Hester van Eeren

Comparative effectiveness of MST and FFT | 5 109 | Repeating the analyses within the complete case sample of 210 adolescents without a court order yielded the same results. Again, MST showed lower scores than FFT on externalizing problems (CBCL: -4.55; 95% CI -8.30 – -.41, YSR: -3.21; 95% CI -6.03 – -.14), with a medium effect size of d = 0.43 and d = 0.33, respectively. As before, no differences were found for the secondary outcomes within this sample. To overcome possible misspecifications in the outcome model for this subgroup in the complete case sample, all covariates were selected except for the level of education of the adolescent, level of education of the primary caregiver, previous treatment, engagement in school or work, relation with mother, internalizing behavioral problems measured with the CBCL and YSR, and total problem behavior measured with the CBCL in a parsimonious model on the outcome. Again, a significant difference between MST and FFT on externalizing problem behavior was found for the CBCL (-3.21; 95% CI -5.89 – -.76), while no difference was found for the YSR (-1.55; 95% CI -4.26 – -.99) (Table 3). Subsample: Youth with a court order In total, 317 (246 MST; 71 FFT) of the 697 adolescents in the study sample had a court order before starting treatment. Of the adolescents who had completedMST, 70.4%were male and 78.2% were born in the Netherlands, while for FFT, 56.1% of the adolescents were male and 91% were born in the Netherlands (for an extensive comparison of the treatment arms, see Table VII in Supplemental Material). Multisystemic Therapy and FFT showed significant differences in terms of age, externalizing behavioral problems measured with the CBCL, internalizing problems measured with the YSR, gender, relation with father, and employment status of the primary caregiver at the baseline (Table VII in Supplemental Material). Balance assessment Before the PS application, the standardized bias was higher than the accepted .25 for age, externalizing problem behavior on the CBCL, internalizing problems on the YSR, gender, relation with father, and employment status of the primary caregiver (Table VII in Supplemental Material). After PS application, balance was not achieved using different PS estimations. Either there were some variables with a standardized bias higher than .25, or there were numerous variables with a standardized bias just below.25. Furthermore, if balance was assessed in the sample with overlapping scores on the PS, roughly 60–80 MST cases had to be removed each time when testing various PS estimations. This indicates that the sample of adolescents assigned to MST could not be balanced to the sample of adolescents assigned to FFT (West et al., 2014). In the complete case sample, 151 adolescents had a court order before treatment (125 MST; 26 FFT). Again, no balance was found between the treatment groups when testing and applying different PS estimations. Analysis of treatment effect Because there was not confidence in assuming balance was achieved in this subsample of youth with a court order, the effectiveness could not be estimated without ensuring the control of allocation bias. The same holds for the complete case sample of 151 adolescents.