基于多维数据挖掘的穿琥宁注射液致药品不良反应发生规律的研究

摘要

OBJECTIVE:To probe into the epidemiological characteristics of adverse drug reactions ( ADR) induced by Chuanhuling injection .METHODS:Case reports on ADR induced by Chuanhuling injection during 1979-2015 were collected by retrieving CNKI , VIP, CBM and Wanfang , Microsoft ACCESS was adopted to construct database , and the Clementine 12.0 was set as the platform to perform data mining .RESULTS: A total of 183 cases of ADR induced by Chuanhuling injection were involved , which were roughly divided into 8 categories in terms of clinical features , including anaphylactic shock (58 cases), hematological damage (46 csaes), skin lesion (35 cases), respiratory system injury (16 cases ) , etc;of 60 cases were combined medication;46 cases of ADR occurred within 5 min after administration .Results of decision tree indicated that the most important classification factor of the clinical diagnoses of ADR was its occurrence time period.The association rule with the highest confidence degree was under the situation of "ADR occurrence time =0-5 min, dosage of dosing fluid =250 ml and concentration of dosing fluid =5％", 88.24％of cases were "the clinical diagnoses of ADR =anaphylactic shock"; the association rule with the highest support degree was the cases with simultaneous appearance of "age group =20-29 years old , concentration of dosing fluid =5％" and "the clinical diagnoses of ADR=anaphylactic shock" accounted for 12.57％of the total cases .The mining results Bayesian network showed that the categories of primary diseases and ADR occurrence time were the main predictors of nervous system damage .The mining results of neural network indicated that drug combination and ADR occurrence time weighted higher in the classification of respiratory system injury .CONCLUSIONS: The characteristics of ADR induced by Chuanhuling injection is associated with ADR occurrence times , drug combinations , categories of primary diseases and patients ' ages, which still needs analysis and verification with the data of larger sample size .%目的:探索穿琥宁注射液致药品不良反应(adverse drug reaction,ADR)发生的流行病学特点.方法:系统检索1979—2015年中国期刊全文数据库、中文科技期刊数据库、中文生物医学期刊数据库及万方数据库等,收集穿琥宁注射液致ADR的详细个案报告,采用Microsoft ACCESS软件构建数据库,以Clementine 12.0软件为数据挖掘平台,开展数据挖掘研究.结果:共纳入穿琥宁注射液致ADR案例183例,其临床表现大致有8类,包括过敏性休克58例、血液系统损害46例、皮肤损害35例、呼吸系统损害16例等;其中60例存在联合用药情况;46例ADR发生于给药开始后5 min内.决策树结果显示,ADR临床诊断的最重要分类因素为ADR发生时间段.置信度最高的关联规则是在"ADR发生时间段(min)=0～5 and配液用量(ml)=250 and配液浓度(％)=5"的情况下,88.24％的案例"ADR临床诊断=过敏性休克";支持度最高的关联规则是"年龄段(岁)=20～29 and配液浓度(％)=5"与"ADR临床诊断=过敏性休克"同时出现的案例占纳入总案例数的12.57％.贝叶斯网络挖掘结果显示,原发疾病类别和ADR发生时间等2个因素是神经系统损害的主要预测因子.神经网络挖掘结果显示,联合用药情况与ADR发生时间等2个因素在呼吸系统损害的分类中有较大权重值.结论:穿琥宁注射液致ADR的特点与ADR发生时间、联合用药情况、原患疾病类别和患者年龄等因素相关,尚需更大样本量的数据分析验证.