Statistical analyses

Using the pre-randomization baseline rates, we assumed the proportion of people with >80% coverage with guideline-recommended medications was 28%. A total sample size of 1000 participants, allowing for a 20% loss to follow-up would have 90% power to detect an absolute improvement of at least 10% using two-sided tests, with p values of <0.05 judged as significant. For the original primary outcome of BP and LDL target attainment, this sample size provided 80% power to detect a 7% absolute improvement, assuming a baseline control rate of 11%. All statistical analyses were conducted blinded to group allocation.

A pre-specified statistical analysis plan that was finalized prior to database lock was followed. The analysis was done by an independent statistician using SAS (version 9.3). Primary analyses were unadjusted, following an intention-to-treat principle and conducted blind to treatment allocation. Multivariate analyses were performed to adjust for any significant differences between each study arm. Pre-specified sub-group analyses were conducted to compare outcomes based on gender, age, baseline, eHEALS, and CVD status (established CVD compared vs. high CVD risk). Mean risk factor levels were compared between groups in terms of relative risks (RR), 95% confidence intervals (CIs), and two-sided p values. Characteristics were compared between groups using independent t tests for continuous or Χ² tests for categorical variables. Mann–Whitney U tests were used where data were not normally distributed.