Hester van Eeren

General discussion | 6 141 | the analysis. Second, when measuring CAFYs, the type of criminal activity that is avoided, the seriousness of the crime, and the number of times it is committed should be defined. Even more, the outcome measure should be clinically relevant and should allow making comparisons with other interventions. A preference weighted measure, like the QALY may be preferred as it adds weights to different outcomes of the interventions. Third, we should be able to assign a value to the defined outcome, the so-called WTP value, in order to judge the cost-effectiveness results against a threshold value. This WTP value depends on the outcome defined and the cost categories included in the analysis, which should be reflected in this WTP value and vice versa. In conclusion, when evaluating the cost-effectiveness of youth care interventions, commonly applied economic evaluation methods are feasible and their results can be interpreted in the same manner as in health care evaluation studies. Use of observational data in treatment evaluation Evaluating the effectiveness of interventions in youth care becomes increasingly important, but due to practical and ethical constraints it is not always possible to randomly allocate adolescents and their families to treatment. As an alternative, research could follow clinical practice in gathering data. In that case, the propensity score (PS) method can be used to control for initial differences between treatment groups. It is thereby of interest to study subgroup effects when using the PS to tailor youth care to adolescents’ situations and needs. One can adjust for subgroup effects in several ways, for example by additionally adjusting for the subgroup or by splitting the dataset into the relevant subgroups. The first manner was studied in a Monte Carlo simulation study in Chapter 4. The research question was whether the subgroups should be added to the outcome model, together with an interaction term between treatment and subgroups, or whether the PS should be made multiple to estimate the specific treatment effects within subgroups. Both methods were found to be feasible, while the latter option (i.e., making the PS multiple on the subgroup and treatment options) gave less biased results compared to the first option (i.e., adding the subgroups to the outcome model). In Chapter 5, two youth care interventions, FFT and Multisystemic Therapy (MST) were compared on their effectiveness using Routine Outcome Monitoring (ROM) data and subgroups were investigated by splitting the dataset. The outcomes were externalizing problems of the adolescent, whether the adolescent was living at home after treatment, was engaged in school or work after treatment, and had had police contacts during treatment. Minor differences were found between the interventions. However, when splitting the dataset into subgroups of adolescents who had a court order before treatment and those who had not, different results were obtained: MST was more effective than FFT in reducing externalizing problems when adolescents had no court order. Because many more adolescents with a court order were assigned to MST than to FFT, the PS could not balance the intervention groups in this subsample. Therefore, no comparative treatment effect could be estimated in this subsample. In general, both Chapter 4 and Chapter 5 showed the applicability of the PS when evaluating youth care, and more importantly, the use of clinical practice data