A94-09 From Clinical Trial to Learning Health System: Reflexive Stakeholder Engagement During the Breathe Oxygen Therapy Trial in East Africa

Rationale By embedding real-time feedback and iterative adaptation, effectiveness-implementation trials can function as prototypes of learning health systems (LHS), linking evidence generation with system improvement. The BREATHE trial (NCT05754034), an RCT evaluating high-flow versus standard-flow oxygen for hypoxemic adults in five East African hospitals, embedded a longitudinal process evaluation to examine and enhance implementation. As part of this, we conducted structured stakeholder feedback sessions at the midpoint of trial enrollment, which was the first formal presentation of emergent themes to frontline staff for validation and co-interpretation. Methods Approximately 50 meeting transcripts, 10 interviews, and over 100 field notes informed the initial coding framework, guided by Consolidated Framework for Implementation Research constructs. In April 2025, 10 feedback sessions (2 per site, 6-8 participants each, across Kenya, Rwanda, and Malawi) invited research staff to review and refine themes. Discussions explored accuracy, relevance, and context-specific adaptations. Transcripts and notes underwent iterative thematic analysis to refine codes. Results Stakeholders affirmed and expanded themes related to trial protocols, consent, intervention perception, clinical-research integration, and health system spillover, among others. Protocol-related concerns included challenges explaining randomization (“The concept of randomization is really tough...”) and perceived lack of equipoise (“You know that if they get [standard oxygen], it will not help...”). Consent dynamics reflected gendered hierarchies and mistrust: “It’s easy when the male voice says yes, all the others follow”; “Why do I have to sign it?” Intervention concerns included unfamiliarity with high-flow oxygen and the perception of oxygen as a marker of poor prognosis (“Most of the time people associate oxygen with end-of-life care”). Teams described workflow integration and evolving trust: “Attendings who were resistant at the beginning [later] apologiz[ed].” Research staff bolstered clinical care and identified gaps: “We check them and help the clinical team titrate.” Moral injury occurred when staff couldn’t intervene despite proximity: “You leave there so broken...it leaves a stain on your heart.” Others highlighted empowerment and learning: “We are making research smart and interesting.” Sites expressed hope for post-trial continuity: “Clinicians are asking: when will you finish the trial so we can have those machines?” Additional quotations are presented in the Table. Conclusion This reflexive stakeholder feedback process, layered onto continuous team reflections, modeled LHS principles of embedded learning and co-produced improvement. It clarified implementation dynamics while advancing adaptation, ethical reflection, and system learning. BREATHE demonstrates how trials in resource-constrained settings can catalyze institutional change and durable improvements in care delivery. This abstract is funded by: Wellcome Trust, Grant Number 222165/Z/20/Z