随机森林倾向性评分方法及其在药品不良反应信号检测中的应用

发布时间：2018-05-22 16:03

本文选题：倾向性评分 + 随机森林　；参考：《中国卫生统计》2016年04期

【摘要】：目的探讨利用随机森林倾向性评分法控制混杂因素的基本思想和步骤,及其在药品不良反应信号检测中的应用。方法利用随机森林计算给定危险因素的条件下研究对象服用双膦酸盐的概率,而后分别通过倾向性评分1:1匹配,1:M匹配和回归调整法控制性别、年龄等混杂因素,分析服药双膦酸盐与骨折发生风险的关系,并与logistic回归倾向性评分法对应结果进行比较。结果随机森林倾向性评分法与logistic回归倾向性评分方法的结果是一致的。其中,倾向性评分1:1匹配样本量损失较大,且与1:M匹配和回归调整法的结果相差较大。结论随机森林倾向性评分法能有效控制药品不良反应信号检测过程中的混杂因素,可以与logistic回归倾向性评分法所得结果相互验证,提高结果的可靠性;但1∶1匹配可能不适用于药品自发呈报系统数据。
[Abstract]:Objective to explore the basic idea and procedure of controlling confounding factors by using random forest tendency scoring method and its application in the detection of adverse drug reaction signals. Methods using random forest to calculate the probability of taking bisphosphonate under the condition of given risk factors, and then to control gender, age and other mixed factors by the inclination score of 1:1 matching 1: M matching and regression adjustment. The relationship between bisphosphonates and fracture risk was analyzed and compared with the corresponding results of logistic regression evaluation. Results the results of the random forest tendency scoring method and the logistic regression tendency scoring method were consistent. Among them, the bias score of 1:1 matching sample size loss is larger, and the result of 1: M matching and regression adjustment method is quite different from that of 1: M matching and regression adjustment method. Conclusion the random forest tendency scoring method can effectively control the confounding factors in the detection of adverse drug reaction signals, and can be verified with the results obtained by logistic regression tendency scoring method, and the reliability of the results can be improved. But 1:1 matching may not be applicable to spontaneous drug reporting system data.
【作者单位】：第二军医大学卫生统计学教研室;
【基金】：国家自然科学基金(No.81373105,81502895)
【分类号】：R954

【相似文献】