Quasi-experimental designs were used to test for intervention impact.
A Wald test was used to test for intervention effect within each FD and also to compare the same effects between the two FDs.
Therefore, rather than incorporate baseline outcome measurements and the abovementioned covariates to the model, the analysis focused on key hypothesis testing for intervention effect, subtype effect, and interaction between both fixed factors across the six domains.
Multilevel mixed effects regression models will be applied to test for intervention effects.
To test for intervention effects, LMMs were run using the ITT approach.
To test for intervention effects, taking into account possible baseline differences, interaction effects of groups by time were calculated.
The test for intervention effect is based on the least square means of the estimated main effect of intervention (estimate of the average of the 6- and 12-month means).
Repeated measures ANOVA, paired sample t-tests, multiple linear regression, and binary logistic regression will be used to test for intervention effects after adjusting for relevant co-variables (e.g. sex, age) and baseline scores.
The GEE analyses allowed to test for intervention effects simultaneously across assessment times considering all available longitudinal information, i.e. each observation irrespective of non-participation at a previous or later follow-up time.
To examine whether this might be related to variation in TCGPS activity the 'Gradual and permanent impact' model (Model 2) was also tested for Intervention B using the quarterly number of new TCGPS patients and total number of quarterly TCGPS consultations as intervention variables, but as with the simple intervention dummy variable, no stable parameter estimates were obtained.
