Thank you so much, @PerPersvensson! I also think confounding by indication is likely, but I cannot yet explain why this would be the direction of the odds ratio. Perhaps the main conclusion is: the causal model here is complex!
1 Like