Previous studies have concluded that when allocation is not controlled for, the results are more likely biased by baseline differences in group characteristics or confounders (e.g., motivation to change health behaviour) [ 15].
Meta-analysis was performed using the two randomised studies plus (i) the unadjusted estimate from the non-randomised study and (ii) the reported adjusted estimate, adjusted for baseline differences in the groups.
Student's t-tests were used to determine any baseline differences in the groups for age, gender, number of previous trials of SSRI's, severity of symptoms in relation to OCD, depressive symptoms, and CGI-S.
Differences in means are adjusted for any baseline differences in outcomes between groups using analysis of covariance, with baseline score as covariate.
Differences in means are adjusted for any baseline differences in costs between groups using analysis of covariance, with baseline cost as covariate.
Despite the randomisation process there were some baseline differences in the two groups, including diagnosis of dementia and previous relationship with someone with dementia,.
This suggests that male students may have more potential to improve attention performance after training, but also that baseline differences in mixed-gender groups should be controlled for.
Moreover, the overall IIEF score improved from 17.3 to 18.6 (P=0.04 and P=0.06, after adjusting for baseline differences) in the intervention group.
Baseline differences in study behavior between groups suggested that the students that used the app, performed better in general.
Additionally, despite propensity matching, we cannot rule out residual confounding from unmeasured baseline differences in health between the groups.
