Development of arandom forest-based predictive model for fear of falling in patients with first-episode cerebral infarction

LIU Lili; ZHU Ruixue; LI Huanhuan

doi:10.3969/j.issn.2096-501X.2025.05.013

Abstract：Objective To explore the influencing factors of fear of falling(FOF)in patients with first-episode cerebral infarction(CI)and to construct a random forest predictive model.Methods One hundred and sixty patients with first-episode CI admitted to the Third Affiliated Hospital of Shandong First Medical University from November 2022 to December 2023 were selected.Based on the presence of FOF,patients were divided into FOF group and non-FOF group.Clinical data of all subjects were collected.Multivariate Logistic regression was used to identify risk factors of FOF,and a random forest model was constructed using R(R4.3.3)software,and the receiver operating characteristic(ROC)curves were plotted to evaluate the model performance.Results Among the 160 patients,68 patients had FOF(42.50%).There were statistically significant differences in age,marital status,walking ability,history of falls,and cognitive impairment between the FOF group and the non-FOF group(P<0.05).Multivariate Logistic regression analysis showed that age 65-80 years old,cognitive impairment,assisted walking,no spouse,and history of falls were risk factors for FOF in patients with first-episode CI(P<0.05).The random forest model showed that the error rate was lowest when the number of trees was 74.The relatively important predictors ranked as follows:Age 65-80 years old,no spouse,cognitive impairment,use of walking aids,and history of falls.Meanwhile,the average reduction in the Gini value was proportional to the importance of each variable in the model.The area under the ROC curve(AUC)for the random forest model was 0.823,compared with 0.788 for the Logistic regression model.The predictive performance of the random forest model was slightly higher than that of the Logistic regression model(Z=2.261,P=0.024).Conclusions Age 65-80 years old,cognitive impairment,use of walking aids,no spouse,and history of falls are risk factors for FOF in patients with first-episode CI.The predictive performance of the random forest model constructed in this study is higher than that of the Logistic regression model.Medical staff could provide patients with more accurate and personalized intervention programs based on the ranking of the genetic importance of the random forest model for FOF in patients with first-episode CI.