Effect of psychotherapy on Intolerance of Uncertainty: A meta-analysis and review (META-ANALYSIS DATA)
Effect sizes, hedges g, gathered from treatment outcome articles on psychological treatments. Published between 1994 and 2022 march. Indicate an overall effect of treatment on Intolerance of uncertainty. The following inclusion criteria was used for selection: (a) Randomized controlled trial on psychotherapeutic treatments for adult patients, reporting outcome measures on Intolerance of Uncertainty either as a primary or secondary measure. (b) Treatments are described in such detail that they are reproducible. (c) Treatment is delivered without augmentations such as psychopharmacological medication or other major therapeutic interventions, unless augmentations are equal across groups within the study. (d) Published in peer-reviewed journal. (e) Group and single patient delivery modes as well as online/digital delivery which are not dependent on sophisticated aps/games/programs beyond video/text communication. Meaning video conference-delivered therapy and online self-study programs that might have been delivered in analog form are included, but treatments based on differentially responding software or other automated functionality were excluded. Quality assessment and data extraction The quality of the eligible studies was assessed independently by the first and second author using the 13-item Joanna Briggs Institute Checklist for Randomized controlled trials (Joanna Briggs Institute, 2017). Each item was classified as either Yes (critieria fulfilled), No (criteria not fulfilled) or Partial (criteria partially fulfilled or unclear description). Any disagreements were discussed and if not resolved successfully decided by the third and fourth author. Coding of the study elements and outcomes was performed by the main author. The elements coded were participant characteristics (number total, males/females, mean age, clinical or non-clinical, primary diagnosis), intervention characteristics (length/number of sessions, face-to-face/online/phone/group/single patient) as well as type of control condition (wait-list, delayed treatment, non-specific therapy, treatment as usual). When there were multiple active groups with independent individual participants, data was extracted for all of them. In the case of Mathur et al (2021), Ladouceur et al., (2000) & Tomei et al., (2018), methodology described in Thalheimer & Cook was used to calculate an effect size.
Steps to reproduce
Statistical analyses were performed in R-studio (v. 4.1.1) using the METAFOR (Viechtbauer, 2010) and CLUBSANDWICH (Pustejowsky & Tipton, 2022) -packages to allow standard error adjustments as some effect sizes shared a common control group and thus a degree of dependence. Rho was set to .8 and sensitivity analysis was performed. Due to the wide inclusion criteria and that large heterogeneity was suspected, Random effects modeling was used. For all comparisons post-treatment with control, Hedges’ g was calculated. Hedges’ g was chosen as an estimate of effect size due to many of the studies small sample size and 95% CI. Variance was calculated using the formula suggested by Borenstein et al. (2009). Intention-to-treat outcome data was used when available. An effect size ≥ 0.80 was considered large, an effect size ≥ 0.50 was considered medium, and an effect size ≥ 0.20 was considered small (Cohen, 1988). Pooled Hedges’ G was recalculated into number needed to treat (NNT) (Kraemer & Kupfer, 2006) for ease of interpretation. Passive control conditions have been shown to generate larger effects than active (Michopoulos et al., 2021; Mohr et al., 2014). It was therefore decided that studies with active and passive controls would be analyzed separately in the current study. Two sub-analyses were thus performed. In the case of studies reporting data on multiple treatments with independent participants, these were included as separate, independent treatments in the analysis.