Report absolute risk (incidence) difference

[msuchard]

No description provided.

[schuemie]

I don't have experience in computing risk differences when you have time-to-event-data with right-censoring (i.e. for which you'd otherwise use a Cox model), as well as propensity score adjustment. Especially computing the variance seems tricky. Any suggested reading?

This paper discusses PS adjustment, but appears to assume a fixed time-at-risk without censoring.

On another note, empirical calibration hasn't been tested on risk differences yet.