Demystifying Clone-Censor-Weighting to Studying Treatment Initiation
Windows: An Example Using Publicly Available Synthetic Medicare Claims
Data
Abstract
Background: Clone-censor-weighting (CCW) can compare treatment
regimens that are initially indistinguishable (such as starting
treatment within specific time windows) without using landmarks or
creating immortal time. The causal contrasts estimated in these cases
and the analyses themselves can become quite complex, however.
Objective: Provide a tutorial on CCW for comparing initiation
windows and illustrate the causal contrasts underlying such comparisons.
Methods: We identified patients with myocardial infarctions
without past aspirin or clopidogrel use in Medicare’s synthetic public
data files. We assigned “clones” to three regimens: 1) initiation
within 30 days; 2) initiation within 90 days; or 3) initiation from
30-90 days. Clones were censored when deviating from their assigned
regimen by failing to initiate treatment in time or by initiating
treatment too early. We addressed informative censoring using inverse
probability of censoring weights (IPCW), calculated weighted 180-day
risks of re-hospitalization or death using Kaplan-Meier methods, and
visualized the portion of the population exposed during the first 90
days to compare exposure distributions underlying regimens.
Results: We identified 1,589 patients experiencing myocardial
infarction with no past medication use. 15% initiated within 30 days
and 26% initiated between 30 and 90 days. After IPCW, the 180-day
outcome risk was 40.2% in the 30-day regimen, 35.7% in the 90-day
regimen, and 35.2% in the 30-to-90 day regimen. Conclusions:
Though CCW can be complex to implement and the effects it estimates can
vary substantially across study populations that initiate treatments at
different times, it is a useful tool for contrasting initiation windows.