OSU Navigation Bar

The Ohio State University University Libraries Knowledge Bank

Dropout from Randomized, Controlled Treatments for Depression

Please use this identifier to cite or link to this item: http://hdl.handle.net/1811/51696

Show full item record

Files Size Format View Description
Cooper_Hayes_Paper.pdf 501.9Kb PDF View/Open Embargoed until February 2017

Title: Dropout from Randomized, Controlled Treatments for Depression
Creators: Cooper, Andrew
Advisor: Strunk, Daniel
Issue Date: 2012-02
Abstract: PURPOSE: Premature treatment termination, also known as "dropout", is a ubiquitous problem in the field of mental health. There is an extensive literature investigating this phenomenon in clinical contexts, including large-scale epidemiological studies (e.g., Wang, 2007) and those conducted in community mental health clinics and private practice (e.g., Mueller & Pekarik, 2000). Estimates of dropout rate in these studies vary widely (26 – 66%; Bados, Balaguer, & Saldana, 2007). Dropout is understudied in the burgeoning field of randomized controlled trials (RCTs) of psychotherapy, even though it has implications for study design, funding, and data analysis (Lane, 2008). Additionally, dropout from RCTs may be easier to define than in clinical contexts, where doing so can be difficult (Reis & Brown, 1999), suggesting possible differences in observed rates. As such, psychotherapy researchers may benefit from an estimate of expected dropout rate specifically derived from RCTs, and from identification of study features that may influence this rate. Meta-analytic techniques provide a useful way of providing an estimate of dropout rate, and of investigating potential predictors of dropout. This study focuses on treatments for major depression, one of the most extensively studied mental health issues. Information about dropout rate was collected for each individual psychotherapy treatment arm (treatment-level) and across the study (study-level) in a collection of RCTs for major depression. Study and treatment characteristics were identified as potential predictors of dropout on the basis of the existing dropout literature, and for theoretical reasons; these included therapy type, treatment duration (both intended and observed), therapist experience, mean age and diversity of the sample, and observed effect of treatment on depressive symptoms. RESEARCH METHODS: Studies were identified using a publicly available database for RCTs of major depression (psychotherapyrcts.org; see Cuijpers, van Straten, Warmerdam, and Andersson, 2008, for details on database creation). Inclusion criteria were: a) individual therapy, b) outpatient setting, c) formal diagnosis of major depression or post partum depression, d) published in English, and e) adequate information for determining dropout rate for psychotherapy condition. Forty-six studies were included in the final dataset, representing 2802 patients. These studies included 69 separate psychotherapy treatment conditions, with an average of 1.5 conditions per study. Overall study sizes ranged from 20 to 681 (mean = 112.1, SD = 114), with individual treatments ranging from 10 to 228 subjects (mean = 40.6, SD = 39.9). The most commonly examined therapy was cognitive-behavioral (CBT). Dropout rates and all predictor variables were rated by a primary coder, with a subset of the sample (11 studies) rated by another coder for comparison purposes. Intra-class correlation coefficients for predictor variables and dropout rates were high (> .9), suggesting very good reliability. However, not all predictor variables were available for all conditions. FINDINGS: Heterogeneity analyses were conducted for both treatment and study-level dropout rates, to determine the degree of variability in dropout rates attributable to real differences (versus measurement error) and to inform analytic efforts. There was significant heterogeneity in dropout estimates, with approximately 72% of variability in the treatment-level dropout rate, and 80% in the study-level rate being attributable to meaningful differences. Heterogeneity also differed on the basis of treatment type; however, therapy types were not equally represented, ranging from 3 to 36 conditions. Due to high heterogeneity in dropout rates, a random effects model was used in all subsequent analyses, to incorporate treatment variability into pooled estimates (Schulze, 2004). The estimated dropout rate at the treatment level was 17%, with a slightly higher 19% estimate at the study level. None of the proposed study- and treatment-level predictors significantly predicted dropout rate. Intended duration predicted dropout at a trend level (p = .09), such that longer studies tended to have higher dropout rates. However, in analyses controlling for intended duration, none of the treatment characteristics predicted dropout. Moderator analyses were not conducted due to concerns about inflated risk of Type I error in the sample (Hedges & Pigott, 2003). IMPLICATIONS: The estimated dropout rate from psychotherapy conditions in individual outpatient RCTs for major depression is 17%, with study level dropout rate slightly higher at 19%. Estimates of dropout may be more or less variable based on treatment type, though this may be an artifact of unequal representation of all treatment types in this sample. None of the selected treatment characteristics predicted dropout rate significantly; however, this may have been due to the high level of heterogeneity in the sample, and incomplete information about predictors. The observed dropout rates in this RCT context appear lower than those genrally reported in strictly clinical samples, although this was not a formal comparison, and there is significant true variability in these estimates (0% - 50% in this sample). This study is restricted to a common type of psychotherapy RCT, and may not generalize to other disorders or RCT formats. However, continued efforts to expand on this research may enhance researchers’ ability to make informed judgments about likely dropout rates in all manner of psychotherapy RCTs.
Embargo: A five-year embargo was granted for this item.
Series/Report no.: 2012. Edward F. Hayes Graduate Research Forum. 26th
Keywords: psychotherapy
dropout
randomized controlled trials
major depression
meta-analysis
Description: Poster Division: Arts, Humanities and Social Sciences: 3rd Place (The Ohio State University Edward F. Hayes Graduate Research Forum)
URI: http://hdl.handle.net/1811/51696
Bookmark and Share
Attribution-NonCommercial-NoDerivs 3.0 Unported This item is licensed under a Creative Commons License:
Attribution-NonCommercial-NoDerivs 3.0 Unported