Constructing a comprehensive nursing risk assessment model for older adults inpatients with multiple chronic conditions: A multicenter cross-sectional study.

AIM: To build a comprehensive nursing risk model for older adults inpatients with multiple chronic conditions to identify nursing risks. METHOD: This study was conducted in Sichuan Province, China from March 2020 to June 2022. We used Python to build logistic regression, decision tree and random forest models. RESULTS: The study included 4458 patients, 2529 were male (56.7 %) and had an average age of 74.31±6.79 years. The mean number of chronic diseases in elderly hospitalized patients was 2.81±0.96. Hypertension (67.0%), diabetes (43.4%), and coronary heart disease (31.4%) are the three diseases with the highest prevalence. There were 552 (12.4%) patients who had nursing risk event. The random forest model has the best comprehensive prediction ability. It has the highest accuracy(0.917), precision(0.667), recall(0.706), Macro-F1(0.933), and AUC values [0.933(95 % CI:0.924-0.943) ] in training set. CONCLUSION: The random forest(AUC=0.933) superior compared to logistic regression and decision tree, suggesting strong potential for clinical implementation.