基于HIS数据的凝血功能障碍伴出血事件自动监测模块的构建与优化——以替加环素为研究案例

Construction and Optimization of Automatic Surveillance Module for Coagulation Dysfunction Complicated with Bleeding Events Based on HIS Data-Tigecycline as A Case Study

目的:依托临床药品不良事件主动监测与智能评估警示系统(ADE-ASAS-Ⅱ)构建凝血功能障碍伴出血事件监测模块,探索复杂诊断条件下药品不良反应(ADR)自动监测方法。方法:以替加环素为研究案例,基于决策树算法将触发器技术与自然语言处理技术相结合,构建量化-文本复合监测模块,并基于某大型三级甲等医院的医院信息系统(HIS)数据挖掘凝血功能障碍相关出血事件。反复优化模块设置,以人工评价结果为“金标准”,比较不同优化方案下模块监测效能以确定最佳设置。统计阳性事件的发生率、基本特征。结果:共监测2019—2022年2345例使用替加环素的患者,人工评价后确定阳性病例83例。模块初始设置下,报警985例,精确率为7.72%,召回率为91.57%,F1值为14.23%。反复优化后,报警475例,召回率不变,精确率提升至16.00%,F1值提升至27.24%;1870例被ADE-ASAS-Ⅱ系统标记为阴性病例,可减少约80%的人工评价工作量。1359例符合纳入与排除标准的患者中(阴性1276例,阳性83例),凝血功能障碍伴出血事件的发生率为6.11%。结论:凝血功能障碍伴出血事件模块的开发提升了ADE-ASAS-Ⅱ系统对复杂诊断条件下ADR的检出效率,拓展了联合结构化与非结构化HIS系统数据识别ADR的监测方法。后续需扩大样本以验证模块稳定性,并持续改进系统算法,以进一步提高监测效率。

OBJECTIVE:To construct the automatic surveillance module for coagulation dysfunction complicated with bleeding events based on adverse drug events active surveillance and assessment system(ADE-ASAS-Ⅱ),and to explore an automated method for monitoring adverse drug reactions(ADR)under complex diagnostic conditions.METHODS:Taking tigecycline as a case study,a quantitative-text composite monitoring module was constructed through the combination of trigger technology and natural language processing technology based on decision tree algorithm,and the coagulation dysfunction-associated bleeding events were mined based on the hospital information system(HIS)data of a large gradeⅢ,level A hospital.The module settings were optimized iteratively,and the manual evaluation results were used as the“gold standard”to compare the monitoring performance of modules under different optimization schemes to determine the optimal settings.The incidence of positive events and their basic characteristics were calculated.RESULTS:A total of 2345 patients using tigecycline were monitored from 2019 to 2022 were monitored,and 83 positive cases were identified after manual evaluation.Under the initial module setting,985 cases were alarmed,with a precision rate of 7.72%,a recall rate of 91.57%and F1 value of 14.23%.After iterative optimization,475 cases were alarmed,the recall rate was stable,the precision rate was increased to 16.00%and the F1 value was increased to 27.24%;1870 cases were marked as negative cases by ADE-ASAS-II,which had reduced approximately 80%of the manual evaluation workload.Of the 1359 patients who met the inclusion and exclusion criteria(1276 negative cases and 83 positive cases),the incidence of coagulation dysfunction complicated with bleeding events was 6.11%.CONCLUSIONS:The development of the module of coagulation dysfunction complicated with bleeding events has improved the detection efficiency of ADE-ASAS-II for ADR under complex diagnostic conditions,and has expanded the monitoring method for identifying ADR by combining structured and unstructured HIS data.Subsequent expansion of the sample is needed to verify the stability of the module and to continuously improve the system algorithm,so as to further enhance the monitoring efficiency.