Difference-in-difference with categorical outcomes

Emily_Chi — Mon, 17 Feb 2025 17:06:50 GMT

I am working on a study with binary outcomes by using difference-in-difference. Traditionally, we know that outcomes usually are continuous in DID model. However, there is a paper discussed using categorical outcomes in DID analysis. Therefore, I am wondering if it is able to procedure in SAS.

(Graves JA, Fry C, McWilliams JM, Hatfield LA. Difference-in-differences for categorical outcomes. Health Serv Res. 2022 Jun;57(3):681-692.)

Also, as my understanding, they used transition effect in DID model. I am unsure is it estimate the difference of probabilities among two groups. Below is the example interpretation, I appreciate more interpretation and knowledge provided.

Example Interpretation:

Transition Control Group (R0R_0R0) Treatment Group (R1R_1R1) Difference (RDD)

Stayed No (P_{00})	70%	60%	-10pp
No → Yes (P_{01})	20%	35%	+15pp
Yes → No (P_{10})	5%	3%	-2pp
Stayed Yes (P_{11})	5%	7%	+2pp

A 15pp increase in P01P_{01}P01 means the intervention led to more patients transitioning to intervention.

Re: Difference-in-difference with categorical outcomes

StatDave — Mon, 17 Feb 2025 17:26:51 GMT

See this note that shows ways the difference in difference (DID) analysis can be done with binary response data.

topic Difference-in-difference with categorical outcomes in Statistical Procedures

Difference-in-difference with categorical outcomes

Re: Difference-in-difference with categorical outcomes