基于检验数据的药品质量风险预警模型研究

Research on Drug Quality Risk Early Warning Model Based on Inspection Data

国家药品评价性抽验的品种相对集中并具有一定的延续性,可利用其海量的检验数据进行质量风险预警,提高药品监管和检验的效能。首先根据药品检验业务和数据的特点选取Apriori、马尔科夫链模型、3σ-AHP、CV-SES等算法,综合分析以上算法在药品检验数据应用上的不足,进而提出原始数据降维、重新设计频繁集生成方法、限值归一化、修正系数函数等改进,最后建立药品质量风险预警模型。采用2013至2016年广东省药品检验所药品抽验和进口检验数据对本模型进行验证,本模型生成的预警信息能有效提高药品质量风险预警的效率和准确度,对加强药品监管具有重要的参考意义。

The variety of the state sampling for drug quality evaluation is relatively concentrated and has certain continuity. The problem we try to solve in this paper is to improve the effectiveness of drug supervision and inspection by using large quantities of drug inspection data for early warning of quality risk. The approach we adopt to solve the problem is characteristics of drug testing data Select Apriori algorithm, Markov chain, 3σ-AHP algorithm, CV-SES(coefficient of variation, single exponential smoothing) Algorithm, through the research on drug quality there are many demerits in the application of drug testing and some improving plans are put forward. Such as to reduces original data dimension, redesign create frequent dataset, scope normalization, correction function for coefficients, and finally establish a drug quality risk early warning model. The impacts on our obtained results are validated by using the 2013-2016 Year drug sampling and testing data of Guangdong institute for drug control, the early warning information generated by this model have important reference significance to improve the efficiency and accuracy of drug quality risk early warning.