Hester van Eeren

Comparative effectiveness of MST and FFT | 5 103 | and there were no missing values for the baseline variables. 1 When adding covariates to the regression analyses in the complete case sample, backward selection of the covariates on the treatment effect was used to find a parsimonious model for the outcome. For the secondary outcomes, it was not possible to also control for covariates, as the number of events needed per covariate in a logistic regression was not met (Peduzzi, Concato, Kemper, Holford, & Feinstein, 1996). Therefore, treatment effects on the secondary outcomes in the complete case sample were estimated without controlling for additional covariates. Subgroup effects Finally, within the study sample, analyses were repeated for the subsamples of youth who had a court order (246 adolescents assigned to MST; 71 FFT) and youth who did not have a court order (168 MST; 202 FFT). Within the complete case sample, analyses were also repeated for the subsamples of youth with (125 MST; 26 FFT) and without (104 MST; 106 FFT) a court order. Here, the PS within each subsample was estimated separately, as recommended by Green and Stuart (2014). Within each subsample, again the balance between the treatment arms was assessed and then the PS was applied by weighting groups by the odds of the estimated PS score (Green & Stuart, 2014). The analyses were performed with IBM SPSS for Windows, version 22 (IBM Corp, 2013) and Microsoft Excel (2013). The 95% confidence intervals were bootstrapped in Stata 12 (StataCorp, 2011). Results This section first describes the sample characteristics, then the balance in the covariates, and finally the treatment effect for respectively all adolescents in the study sample (N = 697), the subsample of adolescents without a court order (N = 370), and the subsample of adolescents with a court order (N = 317). Within each sample, the results of the complete cases analyses are also described. Study sample: All adolescents Within the study sample of 697 adolescents, 422 completed MST and 275 completed FFT. Of the adolescents who completed MST, 67.2% were male and 83.4% were born in the Netherlands. For FFT, 53.6% of the adolescents were male and 95.8% were born in the Netherlands (see Table 1 for an extensive comparison of the treatment arms). Comparing the treatment groups on baseline characteristics showed substantial differences in internalizing, externalizing, and total behavioral problems reported by adolescents. Furthermore, treatment groups differed in gender, country of birth, the adolescent’s living situation, level of education, previous treatment, engagement in school or work, previous court order, previous police contact, and country of birth and employment status of the primary caregiver (Table 1). 1  Because the complete case sample had no missing values on any of the covariates, no missing indicators were needed in estimating the PS and assessing balance of the covariates between the treatment arms