Abstract:Objective To identify the risk factors for frailty in hospitalized cancer patients, and to build a risk prediction model and test the predictive power. Methods Totally, 570 cancer patients hospitalized in two class A tertiary hospitals in Guangzhou were selected and divided into a modeling set of 422 and a validation set of 148.A self-designed sociodemographic data questionnaire, the Fried Frailty Phenotype, M.D. Anderson Symptom Inventory, Hospital Anxiety and Depression Scale, Demoralization Scale Mandarin Version, Nutritional Risk Screening 2002 and Barthel Index were used for survey, and the laboratory indicators such as interleukin-6 and C-reactive protein were collected at the same time.Risk factors of frailty in hospitalized cancer patients were identified through logistic regression analysis to determine the risk prediction model. Results The prevalence of frailty in hospitalized cancer patients was 33.7%.The predictors of frailty were stoma, nutritional status, depression, neuropsychological symptom cluster, gastrointestinal symptom cluster, D-dimer and interleukin-6.The area under the ROC curve of the modeling set and validation set was 0.788 and 0.735 respectively.The Brier score was 0.205 and the calibration slope was 0.625.The decision curve analysis de-monstrated that the nomogram was clinically useful. Conclusion The prevalence of frailty in hospitalized cancer patients is relatively high.The nomogram prediction model has satisfactory predictive power, which can be used for identifying and screening frailty in hospitalized cancer patients.