Influencing factors and countermeasures for long-term retention of emergency patients based on random forest model

Zhu Yun; Zheng Peng; Li Qing

doi:10.3969/j.issn.1002-1949.2025.11.006

Abstract：Objective To explore the influencing factors of long-term retention of emergency patients based on random forest model,and to put forward targeted measures to optimize the emergency process.Methods A total of 480 patients in the emergency department of Huai'an First Hospital Affiliated to Nanjing Medical University from December 2024 to January 2025 were selected.According to the ratio of 7∶3,they were randomly divided into training set(336 cases)and internal validation set(144 cases).According to the emergency retention time,they were divided into＞7 h group and≤7 h group.The clinical data of patients in the training set were collected.In addition,86 emergency patients in our hospital from February to May 2025 were selected for external verification.Logistic regression analysis and random forest algorithm were used to construct the prediction model of long-term retention of emergency patients.The receiver operating characteristic(ROC)curve,calibration curve and decision curve were drawn to evaluate the prediction performance of two prediction models.In addition,86 emergency patients from our hospital from February to May 2025 were selected for external validation,and the predictive performance of the two prediction models was evaluated by using the ROC curve,calibration curve and decision curve.Results Among the 480 emergency patients,the patients with retention time≤7 h were 80.42％,the patients with retention time＞7 h accounted for 19.58％.Age,involving≥2 departments,visiting at 00:00-07:59,and Charlson comorbidity index(CCI)score were risk factors for long-term retention of emergency patients,while green channel visiting and emergency admission were protective factors(P＜0.05).Logistic regression model:logit(P)=2.176+0.156 × age+1.607 × involvement in≥2 departments+1.452 × treatment during 0:00～07:59+1.636 × CCI score-0.974 × green channel treatment-1.008 × emergency admission.The importance of each variable in the random forest model was ranked as green channel visiting,visiting time,CCI score,emergency admission,involving≥2 departments,age.In the internal and external validation,the area under curve(AUC)of the random forest model was higher than that of the Logistic regression model,and the calibration curve was very close to the ideal diagonal.However,the two lines of the random forest model fitted more closely.Both models had obvious positive net benefits,and the random forest model has higher clinical utility.Conclusion Age,involving≥2 departments,visiting time,CCI score,green channel visiting and emergency admission are the influencing factors of long-term retention of emergency patients.The random forest algorithm can effectively identify the key factors of long-term retention of emergency patients,and provide an important basis for optimizing emergency operation and refined management.