How to perform an interrupted time series analysis with multiple treatment groups?

[Stephen Butler]

Hi all,

I'm do a controlled interrupted time series analysis on Stata 15.1.

I have five years of panel data (2014-2018), and my subjects are health clinics. I have c.500 clinics in my treatment group, and c.7000 clinics in my control group. The intervention happened in 2017.

I've tried using the 'ista' command; but that only allows one treatment group (i.e. one clinic), like so:

itsa Dependent_var, treat(1) trperiod(2017) lag(1) posttr figure

I want to have all 500 treatment clinics in the treatment group. Is this possible?

Many thanks.

[Justin Niakamal]

itsa (from SSC) only allows for one treatment group. It sounds like you're better off using a difference in differences setup.

[Stephen Butler]

Thanks Jason, yes I can do a DiD setup - I was looking into time series designs out of curiosity. The pre-intervention trends between treatment and controls are parallel, so a DiD co-efficient should be reliable. Best, Stephen