群组对照试验中应用倾向得分匹配法分析数据对干预效果评价的影响

目的 探讨应用倾向评分匹配法（propensity score matching，PSM）处理群组平行对照试验中横断面调查数据对干预效果评估的影响。方法 以某"十二五"国家科技重大专项的子课题"低档暗娼减少性病艾滋病感染干预研究"数据为例，应用PSM法对干预前后两次横断面调查的低档暗娼人群数据进行匹配，对匹配后样本开展结局变量χ2检验并拟合广义线性混合模型（generalized linear mixed models，GLMM），讨论PSM法对评估结果的影响。结果 以存在显著差别的关键特征变量作为匹配因素进行PSM后，样本量为537，干预前后的两个人群完全可比。GLMM分析结果表明，干预是梅毒感染率降低的主要因素。PSM后数据拟合模型获得的OR值为0.33，与原始数据得到的OR值（0.51）相比降低了0.18，而且前者获得的95%CI（0.16~0.70）比后者（0.27~0.96）更窄，更远离1，将干预措施降低梅毒感染风险的效果从49%提高到了67%。结论 PSM法用于系列横断面调查的群组对照干预试验，可以有效提高不同调查人群之间的可比性，降低人群差异对效果评估的影响，从而提高研究结果的准确性。

The influence of using propensity score matching method to analyze data on the effect evaluation of interventions in the parallel-group controlled trial

Objective To explore the influence on the intervention effect evaluation when applying propensity score matching (PSM) to process the cross-sectional survey data in the parallel-group controlled trial. Methods Data collected from the study of "intervention study on STD/AIDS infection reduction of low-fee female sex workers (FSWs)", a sub-project of a 12th five-year national science and technology major projects, were used as an example. PSM method was applied to match the two cross-sectional surveys data of low-fee FSWs before and after the intervention. The Chi-square test of outcome variables was carried out for the matched samples and the generalized linear mixed model (GLMM) was fitted. The influence of PSM on evaluation results was discussed. Results The sample size was 537 after PSM when using the key characteristic variables with significant difference as the matching factors. The two populations before and after intervention were completely comparable. The results of the GLMM analysis showed that intervention was the major factor for the reduction of syphilis infection. Compared with the OR value (0.51) obtained from the original data, the OR value (0.33) was 0.18 lower when using the data processed by PSM to fit the model. The confidence interval obtained from PSM data was (0.16-0.70), also narrower than original results (0.27-0.96), and farther from the number of 1. The results showed that PSM improved the accuracy of the evaluation and increased the syphilis infection reduction rate caused by intervention from 49% to 67%. Conclusions When applied in the series cross-sectional survey of parallel-group controlled trial,PSM method can effectively improve the comparability of different study populations and reduce the influence of population difference on effect evaluation, so as to improve the accuracy of the research results.