Construction of predictive model of frailty risk in elderly patients with chronic heart failure based on machine learning

HAI Rui; WANG Hui; ZHANG Rong; XU Yaping; YANG Yi

doi:10.12102/j.issn.1009-6493.2024.12.007

Abstract：Objective:To construct a predictive model of frailty risk in elderly patients with chronic heart failure(CHF)based on machine learning,and to provide a new method for accurate prediction of frailty occurrence in clinical elderly patients with CHF.Methods:Clinical data related to CHF patients from the cardiovascular medicine department of a tertiary grade A hospital in Urumqi from January 2023 to May 2023 were collected and randomly divided into training and testing sets in the ratio of 7∶3,with the occurrence of frailty as the outcome variable.The frailty risk prediction models were constructed based on four algorithms:Logistic regression(LR),decision tree(DT),random forest(RF),and support vector machines(SVM).The performance of the models was evaluated based on the area under curve(AUC),accuracy,precision,sensitivity,specificity,F1 value,and the optimal model was selected.Results:A total of 423 patients with CHF were included,182 of whom developed frailty(43%).All four prediction models had high accuracy,and the AUC values of the LR,DT,SVM,and RF models were 0.917,0.863,0.941 and 0.952,respectively,with the RF models having the highest AUC values,and the RF model had the highest accuracy,precision,sensitivity,specificity,and F1 value were the highest.The importance of the feature variables was further ranked based on the RF model,and the top five feature variables were hemoglobin,interleukin-6,albumin,malnutrition,and Charlson Comorbidity Index(CCI)scores.Conclusion:The predictive model of frailty risk in elderly patients with chronic heart failure based on RF machine learning has the best performance,which is helpful for early clinical assessment and prevention of frailty risk.