Hester van Eeren

| Chapter 5 5 | 104 Balance assessment Before the PS application, balance was assessed in all measured baseline characteristics. Table 1 represents the standardized biases. The largest imbalances were found for internalizing problems reported on the YSR (mean T-score: 54.79 for FFT and 50.78 for MST), total behavioral problems measured with the YSR (mean T-score: 57.35 for FFT and 53.59 for MST), gender (53.6% male for FFT and 67.2% male for MST), previous court order (26% had a court order for FFT and 59.4% had a court order for MST), and having police contact before treatment (33.1% had police contact for FFT and 49.2% had police contact for MST) (Table 1). The standardized bias of these baseline variables was higher than the accepted .25 (Table 1). After weighting, balance for all of the covariates was found when the PS model contained all covariates except for the total score of behavioral problems measured by the CBCL (Table 1). Balance was inspected in the sample with overlapping PS scores. As a result, 8 MST and 12 FFT cases were removed from the resulting sample. As Table 1 shows, values for the standardized bias after PS application are all lower than .25. The values of the standardized bias for the missing indicator variables were also lower than .25 (Table III in Supplemental Material shows standardized bias for missing indicators before and after applying the PS). Table 2 shows the variance ratio and five-number summaries of the continuous variables as additional measures for inspecting balance. In the weighted sample, the 2.5 th and 97.5 th percentiles are .78 and 1.22. The estimated variance ratios are within these boundaries, and thus equality between treatment groups using this measure can be assumed. Moreover, the five-number summaries are also roughly equal in the PS weighted sample (Table 2). In the complete case sample of 361 adolescents and their families, balance was achieved when the variables of age, internalizing and externalizing problems measured with the CBCL, parenting stress, gender, country of birth, previous treatment, engagement in school or work, court order, police contacts, and employment status of the primary caregiver were included in the PS estimation. Because balance was inspected in the sample with overlapping PS scores, 49 MST and 3 FFT cases were removed from this sample. Analysis of treatment effect After assessing the balance, the effectiveness of MST and FFT was compared in the outcome model. Table 3 shows no difference in externalizing problem behavior (CBCL: 0.14; 95% CI -3.23 – 3.49, YSR: -0.29; 95% CI -2.45 – 1.90), with a small effect size of d = 0.01 and d = 0.03, respectively. The risk ratios (RR) and risk differences (RD) of the secondary outcomes showed no differences between MST and FFT for the proportion of youth living at home and having had police contact (Table 3). However, a significantly higher proportion of adolescents who had completed MST were engaged in school or work after treatment compared with FFT (RR 1.27; 95% CI 1.06 – 1.57, RD 19.2%; 95% CI 5.2% - 32.9%) (Table 3).