Figure 2: A figure describing the first four steps of
a clone-censor-weighting analysis in the context of comparing effects of
different treatment initiation windows. White diamonds represent index
events and black squares represent prescription records. After
identifying eligible individuals (in this case, by excluding the
individual with treatment prior to their index event), duplicates, or
clones, are created within each treatment regimen of interest (i.e.,
start by day 30, start by day 90, and start between 30-90 days). Those
clones are then censored when they deviate from their assigned regimen
either by failing to initiate by the end of the window (represented by
solid gray circles) or, for the start between 30- and 90-day regimen, by
initiating before day 30 (represented by the checkered gray circle).
Weighting (represented by thicker lines) is then used to upweight those
who were not censored to “stand-in” for those who were. Note that in
regimen B person 4, who initiated close to day 90, is upweighted to
stand in for the person censored for not initiating by day 90, while
person 3 receives a weight of 1 on day 90 and later.