Tang Yang, Zhang Zhengyu, Yu Yue, He Yuxin, Yuan Yuan, Wu Xin, Xu Qian, Niu Jianhua, Wu Xiaoxin, Tan Juntao
Department of Cardiology, Affiliated Banan Hospital of Chongqing Medical University, Chongqing, 401320, People's Republic of China.
Medical Records Department, The First Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, Zhejiang, 310003, People's Republic of China.
Diabetes Metab Syndr Obes. 2025 Jun 9;18:1873-1889. doi: 10.2147/DMSO.S527018. eCollection 2025.
Older patients with type 2 diabetes mellitus (T2DM) often face severe health challenges. This study aims to develop and validate a predictive model for estimating in-hospital death risk in this population.
Clinical data of 17,421 patients with T2DM aged ≥ 65 years admitted to six hospitals in southwest China were collected retrospectively. Model performance was assessed through area under the receiver operating characteristic curve (AUROC) analysis and calibration plots. Clinical utility was evaluated using decision curve analysis (DCA) and clinical impact curve (CIC).
The overall in-hospital death rate was 3.19% (556 cases). Eleven independent predictors were identified: age, gender, history of surgery, Charlson Comorbidity Index score, coronary heart disease, chronic obstructive pulmonary disease, serum levels of creatinine, albumin, glycated hemoglobin, nutritional support drug use, and antibiotic drug use. The multivariable model demonstrated robust predictive accuracy with AUROC values of 0.873 (95% CI: 0.857-0.889) in training set, 0.830 (0.797-0.864) in internal validation set, and 0.834 (0.757-0.911) in external validation set. Bootstrap validation (n=1,000 resamples) confirmed adequate calibration. DCA and CIC analyses revealed substantial clinical net benefit across threshold probabilities. An interactive web-based calculator was implemented for clinical application (https://cqykdxtjt.shinyapps.io/in_hospital_death/).
The prediction model developed in this study demonstrated robust discrimination, calibration, and clinical utility. It can assist healthcare professionals in identifying high-risk older patients with T2DM, facilitating early prevention, detection, and intervention, thereby reducing the risk of in-hospital death in this vulnerable population.
老年2型糖尿病（T2DM）患者常面临严峻的健康挑战。本研究旨在建立并验证一个预测模型，以估计该人群的院内死亡风险。
回顾性收集了中国西南部六家医院收治的17421例年龄≥65岁的T2DM患者的临床资料。通过受试者操作特征曲线下面积（AUROC）分析和校准图评估模型性能。使用决策曲线分析（DCA）和临床影响曲线（CIC）评估临床实用性。
总体院内死亡率为3.19%（556例）。确定了11个独立预测因素：年龄、性别、手术史、Charlson合并症指数评分、冠心病、慢性阻塞性肺疾病、血清肌酐水平、白蛋白、糖化血红蛋白、营养支持药物使用和抗生素药物使用。多变量模型显示出强大的预测准确性，训练集的AUROC值为0.873（95%CI：0.857 - 0.889），内部验证集为0.830（0.797 - 0.864），外部验证集为0.834（0.757 - 0.911）。自抽样验证（n = 1000次重复抽样）证实校准充分。DCA和CIC分析显示在阈值概率范围内有显著的临床净效益。实施了一个基于网络的交互式计算器用于临床应用（https://cqykdxtjt.shinyapps.io/in_hospital_death/）。
本研究开发的预测模型显示出强大的区分度、校准度和临床实用性。它可以帮助医护人员识别高危老年T2DM患者，促进早期预防、检测和干预，从而降低这一脆弱人群的院内死亡风险。
