Building Computational Models to Predict One-Year Mortality in ICU\n Patients with Acute Myocardial Infarction and Post Myocardial Infarction\n Syndrome

Abstract

Heart disease remains the leading cause of death in the United States.\nCompared with risk assessment guidelines that require manual calculation of\nscores, machine learning-based prediction for disease outcomes such as\nmortality can be utilized to save time and improve prediction accuracy. This\nstudy built and evaluated various machine learning models to predict one-year\nmortality in patients diagnosed with acute myocardial infarction or post\nmyocardial infarction syndrome in the MIMIC-III database. The results of the\nbest performing shallow prediction models were compared to a deep feedforward\nneural network (Deep FNN) with back propagation. We included a cohort of 5436\nadmissions. Six datasets were developed and compared. The models applying\nLogistic Model Trees (LMT) and Simple Logistic algorithms to the combined\ndataset resulted in the highest prediction accuracy at 85.12% and the highest\nAUC at .901. In addition, other factors were observed to have an impact on\noutcomes as well.\n

Ähnliche Arbeiten