基于数据挖掘的药品风险监测指标体系构建

The construction of drug risk monitoring index system based on data mining

目的利用大数据对药品全生命周期中可能存在的风险进行分析并建立风险指标体系。方法通过数据挖掘收集可能影响药品安全的因素,对数据进行汇总分析及清洗,梳理风险因素清单,采用德尔菲法对风险因素严重程度、可能性和可监测性进行判断,建立风险指标体系。结果在药品的生产、配送、流通、使用全生命周期中,安全风险来自于产品自身、企业、区域3个层面。其风险指标体系包括:产品层面风险指标有抽检不合格、严重不良反应率、普通不良反应率、文号临期、供应商变更、工艺变更、上报数据异常、退货率异常、工艺控制中关键指标离散度异常、中控报废率异常、冷链异常等;企业层面的风险指标有产品线存在高风险产品、被证实的投诉举报、检查有严重缺陷、行政处罚、政策环境高风险、企业信用异常、行业信用异常、法人失信、企业经济问题、药品质量舆情等。结论基于数据挖掘建立的药品风险监测指标体系可以科学灵敏地对药品风险进行预警从而开展药品智慧监管。

Objective:To analyze the possible risks in the entire life cycle of pharmaceutical products and establish a risk index system for monitoring by using big data.Methods:Collected the factors that may affect drug safety through data mining.Performed cluster analysis to judge the relational sequence and found out the index for risk early warning system design.Synthesized and ranked the risk index using level analysis method and Delphi method.Results:On the basis combining the current index,the index system of monitoring and discerning of pharmaceutical risk was established.In the entire life cycle of drug manufacturing,distribution and use,the safety risks come from three levels:the pharmaceutical product itself,the industries and regional risks.Risk indicators at the product level include sampling inspection unqualified,serious adverse reaction rate,ordinary adverse reaction rate,document number expiration,supplier change,process change,abnormal reported data,abnormal return rate,abnormal dispersion of key indicators in process control,abnormal scrap rate in central control,abnormal cold chain,etc.Risk indicators at the industries level include high-risk medicine lines,confirmed complaint reports,inspection with serious defects under supervision,administrative penalties,high risk of policy environment,abnormal enterprise credit,abnormal industry credit,corporate breach of trust,enterprise economic problems,public opinion on drug quality,etc.Conclusion:The index system of pharmaceutical risk monitoring established based on data mining can scientifically and sensitively provide early warning of drug risks and thus carry out intelligent drug supervision.