摘要
目的:依托临床ADE主动监测与智能评估警示系统-Ⅱ(ADE-ASAS-Ⅱ)构建住院人群癫痫发作自动监测模块,为癫痫发作大样本真实世界研究提供高效的数据挖掘工具。方法:搜集指南、文献、自发报告中与癫痫相关的描述词为初始关键词集,通过预实验对初始关键词集进行初筛分类,利用文本分类技术与决策树建立报警规则,利用ADE-ASAS-Ⅱ的自定义功能与屏蔽功能对模块进行调试,确定模块最佳设置。扩大监测样本量对模块进行验证,对阳性病例的人口学特征及发作原因进行统计描述。结果:以5557例经人工审查的住院患者为测试数据,对模块进行反复调试后,最终确定决策树各分支报警关键词共37个,标题屏蔽关键词12个,模块阳性预测值(PPV)为13.86%,召回率(R)为100.00%。监测我院2021年5月共14549例在院患者,通过纳排甄别得到90例癫痫发作患者,PPV为14.59%,发生率为0.62%,其中急性症状性癫痫发作53例,以强直阵挛发作为主,发作原因以神经系统肿瘤手术最为常见。结论:基于ADE-ASAS-Ⅱ建立的癫痫发作主动监测模块,可以高效、全面、快速的获取住院人群中的目标病例,能够为癫痫发作大样本真实世界研究提供可靠的文本数据挖掘工具。
Objective:To establish a module for automatic monitoring epileptic seizures in hospitalized population based on the adverse drug event active surveillance and assessment system-Ⅱ(ADE-ASAS-Ⅱ),and provide an efficient data mining tool for real-world research on large samples of epileptic seizures.Methods:The initial collection of keywords related to epilepsy were established by searching guidelines,literature and spontaneous reports,which were preliminarily screened and classified through pre-experiments.And the alarm rules were built using text classification technology and decision tree.The module was optimized by the function of custom and shielding of ADE-ASAS-Ⅱto determine the best settings of the alarm rules.The monitoring sample size was expanded to verify the module,and the demographic characteristics and causes of the positive cases were described.Results:A total of 37 alarm keywords in each branch of the decision tree and 12 keywords for title shielding in the module were determined after repeated tests based on 5557 manually reviewed inpatients.The positive predictive value(PPV)of the module was 13.86%,and the recall rate(R)was 100.00%.A total of 14549 hospitalized patients in May 2021 were monitored by the module,among which 90 patients with epileptic seizures were screened by inclusion and exclusion criteria with a PPV of 14.59%and an incidence rate of 0.62%.53 cases were acute symptomatic seizures,and the most frequent type was tonic-clonic seizure.Nervous system neoplasm surgery was the most common reason for epileptic seizure.Conclusion:The active monitoring module of epileptic seizure based on ADE-ASAS-Ⅱcould efficiently,comprehensively and rapidly obtain target cases from the inpatients,which could provide a reliable text data mining tool for real-world research on large samples of epileptic seizures.
作者
卢京川
郭代红
高奥
伏安
李超
郭海丽
王天琳
石廷永
LU Jing-chuan;GUO Dai-hong;GAO Ao;FU An;LI Chao;GUO Hai-li;WANG Tian-lin;SHI Ting-yong(Department of Pharmacy,Medical Supplies Center of Chinese PLA General Hospital,Beijing 100853,China;College of Pharmacy,Chongqing Medical University,Chongqing 400016,China;Kanglianda Software Corporation,Beijing 100028,China)
出处
《中国药物应用与监测》
CAS
2022年第4期248-253,共6页
Chinese Journal of Drug Application and Monitoring
基金
2017年军事医学创新工程重点项目(17CXZ010)
中国研究型医院学会专项课题-临床重点药品的使用监测和评价研究专项(Y2021FH-YWPJ01)。
关键词
癫痫发作
数据挖掘
自动监测
决策树
真实世界研究
Epileptic seizures
Data mining
Automatic monitoring
Decision tree
Real world research
