基于混合模型的药品不良反应数据遮蔽效应消除方法研究

中国药物警戒 ›› 2018, Vol. 15 ›› Issue (6): 343-347.

基于混合模型的药品不良反应数据遮蔽效应消除方法研究

魏建香¹, 张剑吟², 刘美含¹, 李明^3,4, 孙骏³, 徐厚明³

1南京邮电大学物联网学院,江苏南京 210003;
2南京邮电大学计算机学院,江苏南京 210023;
3江苏省药品不良反应监测中心,江苏南京 210002;
4中国药科大学,江苏南京 210009

收稿日期:2018-08-03 修回日期:2018-08-03 出版日期:2018-06-15 发布日期:2018-08-03
作者简介:魏建香,男,博士,副教授·硕导,药品安全风险评估。
基金资助:
国家社会科学基金（14BTQ036)：大数据环境下药品安全突发事件预警与应急管理研究; 国家食品药品监督管理局药品评价中心委托项目：信号检测数据分类依据的研究

Study on Data Masking Effect Reduction Method for Adverse Drug Reaction Based on Hybrid Model

WEI Jian-xiang¹, ZHANG Jian-yin², LIU Mei-han¹, LI Ming^3,4, SUN Jun³, XU Hou-ming³

1 School of Internet of Things, Nanjing University of Posts and Telecommunications, Jiangsu Nanjing 210003, China;
2 School of Computer Science, Nanjing University of Posts and Telecommunications, Jiangsu Nanjing 210023, China;
3 Jiangsu Center for ADR Monitoring, Jiangsu Nanjing 210002, China;
4 China Pharmaceutical University, Jiangsu Nanjing 210009, China

Received:2018-08-03 Revised:2018-08-03 Online:2018-06-15 Published:2018-08-03

摘要/Abstract

摘要： 目的针对传统不相称性测定法在药品不良反应信号检测上存在数据遮蔽现象问题,提出了一种基于混合模型的遮蔽效应消除方法。方法该模型融合了Lasso Logistic回归、移除报告及IC方法。首先利用Lasso Logistic回归法进行数据的首轮移除;第二轮则移除发生频次小于等于4的报告;接着采用IC算法进行信号检测;最后利用我国2010年药品不良反应监测报告中的中药数据进行实验并基于已知不良反应库和国家中心信息通报进行结果的验证。结果实验结果表明,与单一方法比较,此方法在召回率、查准率和F指标上均有良好的表现。结论该混合模型不仅有效地提高了药品不良反应信号检测的准确性,而且具有早期发现真阳性信号的能力。本研究为药品不良反应信号检测提供了一种更可靠的方法。

关键词: 数据遮蔽效应, 混合模型, Lasso Logistic回归, 信号检测

Abstract: Objective To propose a data masking effect reduction method based on hybrid model to solve the problem of masking in adverse drug reactions (ADRs) signal detection by traditional disproportionality measurement. Methods The model was mixed together with Lasso Logistic Regression(LLR), report removal and IC method. The first round of data removal was carried out by Lasso Logistic Regression method and the second round was to remove the reports which frequency is less than or equal to 4. Then, IC algorithm was used for signal detection. Finally, traditional Chinese medicine data of ADR monitoring reports of China in 2010 was used to conduct experiments, and the Results were tested and verified based on known adverse reaction database and Information Bulletin of National ADR Center. Results The experimental Results showed that this model had better performances in recall rate, precision and F index by comparison with each single method. Conclusion The hybrid model not only improves the accuracy of the ADR signal detection, but also has the ability to detect more true positive signals at early stage. This research may provide a more reliable method for ADR signal detection.

Key words: data masking effect, hybrid model, Lasso Logistic Regression, signal detection

中图分类号:

R994.11

魏建香, 张剑吟, 刘美含, 李明, 孙骏, 徐厚明. 基于混合模型的药品不良反应数据遮蔽效应消除方法研究[J]. 中国药物警戒, 2018, 15(6): 343-347.

WEI Jian-xiang, ZHANG Jian-yin, LIU Mei-han, LI Ming, SUN Jun, XU Hou-ming. Study on Data Masking Effect Reduction Method for Adverse Drug Reaction Based on Hybrid Model[J]. Chinese Journal of Pharmacovigilance, 2018, 15(6): 343-347.

参考文献

[1] 钱维,叶小飞,王超,等. 药品不良反应信号检测中混杂因素的控制方法[J]. 中国药物警戒, 2010,7(3):142-144.
[2] Sakaeda T, Tamon A, Kadoyama K, et al.Data Mining of the Public Version of the FDA Adverse Event Reporting System[J]. International Journal of Medical Sciences, 2013, 10(7):796.
[3] Pariente A, Didailler M, Avillach P, et al.A potential competition bias in the detection of safety signals from spontaneous reporting databases[J]. Pharmacoepidemiol Drug Saf, 2010, 19(11):1166-1171.
[4] 侯永芳,王玲,郭秀花,等. 信号检测在药品不良反应监测系统中的应用[J].中国药物警戒, 2012, 9(9):539-541.
[5] 代菲,舒丽芯,储藏,等.简述分析几种信号监测方法在药物不良事件中的应用[J].药学实践杂志, 2012,30(5):380-383.
[6] 傅政,陈文,贺佳,等.药品上市后不良反应监测及信号自动发现方法[J].药学服务与研究,2007,7(6):422-426.
[7] Maignen F, Hauben M, Dogné J M.A mathematical framework to quantify the masking effect associated with the confidence intervals of measures of disproportionality[J]. Therapeutic Advances in Drug Safety, 2017, 8(7):231.
[8] 叶小飞,程刚,郭晓晶,等.药品不良反应信号检测中遮蔽效应的影响及消除方法[J].中国药物警戒, 2013,10(1):33-35.
[9] Wang H W, Hochberg A M, Pearson R K, et al.An experimental investigation of masking in the USFDA adverse event reporting system database.[J].Drug Safety, 2010, 33(12):1117-1133.
[10] Pariente A, Didailler M, Avillach P, et al.A potential competition bias in the detection of safety signals from spontaneous reporting databases[J]. Pharmacoepidemiol Drug Saf, 2010,19(11):1166-1171.
[11] Maignen F, Hauben M, Hung E, et al.A conceptual approach to the masking effect of measures of disproportionality[J]. Pharmacoe-pidemiol Drug Saf, 2014, 23(2):208-217.
[12] Caster O, Norén G N, Madigan D, et al.Large-Scale Regression-Based Pattern Discovery in International Adverse Drug Reaction Surveillance[J]. 2008.
[13] Caster O, Norén G N, Madigan D, et al.Large-scale regression-based pattern discovery: The example of screening the WHO global drug safety database[J]. Statistical Analysis & Data Mining, 2010, 3(4):197-208.
[14] Tibshirani R.Regression Shrinkage and Selection via the Lasso[J]. Journal of the Royal Statistical Society, 2011, 73(3):273-282.
[15] 李涵. 基于数据挖掘技术的分布式网络入侵检测系统的研究[D]. 济南大学, 2003.
[16] 陈友生,缪健,梁义敏,等.常用药品不良反应信号检测方法研究进展[J].中国药物依赖性杂志, 2014,(2):89-92.