Machine Learning-based Risk Prediction Of Malignant Arrhythmia In Hospitalized Patients With Heart Failure

Objectives1.Through this regional multicenter,prospective heart failure(HF)cohort study,we analyzed the epidemiological characteristics and clinical diagnosis and treatment status of hospitalized HF patients in Anhui Province,and conducted horizontal and vertical comparisons with domestic and foreign related studies,so as to provide a reference basis for improving the diagnosis and treatment of HF in Anhui Province and conducting follow-up studies.2.To investigate the incidence of in-hospital malignant arrhythmic(MA)events and mortality due to MA events in HF patients in Anhui Province,and to screen important risk factors for in-hospital MA events in HF patients.3.To address the clinical challenge of MA risk assessment in hospitalized HF patients,explore the feasibility of applying machine learning methods to construct MA risk prediction models and validate the prediction model performance.Methods1.The subjects in this study were obtained from the Anhui Heart Failure Cohort Study(registration number:ChiCTR-POC-16010100).The study subjects were patients who received inpatient treatment for HF at the cardiology departments of the 16 participating hospitals between December 2016 and October 2018.The clinical data of 2794 HF patients were included in the final statistical analysis according to the inclusion and exclusion criteria of the study.Detailed information on patient demographics,HF etiology and comorbidities,history of cardiac surgery,vital signs on admission,ECG and echocardiography indices,laboratory tests and medications taken during hospitalisation were collected.The patient’s in-hospital MA events and deaths were recorded.In this study,MA were defined as sustained ventricular tachycardia or ventricular fibrillation requiring intravenous antiarrhythmic medication or electrical cardioversion or defibrillation intervention.2.The subjects were divided into MA event group and non-MA event group,nd the differences in baseline characteristics between the two groups were analyzed to screen the potential risk factors for MA events.The study subjects were randomly divided into a training set(1964 cases)and a validation set(830 cases)at a ratio of 7:3.Model development and preliminary evaluation were performed in the training set,and model performance was verified with the validation set.The risk prediction models were constructed using Least Absolute Shrinkage and Selection Operator(Lasso)-Logistic regression,Multivariate Adaptive Regression Splines(MARS),Classification And Regression Tree(CART),Random Forest(RF)and the eXtreme Gradient Boosting(XGboost)algorithm.Variable importance was ranked according to the contribution of clinical features to MA events.Receiver operating characteristic(ROC)curve was drawn,and the model discrimination was quantified by the area under the curve(AUC).The calibration curve was drawn,and Brier score was calculated to evaluate the model calibration.In addition,the clinical utility of the Lasso-Logistic model was evaluated by decision curve analysis(DCA).Results1.The median(Q1,Q3)age of the study population was 70(61,77)years,and 60.5%were male.Patients with New York Heart Association(NYHA)functional class Ⅱ-Ⅳ accounted for 18.2%,50.3%,and 31.6%,respectively.The proportion of patients with HFpEF(43.1%)was higher than that of patients with HFrEF(39.0%)and HFmrEF(17.9%).The most common etiology or comorbidity of hospitalized HF patients was hypertension(51.5%),followed by coronary artery disease(45.5%),atrial fibrillation or flutter(35.6%),and diabetes mellitus(22.2%).The utilization rates of ARNI/ACEI/ARB,beta-blockers and spironolactone during hospitalization were 65.6%,64.1%and 89.4%,respectively.The median length of stay was 9(7,12)days,117 patients had an MA event during their hospitalization(4.2%),and 49 patients died in-hospital(1.8%).Of these,14 deaths(0.5%,or 28.6%of total deaths)were due to MA events.Univariate analysis of in-hospital MA events suggested that 56 clinical characteristics were potential risk factors for MA events.2.This study applied 5 machine learning algorithms to construct 6 independent MA risk prediction models.In the training set,the AUC of the XGBoost model was 0.998[95%CI 0.997-1.000],which was higher than the other models(all P<0.001).In the validation set,there was no significant difference in AUC of Lasso-Logistic model 1(AUC:0.867[95%CI 0.819-0.915]),Lasso-Logistic model 2(AUC:0.828[95%CI 0.764-0.892]),MARS model(AUC:0.852[95%CI 0.793-0.910]),RF model(AUC:0.804[95%CI 0.726-0.881])and XGBoost model(AUC:0.864[95%CI 0.810-0.918];all P>0.05),which were higher than that of CART model(AUC:0.743[95%CI 0.661-0.824];all P<0.05).Brier scores for all prediction models were less than 0.25.The DCA results showed that the Lasso-Logistic model had a clinical net benefit,with Lasso-Logistic model 1 outperforming model 2 when the threshold probability was≤40%.Oral-antiarrhythmic drugs,Left bundle branch block(LBBB),serum magnesium,D-dimer and random blood glucose were significant risk factors in half and more of the prediction models.Conclusions1.This study provides a more comprehensive description of the epidemiological characteristics and clinical diagnosis and treatment status of hospitalized HF patients in Anhui region,reporting and analyzing for the first time the incidence of in-hospital MA events and its risk factors in regional HF patients,which provides an important reference for public health intervention and further research.2.The current study findings suggest that machine learning models based on the Lasso-Logistic regression,MARS,RF and XGBoost algorithms can effectively predict the risk of MA in hospitalized HF patients.Among them,Lasso regularization overcomes the shortcomings of traditional statistical analysis methods in complex data processing and feature selection.Compared with other models,the joint Lasso-Logistic model combines prediction accuracy,clinical interpretability and ease of use.3.Oral antiarrhythmic drugs,LBBB,serum magnesium,D-dimer and random blood glucose are important risk factors for MA events during hospitalization in patients with HF and should be given increased attention in clinical work.