Protocol for the Pathways Study: a realist evaluation of staff social ties and communication in the delivery of neonatal care in Kenya

**Background:** Despite significant reductions in mortality among children under 5 years in Kenya, the neonatal mortality rate has remained relatively unchanged. Efforts to improve neonatal care quality include new technologies, routine data collection and feedback systems, and workforce training. However, the 'software' aspects of health systems—defined as 'the ideas and interests, values and norms, and affinities and power that guide actions and underpin the relationships among system actors and elements'—have been neglected in research. These relational factors profoundly influence knowledge acquisition, behaviours, and norms at individual and group levels, and are determinants of quality care. The Pathways Study aims to explore how, why, for whom, and in what circumstances features of health systems 'software' between health workers caring for neonates in Kenyan hospitals influence quality of care targeted by improvement efforts. A preparatory realist synthesis of the literature produced an initial programme theory comprising 35 context-mechanism-outcome configurations (CMOCs) organised under four domains: Social group, Hierarchy, Bridging distance, and Discourse. The Pathways Study will refine this theory for the specific setting of Kenyan neonatal units.

**Methods:** The study is a realist evaluation employing a mixed-methods approach across two phases. Phase 1 involves data collection from two large urban public hospitals in Kenya—Hospital 1 (a large teaching and referral hospital with a neonatal intensive care unit, implementing NEST360) and Hospital 2 (a large county hospital with a smaller neonatal unit, member of the Clinical Information Network). Data collection methods include: (1) non-participant observation of 6 shifts per hospital (6–8 hours each) plus hospital meetings (total 42–56 hours per site); (2) a social network questionnaire administered to all health workers involved in neonatal care (40–120+ participants per site); (3) in-depth semistructured interviews with approximately 35 participants per hospital (total ~70); (4) key informant interviews with management representatives from each hospital and associated training institutions (5–7 persons); and (5) focus group discussions with three groups (5–9 members each): basic nursing graduates, specialist neonatal/paediatric nursing graduates, and nurse educators. Additional qualitative data will be collected from a purposively sampled nursing graduate group and nurse educators. Phase 2 involves a stakeholder co-design workshop (8–12 participants) to discuss and refine phase 1 findings and develop recommendations for practice. Inclusion criteria require participants to be aged 18 years or over, work in delivery rooms, postnatal wards, neonatal units, or paediatric wards, and be able to provide informed consent. Data collection tools were informed by the prior realist synthesis and will be piloted before use. Researchers will adopt a reflexive approach throughout.

**Key Results:** This is a study protocol; no results are reported. The planned analyses include: thematic analysis of qualitative data using NVivo software; grounded theory approach (open-coding → axial-coding → selective-coding) for sociocognitive skills data; social network analysis using Gephi and/or R to calculate metrics such as density and centrality and develop sociograms; and realist evaluation using retroduction to identify demi-regularities and develop CMOCs. The initial programme theory from the literature will be tested, confirmed, refuted, or refined against primary data. A refined middle-range programme theory and recommendations for practice will be co-developed as main outputs.

**Clinical Implications:** The findings will inform the design and targeting of interventions to improve quality of neonatal care in Kenyan hospitals by addressing workforce social ties and communication. The study will contribute to understanding how sociocognitive skills (e.g., communication, teamwork, leadership, situational awareness) influence the use of clinical competencies by nurses. Recommendations are likely to include how and when these skills can be incorporated into basic and specialised nursing programmes and in-service continuous professional development. Findings may also be applicable to wider healthcare training programmes and will contribute to the international literature on social networks of the healthcare workforce, which is almost entirely derived from high-income settings. The study has received ethical approval from Kenya Medical Research Institute (KEMRI/SERU/CGMR-C/241/4374) and Oxford Tropical Research Ethics Committee (OxTREC 519-22).