Li Zhi-wen, Liu Jian-meng, Ren Ai-guo
Institute of Reproductive and Child Health, Peking University, Beijing 100191, China.
Zhonghua Liu Xing Bing Xue Za Zhi. 2010 Feb;31(2):223-6.
In this article, we presented the rationale and calculation procedures of a propensity score weighting method, with its application in epidemiological studies. The rationale for propensity score weighting method is similar to those for traditional standardization methods. Propensity score is used to estimate the weight for each individual. As the propensity score serves the function of observed covariates, the propensity score weighting can balance the distribution of the observed covariates between the comparison groups. There are two weighting methods according to the target standard populations: the Inverse probability of treatment weighting (IPTW) and the Standardized mortality ratio weighting (SMRW). Results of the example show that the distribution of the covariates tended to be consistent after weighting, and the IPTW and SMRW methods showed similar effect estimates. Propensity score weighting method can effectively balance the distribution of the confounding factors between the compared groups in non-randomized controlled trials.
在本文中，我们介绍了倾向得分加权法的基本原理和计算程序，及其在流行病学研究中的应用。倾向得分加权法的基本原理与传统标准化方法相似。倾向得分用于估计每个个体的权重。由于倾向得分起到观察协变量的作用，倾向得分加权可以平衡比较组之间观察协变量的分布。根据目标标准人群有两种加权方法：治疗逆概率加权法（IPTW）和标准化死亡率比加权法（SMRW）。实例结果表明，加权后协变量的分布趋于一致，IPTW和SMRW方法显示出相似的效应估计值。倾向得分加权法可以有效平衡非随机对照试验中比较组之间混杂因素的分布。
