Abstract Introduction Despite trials consistently showing the benefits of behavioral interventions for insomnia such as Cognitive Behavioral Therapy for Insomnia (CBT-I) and mindfulness based interventions, attempts to scale these efforts up to broad implementation have not always resulted in the same levels of effect. It is possible that study design factors that are often underreported — such as provider number and characteristics — may result in decreased generalizability of trial results. Methods We searched PubMed and PsycInfo for RCTs of behavioral sleep interventions delivered to adults or adolescents with insomnia by a human provider. We used the Covidence platform for review of the abstracts and full text, requiring consensus by two team members at each stage. Quantitative meta-analysis was conducted using StataSE to explore effect size and heterogeneity, with meta-regression to test whether low or unstated provider number was associated with an increased effect size. Results Of the 142 included studies, 75 reported ISI scores at baseline and follow-up and are analyzed here; only 49 reported the number of providers (including 21 with a single provider), and only 5 reported whether effect sizes differed across providers. After meta-analysis, the resulting weighted Hedge’s g across all included studies of 1.67 (95%CI 1.11 to 2.22) represents a large effect size. However, visualization demonstrated potential publication bias and substantial heterogeneity of treatment effects (HTE). In the meta-regression, even after adjusting for known drivers of HTE (year of study, time to follow-up, baseline severity), the number of intervention providers remains a significant driver, with studies having 4+ providers showing significantly worse effect sizes than studies with fewer providers (adjusted difference in Hedge’s g of -3.10, 95%CI -6.02 to -0.17). Conclusion The most readily evaluated mechanisms of our observed results include publication bias and challenges with intervention fidelity. More challenging to evaluate is the possibility that the effect sizes of some intervention trials (especially those with 1-3 providers) may rest at least in part in factors outside the intervention itself, such as in the personal characteristics of the providers. Support (if any)
Abudu et al. (Fri,) studied this question.