The original CHAT paper had a total of 226 subjects in the surgery arm (early adenotonsillectomy), however in the version 11, there are only 224 subjects in the surgery arm using variable 'had_surgery'. Is there an explanation for the different number of subjects? Is there another surgery arm variable I should use?
I suggest using arm (https://sleepdata.org/datasets/chat/variables/arm) to determine the randomized treatment arm (n=226 in surgery arm).
I'm guessing had_surgery also includes subjects who crossed over from the watchful waiting arm. Some subjects randomized to surgery did not ultimately have surgery, too.