数据挖掘在疫苗风险监测中的应用研究

Research of Data Mining in Monitoring Risk of Vaccines

目的探索适合我国的疫苗不良事件数据挖掘方法。方法检索各种数据挖掘有关文献,对其进行研究、分析和比较。结果目前各国普遍使用的信号检测方法主要有频数法和贝叶斯法,前者主要有比例报告比法、报告比值比法、MHRA法等;后者包括贝叶斯判别可信区间递进神经网络模型与相对比值法等。结论在疫苗风险监测中应用信号检测方法是可行性,建议综合采用BCPNN、PRR、ROR、MHRA等多种方法。

Objective To probe into data mining methods for vaccines adverse event in China.Methods Literatures concerning data mining methods were reviewed for further analysis and comparison.Results Frequency method and Bayesian method were the two main data mining algorithms used in pharmacovigilance in various countries nowadays.The former includes proportional reporting ratio（PRR）,reporting odds ratio（ROR） and MHRA and the latter includes Bayesian Confidence Propagation Neural Network（BCPNN） and relative rate（RR）.Conclusion Data mining in monitoring risk of vaccines is feasible.It is suggested that BCPNN,PRR,ROR and MHRA are integratedly used in the signal detection.