201 Summary and General discussion 7 and (2) continue to develop approaches to optimize effectiveness for individual forensic psychiatric patients. Understanding effectiveness of social network interventions RCTs are beyond dispute the “gold standard” in determining intervention effectiveness. However, we do agree with previous researchers that debated the use of RCTs to address effectiveness of complex interventions, such as (informal) social network interventions (Mead et al., 2010; Priebe et al., 2020; Wolff, 2000). RCTs examining complex interventions are likely to face pragmatic and methodological challenges (Sanson-Fisher, Bonevski, Green, & D’Este, 2007; Siette et al., 2017). Looking into the pragmatic challenges, for example, we have learned that performing an RCT among forensic psychiatric outpatients was time-consuming. Our inclusion period lasted about 3 years and a lot of effort and time was invested in the data collection procedure (i.e., about 4.5 years). Moreover, despite these efforts, trials like ours still are likely to encounter compliance problems (i.e., low initial participation rates and low overall adherence to the intervention) and study dropout, compromising sample sizes and effect estimates (Ali et al., 2021; Harris et al., 1999). Furthermore, we want to emphasize the ethical concerns that might arise due to the randomization procedure, as it is undesirable towithhold a potential beneficial intervention from vulnerable patient populations (Sanson-Fisher et al., 2007; Wolff, 2000). Regarding the methodological challenges, besides relatively small sample sizes that are not unlikely in RCTs among psychiatric populations, we acknowledge the difficulty of implementing standardized intervention protocols as well as the fact that forensic patient populations are often heterogenous. Thus, after conducting an RCT to examine a social network intervention among forensic patients, we might not sufficiently be able to determine what works for which individual patient and when (Purgato et al., 2021; Sanson-Fisher et al., 2007; Wolff, 2000). Consistently, it was argued that RCTs insufficiently provide information on an individual level and are not ideal for evaluating effects of complex social interventions (Purgato et al., 2021; Sanson-Fisher et al., 2007) – “…the RCT model is unlikely to yield valid, reliable and generalizable inferences without becoming more complex in design and more sensitive to issues of selection bias, unmeasured variables and endogeneity.” (Wolff, 2000). Therefore, RCTs should be conducted under the condition of sufficient power (e.g., obtaining larger samples through multi-center trials and evaluating multiple follow-up timepoints) as well as data on patient characteristics (e.g., age, diagnosis, and socio-environmental characteristics) in order to investigate moderator factors that are likely to influence the treatment effects (Sanson-Fisher et al., 2007). Given the limitations above, we support the view that alternative experimental designs should be explored, such as randomized encouragement trials and matched case-control studies (West et al.,
RkJQdWJsaXNoZXIy MjY0ODMw