Treatment effect bounds: An application to Swan-Ganz catheterization

详细信息	查看全文 | 推荐本文 |

• 作者：Jay Bhattacharya^a ; ^{jay@stanford.edu} ; Azeem M. Shaikh^b ; ^{amshaikh@uchicago.edu} ; Edward Vytlacil^c ; ^{edward.vytlacil@yale.edu}
• 关键词：Partial identification ; Average treatment effect ; Swan&ndash ; Ganz catheterization ; Threshold crossing model ; Simultaneous equation model
• 刊名：Journal of Econometrics
• 出版年：2012
• 期刊代码：31_03044076
• 类别：mt
• 出版时间：June, 2012
• 卷：168
• 期：2
• 页码：223-243
• 文件大小：1181 K

摘要

We reanalyze data from the observational study by on the impact of Swan-Ganz catheterization on mortality outcomes. The study by assumes that there are no unobserved differences between patients who are catheterized and patients who are not catheterized and finds that catheterization increases patient mortality. We instead allow for such differences between patients by implementing both the instrumental variable bounds of , which only exploits an instrumental variable, and the bounds of , which exploit mild nonparametric, structural assumptions in addition to an instrumental variable. We propose and justify the use of indicators of weekday admission as an instrument for catheterization in this context. We find that in our application, the bounds do not indicate whether catheterization increases or decreases mortality, where as the bounds reveal that at least for some diagnoses, Swan-Ganz catheterization reduces mortality at 7 days after catheterization. We show that the bounds of remain valid under even weaker assumptions than those described in . We also extend the analysis to exploit a further nonparametric, structural assumption-that doctors catheterize individuals with systematically worse latent health-and find that this assumption further narrows these bounds and strengthens our conclusions. In our analysis, we construct confidence regions using the methodology developed in . We show in particular that the confidence regions are uniformly consistent in level over a large class of possible distributions for the observed data that include distributions where the instrument is arbitrarily 鈥渨eak鈥?