Key points (98/100 words)
- Clone-censor-weighting (CCW) can compare the effects of different
treatment initiation windows.
- The causal contrast estimated by CCW when studying initiation windows
varies depending on when the population naturally initiates, making
information on initiation timings essential context for estimates.
- We provide straightforward analytic code for a CCW analysis within a
cohort constructed from publicly available synthetic Medicare claims
data.
- We demonstrate how to construct inverse probability of censoring
weights that only allow recent initiators to “stand in” for those
censored at the end of windows to avoid selection bias when exposure
effects vary over time.