Developing a program theory for implementing a Danish social prescribing model: an ethnographic study.

Introduction: Social prescribing (SP) is increasingly promoted internationally as a cross-sector strategy to address complex health and social needs by linking primary care patients with community-based, non-clinical support. Despite increasing policy uptake and expanding evaluations of outcomes, evidence on how SP models are implemented and become embedded in routine practice across diverse welfare and organizational contexts remains limited and fragmented. Existing studies tend to catalogue implementation activities and contextual barriers without sufficiently explaining how implementation strategies operate through actors, activate underlying mechanisms, and interact with contextual conditions. We aimed to address this gap by developing an empirically grounded, mechanism-based program theory for implementing a Danish SP model. Specifically, we examined which implementation strategies were distributed across key actor groups, and how strategies performed by specific actors were assumed to activate mechanisms under specific contextual conditions to generate implementation outcomes. Methods: We conducted an ethnographic field study of the early implementation phase (from mid-2024 to mid-2025) of a Danish SP model (Southwest CPH) initiated by general practitioners (GPs) in Copenhagen in collaboration with link workers (LWs), a Volunteer Centre, and the DaneAge Association. Data comprised 68 days of participant observations supplemented by documents (e-mails, training materials, and reports). Analysis followed an abductive five-step approach: (1) mapping implementation strategies using the ERIC taxonomy, (2) analysing the distribution of strategies across actors, (3) examining systemic interaction patterns across clinical, community, and governance settings, (4) identifying actor logics and mechanisms informed by cultural models, and (5) synthesizing findings into a mechanism-based program theory drawing on realist evaluation, systems theory, and Proctor's implementation outcomes. Findings were iteratively validated through two rounds of analytic dialogue with all project members. Results: Implementation strategies spanned six ERIC clusters and were distributed across four actor groups in a differentiated yet complementary pattern. GPs primarily led dissemination, policy-related activities, and infrastructural anchoring; LWs concentrated on training, relationship-building, and iterative learning; the Volunteer Centre facilitated local networking and community infrastructure; and the DaneAge Association contributed to national-level advocacy and legitimacy work. Synthesis identified five cross-cutting mechanism families through which strategies were assumed to contribute to implementation outcomes under specific contextual conditions: (1) legitimacy, (2) standardization and normalization, (3) relational sense-making and trust-building, (4) network weaving and coalition formation, and (5) knowledge visibility and mobilization. The resulting program theory specifies how combinations of strategies enacted across actors and system levels are expected to generate key implementation outcomes, including adoption, feasibility, fidelity, acceptability, penetration, and sustainability. Conclusions: Implementation of SP should be understood as a system-embedded, multi-actor process in which strategies generate effects only through actor-specific enactment and the activation of mechanisms under particular contextual conditions, rather than as isolated or transferable techniques. The empirically grounded program theory developed in this study provides an analytically transferable framework for examining and guiding adaptive SP implementation and scale-up in welfare systems characterized by strong professional autonomy and decentralized governance.