ABSTRACT
AIMS: To describe the implementation process and fidelity of two versions of a guideline-based, multicomponent intervention to reduce physical restraints in nursing homes and to identify factors that might explain the heterogeneity of effects between different clusters. DESIGN: Mixed methods evaluation of the implementation process (dose delivered, dose received, response, and adaption) alongside a pragmatic three-arm cluster randomized controlled trial. METHODS: Quantitative and qualitative process data were collected during the study period (February 2015-February 2017). Quantitative data from questionnaires and short surveys were analysed by descriptive statistics. Qualitative data from focus groups and semi-structured interviews were analysed using content analysis. An in-depth analysis was conducted by contrasting responding and non-responding clusters regarding the intervention goal and primary outcome. RESULTS: Both interventions were implemented as planned in all clusters: we found no deviations from the protocol regarding the dose delivered to and received by the clusters. Satisfaction of staff targeted by the interventions was high. The in-depth analysis did not reveal any pronounced variation in the degree of implementation or adoption in clusters with a good or nearly no response to the interventions or factors explaining different study effects. CONCLUSION: Although both versions of a guideline-based multicomponent intervention to prevent physical restraints in nursing homes were implemented as planned and the response was generally acceptable, the interventions' goal to change nursing practice towards a least-restraint policy was not achieved by the entire nursing staff in all of the clusters. No factors could be identified that might explain the different effects of the interventions. IMPACT: For some nursing homes, different approaches than addressing nurses' attitudes and institutional policies might be needed to sustainably reduce the use of physical restraints; however, the process evaluation did not reveal characteristics that might have hampered or facilitated the effectiveness of the intervention.