Stata se statistical software release 17

The sample size of 19 RACFs was deemed to have sufficient statistical power. To estimate the statistical power of the helix, counterbalanced, three-strategy, and three-context area (medication type) trial design, we performed 1000 Monte Carlo simulations using Stata/SE Statistical Software: Release 17.0 (StataCorp. 2021). The sample size analysis was conducted at the RACF level based on the primary outcome (i.e. proportion of RACF-level Guideline concordant use of antipsychotics, benzodiazepines and antidepressants). The sample size was based on conducting analysis using a linear mixed model employing fixed, categorical factors for medication type and guideline implementation strategy and RACF as a random intercept. We assumed a minimum important effect size of an absolute change in the proportion of concordant medication category takers of 5% for the level 2 strategy relative to level 1 strategy and of 15% for the level 3 strategy relative to level 1 strategy. Given these assumptions, our sample size of 19 RACFs provides > 99% power to detect the absolute change in proportion of concordant medication category takers of 15% and 81% power to detect the absolute change in proportion of 5%.

All primary analyses will be conducted using the intention-to-treat (ITT) principle. Level 3 intervention will be compared to level 1 and level 2 at the 6-month time point as a primary outcome (Table 4).
Per protocol analysis will also be undertaken with the per protocol set including all RACFs without a major protocol deviation. Major protocol deviations may include, but are not limited to, the following:

Loss of knowledge broker

Incomplete delivery of intervention (e.g. local action plans, QUM service)

We anticipate that the primary analysis will be conducted using a linear mixed model treating the RACF as a random effect, with the strategy level and medication contexts as fixed effects. Multiple imputation will be used to impute missing covariate values where necessary. Data will be analysed using Stata/SE Statistical Software: Release 17.0 (StataCorp. 2021).