Philosophiae Doctor - PhD / Introduction: National community health worker (CHW) programmes are to an increasing extent being implemented in health systems globally, mirrored in South Africa in the ward-based outreach team (WBOT) strategy. In many countries, including South Africa, a major challenge impacting the performance and sustainability of scaled-up CHW programmes is ensuring adequate support from and supervision by the local health system. Supervisory systems, where they exist, are usually corrective and hierarchical in nature, and implementation remains poor. In the South African context, the absence of any guidance on CHW supportive supervision has led to varied practices across the country. Improved approaches to supportive supervision are considered critical for CHW programme performance. However, there is relatively little understanding of how this can be done sustainably at scale, and effective CHW supervisory models remain elusive. Research to date has mostly positioned supervision as a technical process rather than a set of relationships, with the former testing specific interventions rather than developing holistic approaches attuned to local contexts. This doctoral study was exploratory in nature, seeking to generate an in-depth and contextualised understanding of the supervision phenomenon in one specific district in the North West Province (NWP) in South Africa. Using co-production methodology in an iterative approach, the study culminated in the formulation of a supportive supervision framework with CHWs and other frontline actors. Methods: The study was based on a holistic conceptual framework of supportive supervision, which was viewed as comprising three core functions ‒ accountability, development and support ‒ embedded in a complex and multi-level system of resources, people and relationships. To address the study objectives, the research used a mix of qualitative and quantitative methods. Three studies were conducted in a phased process: study 1 comprised a qualitative description of policy and practices in two districts related to the supervision of WBOTs; study 2 identified the main actors and mapped the supervisory system of WBOTs in the district, using social network analysis (SNA); and study 3 involved a qualitative exploration of workplace and interpersonal trust factors in the district and the supervisory system of WBOTs in the district. These three studies provided inputs for a workshop aimed at developing recommendations for a district-level, WBOT supportive supervisory framework. Four published papers reporting on the research conducted are presented in this thesis. It should be noted that the research was conducted during a turbulent political and administrative period in the NWP, when the WBOT programme changed from being a flagship programme for the country to one in crisis. This shifting context needs to be borne in mind when the findings are viewed and interpreted. Results: The study identified weaknesses in both the design and implementation of the supervisory system of WBOTs, with the absence of clear guidance resulting in WBOTs and PHC facilities performing their roles in an ad hoc manner, defined within local contexts. The study documented evidence of high internal cohesion within WBOTs and (where present) with their immediate outreach team leaders (OTLs). However, the relationships between WBOTs and the rest of the primary health care (PHC) and district health system were characterised by considerable mistrust – both towards other workers and the system as a whole. This occurred against a backdrop of increasing OTL vacancies, and the perceived abandonment of WBOT training and development systems and career opportunities. These findings are not dissimilar to those reported previously on the WBOT programme in South Africa and in programmes in other low-resource settings. Nevertheless, through its in-depth, exploratory and participatory approaches, this study provides additional insights into the phenomenon of supportive supervision. Firstly, in conceptualising supportive supervision as a set of ‘bundled’ practices within complex local health systems, the findings reflected the complexity of everyday realities and lived experiences. Secondly, through the embedded nature of the research and the phased data-collection process, the study was able to observe the impact of wider health system contexts and crises on the coalface functioning of the WBOT programme. Thirdly, the study emphasised how supportive supervision depends on healthy relational dynamics and trust relationships, and, finally, how a co-production approach can translate broad guidance, experience and theoretical understanding into meaningful, local practice owned by all the actors involved. Ultimately, the process of engagement, building relationships and forging consensus proved to be more significant than the supportive supervision framework itself. Conclusion: The lack of explicit, coherent and holistic guidance in developing CHW supportive supervision guidance and the failure to address supervision constraints at a local level undermine the performance and sustainability of CHW programmes. Effective supportive supervisory systems require bottom-up collaborative platforms characterised by active participation, sharing of local tacit knowledge and mutual learning. Supervisory systems also need to be designed in ways that promote relationships and generate trust between CHW programmes, other actors and the health system.