Machine Learning Algorithms Could Be Used For The Establishment Of Prediction Model In Patients With Type 3 Cardiorenal Syndrome

Objectives: Acute kidney injury(AKI)is a common complication of critically ill patients,and its prevalence is more than 50%.Type 3 cardiorenal syndrome(CRS)refers to acute heart injury and /or dysfunction caused by AKI,which can lead to poor prognosis and death.This study aimed to establish prediction models by using machine learning algorithms and predict the occurrence and death of type 3 CRS in AKI patients.Methods: Clinical and laboratory data of patients with AKI were filtered from the Medical Information Mart for Intensive Care-IV(MIMIC-IV)database.These patients were then divided into two groups according to the incidence of cardiorenal syndrome.Prediction models were established by using algorithms of machine learning,including K nearest neighbors(KNN),support vector machine(SVM),logistic regression(LR),and random forest(RF).The area under receiver operating characteristic curve(AUC)was used to evaluate performance of prediction models.Results: A total of 6716 patients with AKI were included in this study.There were 2625 patients developed Type 3 CRS in total,among whom 665 patients died.Thirty-six variables were used to build models.The most significant risk factors associated with Type 3 CRS were age,thrombocyte,and serum phosphorus.The AUC value of the RF model was 0.91,which was better than 0.84 of LR,0.79 of KNN,and 0.83 0f SVM.For 2625 Type 3 CRS patients,machine learning prediction model found that systolic blood pressure level,aspartate aminotransferase level and lactic acid level had the great influence on patients with Type 3 CRS,and the best prediction performance was RF:AUC value was 0.93,while LR value was 0.77,KNN value was 0.76,and SVM value was 0.77.Conclusions: These findings proved that RF predictive model could effectively predicted the incidence of type 3 CRS in AKI patients and was better than other machine learning prediction models,which helps to identify high-risk AKI patients with poor prognosis and make early CRS prevention strategies.