This is an author produced version of Patterns of therapist variability: Therapist effects and the contribution of patient severity and risk.. White Rose Research Online URL for this paper: http://eprints.whiterose.ac.uk/111248/

Objectives: To investigate the size of therapist effects using multilevel modeling (MLM), to compare the outcomes of therapists identified as above and below average, and to consider how key variables, in particular patient severity and risk and therapist caseload, contribute to therapist variability and outcomes. Method: We used a large practice-based data set comprising patients referred to the UK’s National Health Service primary care counselling and psychological therapy services between 2000 and 2008. Patients were included if they had received ≥ 2 sessions of one-to-one therapy (including an assessment), had a planned ending to treatment and completed the Clinical Outcomes in Routine Evaluation-Outcome Measure (CORE-OM) at preand posttreatment. The study sample comprised 119 therapists and 10,786 patients, whose mean age was 42.1 years, and 71.5% were female. MLM, including Markov chain Monte Carlo procedures, were used to derive estimates to produce therapist effects and to analyze therapist variability. Results: The model yielded a therapist effect of 6.6% for average patient severity but it ranged from 1%-10% as patient non-risk scores increased. Recovery rates for individual therapists ranged from 23.5% to 95.6% and greater patient severity and greater levels of aggregated patient risk in a therapist’s caseload were associated with poorer outcomes. Conclusions: The size of therapist effect was similar to those found elsewhere but the effect was greater for more severe patients. Differences in patient outcomes between those therapists identified as above or below average were large and greater therapist risk caseload rather than non-risk caseload was associated with poorer patient outcomes.