Who profits from new guidelines, which method to choose?

I have a register with 500 patients treated according to the old guideline and 300 patients treated according to the new guideline. The success of treatment is measured by a continuous variable Y. Under the new guideline Y after treatment should be generally lower compared to the old guideline.

Which group of patients did achieve lower Y values and which did not?

I started with a quantile regression Y_aftertreatement ~ (Y_beforetreatement+age+gender+conf2+conf3+…+confn)*guidline

Where guidline=1 if new guideline and guidline=0 old guideline.

I thought the significant interaction terms conf2*guideline (etc.) would tell me which patient characteristics profit from the new guideline and which do not.

Is there a better approach? How do I include medication? Medication (dose) also changed with the guidelines.

I am grateful for any tip.

what time frame do the data represent? maybe time needs to be accounted for first?

1 Like

Agree time is of the essence here, as there may be secular trends — perhape even secular trends that drove the guideline change. Is the care episodic (like a surgical intervention) or chronic? In the latter case, can you observe within-patient shifts from old to new guidline? Are you able to detect which guideline was being followed, just looking at the record of care? If so, differential adoption rates across providers might give you a whiff of an instrumental variable.