预测拉氧头孢致凝血功能异常列线图模型的建立与验证  

Development and validation of a nomogram for predicting latamoxef-induced coagulation disorders

作　　者：白荷荷[1] 王金萍[1] 聂晓静[1] 马莉[1] 彭莉蓉[1] 王冠 Bai Hehe;Wang Jinping;Nie Xiaojing;Ma Li;Peng Lirong;Wang Guan(Department of Pharmacy,Xi′an Central Hospital,Xi′an 710003,China)

机构地区：[1]西安市中心医院药剂科,陕西西安710003

出　　处：《实用药物与临床》2023年第12期1082-1088,共7页Practical Pharmacy and Clinical Remedies

基　　金：西安市科技项目(22YXYJ0015);西安市中心医院科研项目(2022YB06)。

摘　　要：目的探讨成人住院患者拉氧头孢致凝血功能异常的预测因子,构建简便易行的列线图模型,为临床决策提供参考。方法回顾性分析2021年1月至2022年12月西安市中心医院成人住院患者资料,基于中国药物警戒系统主动监测拉氧头孢致凝血功能异常的病例。对拉氧头孢致凝血功能异常及非拉氧头孢致凝血功能异常的患者进行1∶4倾向匹配,以7∶3随机构建训练组和验证组。采用Logistic回归分析筛选拉氧头孢致凝血功能异常的独立预测因子,利用R软件rms包构建列线图模型,通过受试者操作特征曲线(Receiver operating characteristic curve,ROC曲线)和校正曲线对列线图模型进行评价。结果拉氧头孢致凝血功能异常的发生率为7.09%。手术(OR=7.030,95%CI:3.172~15.580)、肝功能不全(OR=3.558,95%CI:1.113~11.369)、肾功能不全(OR=2.492,95%CI:1.110~5.593)、累积DDDs(OR=1.155,95%CI:1.075~1.240)和营养风险(OR=10.127,95%CI:4.488~22.851)是拉氧头孢致凝血功能异常的独立预测因子。根据独立预测因子构建的列线图模型显示,训练组和验证组ROC曲线下面积(Area under the curve,AUC)分别为0.902(95%CI:0.864~0.940)和0.904(95%CI:0.832~0.944),具有良好的区分度。校准曲线显示,拉氧头孢致凝血功能异常的预测概率与实际观测概率具有较好的一致性。结论本研究构建的列线图模型具有良好的预测性能,有助于临床医生更快、更准确地识别拉氧头孢致凝血功能异常的潜在风险。Objective To explore the predictive factors of latamoxef-induced coagulation disorders in adult inpatients,and construct a simple and feasible nomogram model,which can provide references for clinical decision-making.Methods The data of adult inpatients in Xi′an Central Hospital from January 2021 to December 2022 were retrospectively analyzed,and the cases of latamoxef-induced coagulation disorders were actively monitored based on the Chinese pharmacovigilance system(CHPS).The patients with and without latamoxef-induced coagulation disorders were matched by 1∶4 preferentially,and the training group and the verification group were randomly constructed at 7∶3.Logistic regression analysis was used to screen the independent predicators of latamoxef-induced coagulation disorders,and the rms package of R software was used to construct the nomogram model.The nomogram model was evaluated by receiver operating characteristic curve(ROC curve)and calibration curve.Results The incidence of latamoxef-induced coagulation disorders was 7.09%.Surgery(OR=7.030,95%CI:3.172~15.580),hepatic dysfunction(OR=3.558,95%CI:1.113~11.369),renal dysfunction(OR=2.492,95%CI:1.110~5.593),cumulative DDDs(OR=1.155,95%CI:1.075~1.240)and nutritional risk(OR=10.127,95%CI:4.488~22.851)were independent predictors of latamoxef-induced coagulation disorders.The nomogram model constructed based on independent predictive factors showed that the area under the curve(AUC)of the training group and the validation group was 0.902(95%CI:0.864~0.940)and 0.904(95%CI:0.832~0.944),respectively,with good discrimination.The calibration curve showed that the predicted probability of latamoxef-induced coagulation disorders was in good agreement with the actual observation probability.Conclusion The nomogram model constructed in this study has good predictive performance,which helps clinicians to identify the potential risks of latamoxef-induced coagulation disorders more quickly and accurately.

关 键 词：拉氧头孢 凝血功能异常 列线图 预测模型 药物不良反应 中国药物警戒系统 

分 类 号：R969[医药卫生—药理学]