Li Xiuyan, Lei Wanlin, Wang Maofeng, Xu Lili
Intensive Care Unit, Affiliated Dongyang Hospital, Wenzhou Medical University, Dongyang, Zhejiang, 322100, People's Republic of China.
Department of Biomedical Sciences Laboratory, Affiliated Dongyang Hospital, Wenzhou Medical University, Dongyang, Zhejiang, 322100, People's Republic of China.
Risk Manag Healthc Policy. 2025 Jun 25;18:2107-2120. doi: 10.2147/RMHP.S529488. eCollection 2025.
Third-generation cephalosporins, while widely prescribed, carry underquantified thrombocytopenia risks in older adults. This study aimed to develop and validate a clinical prediction model for cephalosporin-associated thrombocytopenia in hospitalized patients aged over 65 years.
A retrospective cohort (2019~2023) initially included 45,779 cephalosporin treated patients. After applying exclusion criteria, 12,917 patients were analyzed. Predictors were selected via LASSO regression, with backward elimination multivariate logistic regression constructing a nomogram. Model performance was assessed using AUC, calibration curves, and decision curve analysis (DCA) in training and testing sets.
The final model identified eight predictors: baseline platelet count (PLT), red blood cell count (RBC), presence of tumor, renal insufficiency (RI), liver cirrhosis (LC), meropenem use, use of antifungal drugs (AD), and daily usage frequency (DUF). It demonstrated strong discrimination (training AUC 0.82 [95% CI 0.79-0.85]; testing AUC 0.80 [0.76-0.84]) and calibration (Brier score 0.057). DCA confirmed clinical utility across wide risk thresholds.
This nomogram tool enables rapid thrombocytopenia risk assessment in elderly patients receiving cephalosporins. Clinically, it guides antibiotic selection by quantifying comorbidity-drug interactions, and improves toxicity monitoring accuracy in complex geriatric cases with polypharmacy.
第三代头孢菌素虽然广泛应用于临床，但老年患者使用时血小板减少风险的量化尚不明确。本研究旨在建立并验证一个针对65岁以上住院患者头孢菌素相关性血小板减少的临床预测模型。
一项回顾性队列研究（2019年至2023年）最初纳入了45779例接受头孢菌素治疗的患者。应用排除标准后，对12917例患者进行了分析。通过LASSO回归选择预测因素，并采用向后逐步排除法进行多变量逻辑回归分析以构建列线图。在训练集和测试集中，使用AUC、校准曲线和决策曲线分析（DCA）来评估模型性能。
最终模型确定了8个预测因素：基线血小板计数（PLT）、红细胞计数（RBC）、肿瘤的存在、肾功能不全（RI）、肝硬化（LC）、美罗培南的使用、抗真菌药物（AD）的使用以及每日使用频率（DUF）。该模型显示出较强的区分能力（训练集AUC为0.82 [95% CI 0.79 - 0.85]；测试集AUC为0.80 [0.76 - 0.84]）和校准度（Brier评分为0.057）。DCA证实了该模型在广泛风险阈值范围内的临床实用性。
该列线图工具能够快速评估老年头孢菌素使用者的血小板减少风险。在临床上，它通过量化合并症与药物之间的相互作用来指导抗生素的选择，并提高了在复杂的老年多药联用病例中毒性监测的准确性。
