Construction and verification of risk prediction model for unplanned readmission of patients with tumor

HE Xueyang; ZHANG Xin; LI Qin; YUAN Jifeng; WANG Lifang; SONG Jianbo

doi:10.12102/j.issn.1009-6493.2025.10.003

Abstract：Objective:To explore the influencing factors of unplanned readmission of patients with tumor,to construct a risk prediction model,and to verify its effectiveness.Methods:A total of 4 818 patients with solid tumors admitted to the cancer center of a tertiary grade A hospital in Shanxi province were selected as the research subjects from January 2023 to May 2024.The influencing factors of unplanned readmission of patients with tumor were explored using Logistic regression analysis.A predictive model was constructed,and its effectiveness was evaluated.Results:Out of 4 818 cases,269 patients experienced unplanned readmission,with incidence rate of 5.58%.The results of Logistic regression analysis showed that body mass index,Eastern Cooperative Oncology Group(ECOG)score,head and neck tumors,previous total cycles of chemotherapy,previous total cycles of immunotherapy,radiation therapy administered during last hospitalization,targeted therapy administered during last hospitalization,and neutrophil to lymphocyte ratio(NLR)were the influencing factors of unplanned readmission of patients with tumor(P<0.05).A prediction model was constructed based on the influencing factors.Area under the curve(AUC)of receiver operator characteristic of the model was 0.789(95%CI 0.754-0.823).AUC of receiver operator characteristic of the model after internal validation using the Bootstrap method was 0.771(95%CI 0.761-0.776).The Hosmer-Lemeshow test results showed P=0.158.Calibration curve indicated that the predicted probability was close to the actual probability.Conclusions:The occurrence of unplanned readmission of patients with tumor was influenced by body mass index,ECOG score,head and neck tumors,previous total cycles of chemotherapy,previous total cycles of immunotherapy,radiation therapy administered during last hospitalization,targeted therapy administered during last hospitalization,and NLR.The risk prediction model for unplanned readmission of patients with tumor constructed based on these influencing factors has good efficacy.