Barriers and facilitators to the implementation of a complex nursing intervention: a human-in-the-loop LLM-Assisted...

Implementation of complex nursing interventions remains challenging, with success varying substantially across healthcare settings. Most implementation research assesses barriers and facilitators at single timepoints, limiting understanding of how these determinants evolve. Therefore, we aimed to understand how barriers and facilitators evolve over time during different implementation phases across three categories of implementation outcomes.

Following the "Exploration, Preparation, Implementation, and Sustainment" (EPIS) framework, this longitudinal qualitative study examined barriers and facilitators during implementation of the PREDOCS nursing intervention across twelve Dutch cardiothoracic centers from 2016 to 2019. We conducted 252 structured interviews across eight measurement waves with diverse healthcare professionals. Data were systematically coded for barriers and facilitators using the Capability, Opportunity, Motivation-Behaviour (COM-B) model through a hybrid approach combining Large Language Model (LLM)-assisted analysis with rigorous human verification. Hospitals were categorized into three implementation outcome groups: Implementation (n = 4), Delayed-Implementation (n = 5), and No-Implementation (n = 3).

Barriers and facilitators evolved systematically across EPIS phases. Motivation remained universally high across all outcome groups throughout all phases. Fundamental knowledge gaps about intervention components during the Exploration phase were present in all three No-Implementation hospitals and absent in the Implementation group. Opportunity-domain factors, specifically the presence or absence of organizational infrastructure, resources, and management support, differed systematically between outcome groups across phases. As implementation progressed, the Implementation group reported a shift in the Capability domain from training barriers to available training and competent staff, a pattern absent in the Delayed and No-Implementation groups. Successful hospitals reported compensatory opportunity facilitators alongside persistent barriers, whereas hospitals that discontinued implementation reported sustained belief in intervention value in the absence of enabling organizational conditions.

Implementation success depends less on motivation-building than on developing organizational infrastructure to translate intention into practice. Implementation determinants are not static: early knowledge gaps and absence of organizational facilitators functioned as early signals of later failure, and support strategies should therefore be phase-specific and longitudinally monitored rather than assessed at a single point in time.