Development Of Readmission Risk Prediction Model For Patients With Chronic Heart Failure Based On Random Survival Forest And Cox Risk Ratio Regression

Objectives:The purpose of this study was to describe the status of readmission in patients with chronic heart failure,and to explore the risk factors of readmission in patients with chronic heart failure by combining random survival forest and Cox proportional hazards regression,and to construct a readmission risk prediction model.For medical staff to accurately screen high-risk groups in clinical work,realize early individualized nursing intervention,improve patient survival prognosis,reduce avoidable readmission rate,reduce waste of medical resources,and enrich my country’s chronic heart failure readmission prevention strategy and accurate risk prediction for reference.Methods:This study is a prospective cohort study.From September 2020 to April 2021,402 patients with chronic heart failure in the cardiology ward of a tertiary hospital in Shandong were selected by convenience sampling method,and then the included patients were followed up for half a year.Patient-related data were collected at admission and during follow-up.Using R4.1.0,the patients included in the study were randomly divided into the modeling group and the validation group according to the ratio of 7:3.Based on the random survival forest algorithm,the feature importance of all the independent variables in the modeling group data was sorted.The best combination of variables was selected by combining the depth method,and the independent variables obtained by the above screening were subjected to univariate and multivariate Cox proportional hazards regression analysis to explore the independent risk factors of readmission in patients with chronic heart failure,and to construct chronic heart failure.The prediction model of readmission risk in heart failure patients,the results are expressed in the form of nomogram,and at the same time,internal validation is performed in the data of the modeling group,and external validation is performed in the data of the validation group.HosmerLemeshow test)and clinical decision curve(Decision Curve Analysis,DCA)to evaluate the constructed chronic heart failure readmission risk prediction model.Results:1.A total of 402 patients with chronic heart failure were included in this study,with an average age of 61.78±13.73 years,including 267 males(66.4%)and 135 females(33.6%).In the modeling group(n=303 cases),the average age was 62.17±13.72 years,of which 201 were males(66.3%);in the validation group(n=99 cases),the average age was 60.57±13.78 years,of which 66 were males(66.7%).There was no statistical difference in most variables between the two groups(P>0.05),indicating that the modeling group and the validation group in this study basically conformed to complete randomization.2.During the six-month follow-up,150 of the 402 chronic heart failure patients had readmission,and the 6-month readmission rate was 37.31%.3.The random survival forest method was used to sort the importance of all 54 variables.Through the combination of the VIMP method and the minimum depth method,the best variable combination containing 17 variables was finally screened,namely age,chronic disease outpatient insurance,Type of medical insurance,number of hospitalizations in the year before this admission,years of heart failure,discharge method,body mass index,Charlson comorbidity index,CONUT score,average red blood cell width,hemoglobin,lymphocytes,cystatin C,blood urea nitrogen,NT-proBNP,serum high-sensitivity troponin and serum potassium.The above variables were analyzed by log-rank method,and 12 variables were statistically significant(P<0.05),which were the chronic disease outpatient insurance,the number of hospitalizations in the year before the current admission,the way of leaving the hospital,the CONUT score,Mean erythrocyte width,lymphocytes,hemoglobin,cystatin C,urea nitrogen,NT-proBNP,serum high-sensitivity troponin,and serum potassium.4.The multivariate Cox analysis showed that the chronic disease outpatient insurance(P=0.003),the way of leaving the hospital(P<0.001),the number of admissions within one year(P=0.006),the CONUT score(P=0.002),the cystatin C(P=0.004),mean red blood cell width(P=0.021)and serum high-sensitivity troponin(P=0.016)were the independent risk factors for readmission in patients with chronic heart failure.5.The AUC of the prediction model constructed in the modeling group based on the above independent risk factors is 0.785,the AUC of the model after internal verification is 0.732,and the AUC of the model after external verification in the verification group is 0.720,which proves that the model has a high degree of discrimination;The model fitting curves of the two groups show that the calibration curve and the standard curve basically fit,which proves that the probability of readmission in patients with chronic heart failure predicted by the model is more consistent with the actual situation;the clinical decision curves of the two groups are displayed within a certain range,The model can obtain better clinical benefit.Conclusions:In China,the readmission rate of patients with chronic heart failure remains high,and the persistent illness and repeated admissions form a vicious circle.It suggests that medical staff should identify patients with high risk of readmission in clinical work and conduct individualized interventions as soon as possible.Maximize control of exacerbations and reduce readmissions in patients with chronic heart failure.Insured participation in chronic disease outpatient clinics,way of leaving the hospital,number of admissions within one year,CONUT score,cystatin C,mean red blood cell width and serum high-sensitivity troponin were independent risk factors for readmission in patients with chronic heart failure.In this study,the random survival forest combined with the traditional Cox proportional hazards regression modeling method,the constructed chronic heart failure readmission risk prediction model has good predictive performance,and clinical nurses can use this risk prediction model to identify high-risk patients.Implement a nurse-led,evidence-based,research-based chronic heart failure education program to improve patients’ ability to self-care,reduce their readmission rate,and reduce medical resource waste.