Realist assessment of fidelity during the implementation of the PARTNERS collaborative care intervention for people with diagnoses of severe mental illness within a cluster randomised controlled trial.

Many with severe mental illnesses are underserved by disjointed service provision. PARTNERS aims to address this via collaborative care with recovery-based coaching. PARTNERS was evaluated in a randomised controlled trial. Understanding how intervention delivery compared to the model, why this was, and under what circumstances, aids interpretation of trial results and optimisation of future implementation. This paper reports the results of a Realist assessment of fidelity, exploring delivery compared to model and refining programme theory. Practitioners, service users, supervisors, primary care representatives, and researchers (n = 39) were interviewed. Additional data included session recordings, follow up interviews, practitioner reflective logs, supervision logs, contact data, service user surveys, and meeting minutes. A framework analysis with evaluative coding was used to assess the extent to which delivery matched the Realist initial programme theory, and how, why and under what circumstances this was the case. Retroductive analysis was used to refine the programme theory. Delivery was good, but varied by practitioner and over time. Delivery improved over time, as practitioner understanding of the intervention increased. Refinements to the programme theory include training leading to practitioners forming collaborative relationships with service users most of the time, but unidentified contextual factors causing variation in consistency. Whether training led to practitioners liaising across different bodies was dependant on the contextual factors of existing relationship skills and previous connections. System-level difficulties in providing consistent supervision made it difficult to assess the impact of this mechanism on delivery. Variation in delivering means caution should be applied when interpreting trial results. Implementation of practitioner-level change without implementing system-level change limits the ability to fully implement the model and to draw conclusions as to effectiveness. Current changes to NHS community mental health care may make this more achievable. Further research is needed to understand the role of supervision and optimal training. Trial registration: This is the realist process evaluation of the cluster randomised controlled trial ISRCTN95702682. REC approval: West Midlands-Edgbaston Research Committee 29/06/2017, ref: 14/WM/0052 (trial registration number ISRCTN95702682).