostly generalized to predict risk, which questions their usefulness. This paper Proposes a simple risk prediction model based on machine learning that applies to all types of SCA and focuses on mortality. ACS and focuses on mortality and relevant in-hospital complications. The DBMIST-US technique is competitive with other undersampling and oversampling techniques to address the problems of class imbalance and overlap. and class overlap. A representative sample was found to be sufficient to optimize the performance of the classifiers. classifiers. Clinical practice is focused on prioritizing sensitivity over accuracy and specificity; however, this technique can be However, this technique can be used to develop a risk calculator that can save a similar number of patients while lowering costs due to of patients while decreasing costs because the number of false positives would decrease significantly, leading to an increase in specificity. resulting in an increase in specificity.