I have a dataset of 170 participants that were followed over 5 visits.

Main Exposure is a point in time (ex. January 1, 2010; before=0, after=1).

All participants have at least a baseline, a follow-up before the time point, a follow-up after the time point.

Outcome is health service use (HSU). (yes/no)

I have insurance status as a covariate (yes/no).

I want to know whether the time point (proxy a policy implementation) made a difference in health service use.

I want to know if insurance after the timepoint was the reason that HSU changed (ex: completely wiped out the effect/mediated).

I tried a total effect model and an outcomes model to see if there was a big difference in estimates after adjusting for insurance.

GEE total effect adjusted model: HSU = B0 +B1(TIMEPOINT) +Error

GEE outcomes adjusted model: HSU = B0 +B1(TIMEPOINT) +B2 (INSUR)+Error

PostTimePoint OR: 1.94 (1.41-2.66)

PostTimePoint aOR GEE Total effect adjusted model: 2.06 (1.46-2.91)

PostTimePoint aOR GEE Outcomes adjusted effect model: 1.92 (1.34-2.77)

I am having a hard time understanding why the effect didn't wash away and how to interpret the process. Any help would be great. Thanks!