双重差分模型在社区干预研究效果评价中的应用

Application of Difference in Difference Approach in Community Intervention Trial

目的介绍双重差分模型的原理和结构,及在社区干预准实验设计研究中的应用。方法以"农村初级卫生保健项目(2001～2005周期)"家庭问卷调查资料为例,采用stata9.2软件,拟合双重差分模型,并比较无协变量模型和含协变量模型的双重差分估计量。结果两种模型的差分估计量接近,含协变量模型考虑了控制变量的影响,差分估计量更加准确。结论双重差分法在社区干预准实验设计研究的效果评价中是一种适宜的方法。

Objective To introduce the principle and structure of difference in difference approach, and to apply in semi-experimental designed community intervention trial. Methods The data of ＂Rural Primary Health Care（2001-2005 periods）＂ was used to fit the model with no covariates and with covariates by stata 9.2 version. Two difference-indifference estimators were compared. Results Two difference-in-difference estimators were nearly equal. The model with covariates was more accurate after controlling some associated factors. Conclusion Difference in difference approach is suitable to evaluate the intervention effect in semiexperimental designed community intervention trial.