Power calculations to detect an effect size=0.35 of 1 SD between the intervention and usual care periods, assuming an intracluster correlation of 0.1 and alpha of 5% for a two-sided test for a stepped wedge cluster randomised controlled trial with nine clusters and three steps