Development of a machine learning‑based prediction model for frailty in patients with breast cancer during chemotherapy

The aim of the present study was to investigate the prevalence and risk factors of frailty in patients with breast cancer undergoing chemotherapy, and to construct and validate a predictive model for frailty.Using a convenience sampling method, 439 patients with breast cancer receiving post-operative chemotherapy from January to November, 2024 were enrolled as the modeling set.Frailty was assessed using the Chinese version of the Groningen Frailty Indicator.Data were collected using a general information questionnaire, the Hospital Anxiety and Depression Scale, the Social Support Rating Scale, the Nutritional Risk Screening 2002, the Self-rating Scale of Body Image and the Instrumental Activities of Daily Living scale.Logistic regression (LR) was used to analyze risk factors.Of note, four machine learning algorithms, namely LR, random forest (RF), support vector machine (SVM) and XGBoost were applied to develop prediction models.Model performance was evaluated through internal validation (70:30 random split of the modeling set) and external validation (another 130 patients).The results revealed that among the 439 patients, 273 (62.17%) developed frailty.Multivariate LR identified anxiety [odds ratio (OR), 1.264; 95% confidence interval (CI), 1.144-1.396],low social support (OR, 0.923; 95% CI, 0.890-0.958)and body image disturbance (OR, 1.025; 95% CI, 1.001-1.050)as independent risk factors for frailty (P<0.05).Among the machine learning models, RF exhibited the optimal performance, with an internal validation AUC of 0.97 (95% CI, 0.94-0.99)and accuracy of 89.23%; external validation yielded an AUC of 0.78 (95% CI, 0.69-0.87)and an accuracy of 73.85%.Feature importance ranking indicated that social support, anxiety, body image disturbance and depression contributed most to the prediction.On the whole, the present study demonstrates that the incidence of frailty is high among patients with breast cancer during chemotherapy, with anxiety, insufficient social support and body image disturbance being key contributing factors.The prediction model based on the RF algorithm demonstrates good discriminatory ability and may be used for the early identification of patients who are at a high risk of frailty, providing a reference for targeted interventions.