Can implementation failure or intervention failure explain the result of the 3D multimorbidity trial in general practice: mixed methods process evaluation

Objectives During a cluster randomised trial, (the 3D study) of an intervention enacting recommended care for people with multimorbidity, including continuity of care and comprehensive biennial reviews, we examined implementation fidelity to interpret the trial outcome and inform future implementation decisions. Design Mixed-methods process evaluation using cross-trial data and a sample of practices, clinicians, administrators and patients. Interviews, focus groups and review observations were analysed thematically and integrated with quantitative data about implementation. Analysis was blind to trial outcomes and examined context, intervention adoption, reach and maintenance, and delivery of reviews to patients. Setting Thirty-three UK general practices in three areas. Participants The trial included 1546 people with multimorbidity. 11 general practitioners, 14 nurses, 7 administrators and 38 patients from 9 of 16 intervention practices were sampled for an interview. Results Staff loss, practice size and different administrative strategies influenced implementation fidelity. Practices with whole administrative team involvement and good alignment between the intervention and usual care generally implemented better. Fewer reviews than intended were delivered (49% of patients receiving both intended reviews, 30% partially reviewed). In completed reviews >90% of intended components were delivered, but review observations and interviews with patients and clinicians found variation in style of component delivery, from 'tick-box' to patient-centred approaches. Implementation barriers included inadequate skills training to implement patient-centred care planning, but patients reported increased patient-centredness due to comprehensive reviews, extra time and being asked about their health concerns. Conclusions Implementation failure contributed to lack of impact of the 3D intervention on the trial primary outcome (quality of life), but so did intervention failure since modifiable elements of intervention design were partially responsible. When a decisive distinction between implementation failure and intervention failure cannot be made, identifying potentially modifiable reasons for suboptimal implementation is important to enhance potential for impact and effectiveness of a redesigned intervention.