背景：生存期延长的乳腺癌（BC）患者的虚弱发生率较高。本研究旨在开发和验证一种结合社会人口统计学因素（SF）和疾病相关因素（DRF）的新型模型，以识别具有延长生存期的BC患者的虚弱。
方法：这项横断面研究检查了一个大型城市学术医疗中心的1167名患者的数据。在训练集中构建了三种类型的预测模型（817名患者）：SF模型，DRF模型，和SF+DRF模型(组合模型)。使用受试者工作特征(ROC)曲线评估模型性能和有效性，校准图和决策曲线分析(DCA)。然后该模型随后在验证集上进行验证。
结果：延长生存期的BC患者衰弱发生率为35.8%。我们确定了六个独立的危险因素，包括年龄，健康状况,化疗,内分泌治疗,合并症和口服药物的数量。最终,我们构建了一个脆弱的最优模型(组合模型C)。预测模型在训练集AUC：0.754，（95％CI，0.719-0.789；灵敏度：76.8％，特异性：62.2%）和验证集AUC：0.805，（95%CI，0.76-0.85），灵敏度：60.8%，特异性：分别为87.1%）。为训练集和验证集构建了预测列线图。进行了校准和DCA，这表明临床模型具有令人满意的校准和临床实用性。最终,我们将预测模型实现为移动友好的Web应用程序,该应用程序为BC提供准确和个性化的预测.
结论：本研究表明，延长生存期的BC患者的虚弱患病率为35.8%。我们开发了一种筛选脆弱的新模型，这可能为虚弱筛查和预防提供证据。
BACKGROUND: Breast cancer (BC) patients with extended survival show a higher incidence of frailty. This study aimed to develop and validate a novel model combining sociodemographic factors (SF) and disease-related factors (DRF) to identify frailty in BC patients with extended survival.
METHODS: This cross-sectional study examined data from 1167 patients admitted to a large urban academic medical centre. Three types of predictive models were constructed in the training set (817 patients): the SF model, the DRF model, and the SF + DRF model (combined model). The model performance and effectiveness were assessed using receiver operating characteristic (ROC) curves, calibration plots and decision curves analysis (DCA). Then the model was subsequently validated on the validation set.
RESULTS: The incidence of frailty in BC patients with extended survival was 35.8%. We identified six independent risk factors including age, health status, chemotherapy, endocrine therapy, number of comorbidities and oral medications. Ultimately, we constructed an optimal model (combined model C) for frailty. The predictive model showed significantly high discriminative accuracy in the training set AUC: 0.754, (95% CI, 0.719-0.789; sensitivity: 76.8%, specificity: 62.2%) and validation set AUC: 0.805, (95% CI, 0.76-0.85), sensitivity: 60.8%, specificity: 87.1%) respectively. A prediction nomogram was constructed for the training and validation sets. Calibration and DCA were performed, which indicated that the clinical model presented satisfactory calibration and clinical utility. Ultimately, we implemented the prediction model into a mobile-friendly web application that provides an accurate and individualized prediction for BC.
CONCLUSIONS: The present study demonstrated that the prevalence of frailty in BC patients with extended survival was 35.8%. We developed a novel model for screening frailty, which may provide evidence for frailty screening and prevention.
