The initial sample size calculations were based on showing a difference in the square root of the asthma quality of life score. We assumed a SD of 0.61, 200 identified asthmatic patients per group, 20 practices per group, and a mean baseline quality of life score of 2.25. We estimated that we had 80% power to detect a fall in the score to be between 1.56 (intraclass correlation coefficient 0.15) and 1.80 (0.01) in the intervention group. Based on an assumed asthma prevalence of 5.5%, and 60% survey response rates, we calculated that 12 200 patients were required to identify 400 subjects with asthma. The funders requested that the survey should be doubled to increase the power of the study. The eventual survey sample size was therefore 24 400.