Estimating variability in outcomes attributable to therapists: A naturalistic study of outcomes in managed care.

Abstract

To estimate the variability in outcomes attributable to therapists in clinical practice, the authors analyzed the outcomes of 6,146 patients seen by approximately 581 therapists in the context of managed care. For this analysis, the authors used multilevel statistical procedures, in which therapists were treated as a random factor. When the initial level of severity was taken into account, about 5% of the variation in outcomes was due to therapists. Patient age, gender, and diagnosis as well as therapist age, gender, experience, and professional degree accounted for little of the variability in outcomes among therapists. Whether or not patients were receiving psychotropic medication concurrently with psychotherapy did affect therapist variability. However, the patients of the more effective therapists received more benefit from medication than did the patients of less effective therapists.

Therapist effects and IAPT Psychological Wellbeing Practitioners (PWPs): A multilevel modelling and mixed methods analysis

“…The rate of 9% is comparable to previous studies of therapist effects during routine delivery of the traditional 'low volume, high contact' psychotherapies, where the percentage of variance attributable to therapists ranges from 5-8% (Lutz et al, 2007;Wampold & Brown, 2005). Even the lower bound of the 95% probability interval (5.8%) falls within this range.…”

“…the outcomes for patients seen by the same practitioner are likely to be similar in some way and different from the outcomes for patients seen by another practitioner), multi-level modelling (MLM) is advocated as an appropriate method for assessing higher level (in this case, PWP) effects (Goldstein & Speigelhalter, 1996, Raudenbush & Bryk, 2002, Wampold & Brown, 2005. Other methods include generalised estimating equations which estimate the parameters of a generalised linear model (Hardin & Hilbe, 2003).…”

“…Recent studies of higher intensity treatments in psychological therapy services have also used MLM to estimate therapist effects (e.g. Wampold & Brown, 2005, Saxon & Barkham, 2012.…”

“…Other methods include generalised estimating equations which estimate the parameters of a generalised linear model (Hardin & Hilbe, 2003). MLM allows for the partitioning of the total outcome variance between level 1 (patient level) and level 2 (PWP level) with the proportion of total variance at level 2 equating to the therapist effect (Raudenbush & Bryk, 2002;Wampold & Brown, 2005). MLM has been extensively used in other areas of research, most notably in the study of comparative effectiveness of schools, where pupils are nested within teachers or classes, which are in turn nested within schools (Goldstein & Speigelhalter, 1996).…”

“…Pre-treatment scores and interactions between the measures were controlled for by the inclusion of both PHQ-9 and GAD-7 pre-treatment scores, centred around their grand means in both models (Hoffmann & Gavin, 1998;Wampold & Brown, 2005). Variables in the model were considered statistically significant when their coefficients were more than 1.96 times their standard errors.…”