Looking at longer term effects, all methods showed a trend with the treatment effect becoming more negative through time, with truncated IPW estimating the largest difference in ppFEV 1 between those taking and not taking treatment of −8.81% (95% CI, −10.50 to −7.12, P < .001) and HA‐MSM the smallest effect of −1.52% (95% CI, −3.30 to 0.27, P = .097).
← all excerpts
Estimating long-term treatment effects in observational data: A comparison of the performance of different methods under real-world uncertainty.
1
—
—