Prediction Of Death Risk In Patients With Severe Myocardial Infarction Based On Machine Learning Algorithm

The treatment of patients with acute myocardial infarction in the first 24 hours can greatly affect their mortality.However,when patients with acute myocardial infarction are treated,there are differences between different countries,even between different regions of the same country and different doctors.Machine learning can effectively reduce the differences of subjective factors by learning rules and summarizing the characteristics of experience from data.In this paper,the machine learning classification model is used to predict the in-hospital outcome of patients,and the scoring card is constructed combined with the model to evaluate the patient’s condition,which can minimize the subjective differences and assist doctors in making reasonable decisions.First,2261 patients with acute myocardial infarction were screened from the Medical Information Mart for Intensive Care III(MIMIC-III)and the maximum and minimum values of 79 indicators on the first day of admission were obtained.For missing values,this study used random forest interpolation to fill in the two categories of in-hospital death and survival respectively.As it is common in ICU that the indicators of patients are far beyond the normal range or the indicators differ greatly between patients,the outliers are not processed.In the feature engineering part,nine features were selected by adding L1 regularization logistic regression model,these features are chi-square boxed and WOE coded,then over-sampling is used to deal with the problem of category imbalance in the data.Then,six machine learning classification models,including decision tree,support vector machine and random forest,were used to predict the in-hospital outcomes of patients.By comparing the performance of each model,it was concluded that the Logistic Regression model had the best prediction effect,with an accuracy of 0.9072,sensitivity of 0.9268 and specificity of 0.9045,the AUC value is0.9818.Compared with existing studies,all indicators performed well,which could better predict the in-hospital outcomes of patients.Finally,the logistic regression model was used to establish a scoring system for acute myocardial infarction patients.By comparing the score card AMI＿FIN constructed in this paper with the existing scoring systems acute physiology score(APS-III),oxford acute disease severity score(OASIS)and simplified acute physiology score(SAPS-II),the results showed that AMI＿FIN had better prognostic effect on patients with acute myocardial infarction.