Identifying matched treated and control groups after using teffects psmatch

[Tim Anderson]

Hello,

I am planning to use teffects psmatch to examine the impact of non-randomized exposure to a medication on a group of about 15,000 patients. I understand how to use the psmatch2 program to generate a post match treated and control group which can then be used to standardized mean differences in order to gauge the post-match balance of covariates.

However, I am uncertain how to calculate the same post-match standardized mean differences for model covariates using the teffects psmatch program, as the output of the teffects psmatch does not include a variable indicating matched treated and control patients.

Appreciate your advice on this topic.

Thanks,
Tim

[Tim Anderson]

Hi All,

Let me clarify this question a bit, as I have read the following Statalist thread which was helpful in partially clarifying this issue, and the tebalance summarize command does provide a way to calculate standardized mean differences pre and post match when looking at a single outcome.

https://www.statalist.org/forums/for...score-matching

However, I remain unclear on how to use this approach if looking at multiple outcomes in the same cohort. For example, I am looking at one primary outcome (mortality) and two secondary outcomes (hospital readmission and falls) accounting for 20+ covariates in my propensity score.

Ideally I would like to be able to report standardized mean differences pre and post match for each covariate in a typical Table 1. However, using the teffects psmatch and tebalance summarize commands requires that I generate a new ps match for each outcome separately, rather than use the same matched cohort throughout, and thus my "Table 1" of standardized mean differences would be different for each outcome. Am I misinterpreting how to use tebalance?

Thanks,
Tim