Prediction model for risk of frailty in elderly patients with heart failure with preserved ejection fraction based on machine learning
10.3760/cma.j.cn211501-20240424-01021
- VernacularTitle:基于机器学习的老年射血分数保留型心力衰竭患者衰弱风险预测模型研究
- Author:
Zhen ZHANG
1
;
Rui HAI
;
Rong ZHANG
;
Hui WANG
;
Yaping XU
;
Dan JIANG
Author Information
1. 新疆医科大学第一附属医院心力衰竭科，乌鲁木齐 830054
- Publication Type:Journal Article
- Keywords:
Ejection fraction preserved heart failure;
Frailty;
Machine learning;
Prediction model
- From:
Chinese Journal of Practical Nursing
2024;40(36):2849-2855
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To construct a frailty risk prediction model for elderly heart failure with preserved ejection fraction (HFpEF) patients, and to provide a new method for accurate prediction of the occurrence of frailty in clinical elderly HFpEF patients.Methods:A cross-sectional study method was used to collect clinical data related to HFpEF patients from the Department of Cardiovascular Medicine of the First Affiliated Hospital of Xinjiang Medical University from June to November 2023 using convenience sampling method, and were randomly divided into a training set and a test set in a ratio of 7∶3, based on Logistic regression, support vector machines (SVM) and random forests, respectively, to construct a frailty risk prediction model. The performance of the models was evaluated based on the area under curve (AUC), accuracy, precision, sensitivity, specificity, F1 value.Results:A total of 319 patients with HFpEF were included, 210 males and 109 females, aged 68(62,77) years, 133 of whom developed frailty (41.7%). The dataset was divided into a training set of 223 cases and a test set of 96 cases, and all three prediction models had high accuracy, and the AUC values of the Logistic regression, SVM, and random forests models were 0.874, 0.924 and 0.884, respectively, with the SVM model having the highest AUC value and the random forests model having the highest sensitivity (0.833), specificity (0.850), accuracy (0.844), and F1 value (0.872). The importance of the feature variables was further ranked based on the random forests model, and the top five feature variables were age, interleukin-6, albumin, malnutrition, hemoglobin.Conclusions:The random forests models have the best overall predictive efficacy, which is helpful for early clinical assessment and prevention of their frailty risk.
