论文部分内容阅读
疑似预防接种异常反应(Adverse Events Following Immunization,AEFI)监测是疫苗警戒研究的主要方式之一,其主要目的是侦测疫苗的安全性信号。由于被动AEFI监测系统存在数据质量不稳定、疫苗接种剂次难以收集等问题,造成传统的以分母为基础的研究方法难以实施。近数十年来,疫苗警戒研究者和统计学家开始利用以分子为基础的数据挖掘技术(Data Mining Algorithms,DMAs)对AEFI监测数据进行安全性信号的侦测。目前疫苗警戒研究中最常见的DMAs方法为比例失衡分析,主要分为传统频数方法和贝叶斯方法,计算的指标主要有比例报告比、报告比值比和相对报告比;以及贝叶斯置信传播神经网络的信息组件和模糊贝叶斯伽马-泊松收缩论的经验贝叶斯几何均数。现对DMAs的基本原理及其应用进行综述。
Adverse Events Following Immunization (AEFI) monitoring is one of the major approaches to vaccine vigilance research, and its primary goal is to detect the safety profile of the vaccine. Due to the instability of data quality and the difficulty of collecting vaccination agents in passive AEFI surveillance system, the traditional denominator-based research methods are difficult to implement. For nearly a decade, vaccine vigilance researchers and statisticians have begun exploiting molecular-based data mining algorithms (DMAs) to detect safety signals from AEFI surveillance data. At present, the most common DMAs methods in vaccine vigilance research are proportion imbalance analysis, which are mainly divided into traditional frequency method and Bayesian method. The calculated indexes mainly include ratio report ratio, report ratio ratio and relative report ratio; and Bayesian belief distribution Information Elements of Neural Networks and Empirical Bayesian Geometric Mean of Fuzzy Bayesian Gamma - Poisson Shrinkage Theory. The basic principles of DMAs and their applications are reviewed.