“…The sample size was calculated for the three primary outcomes: bingeâdrinking frequency, number of alcoholic drinks on a typical occasion, and alcoholârelated problems at 3âmonth followâup. To detect an anticipated effect size of d = 0.26, with a power of 80% and with a type I error of 5% for each of these outcomes, 306 patients (153 patients per group) had to be included, if randomization had occurred at patient level. With an assumed intraâcluster correlation of 0.05 and an average of 1.264 patients per cluster, we calculated a design effect for cluster randomization of 1.013, which increased the required sample size to 312 patients (156 patients per group).…”