Abstract
South Africa has a high burden of drug-resistant tuberculosis (DR-TB). A policy to decentralize DR-TB treatment from specialized central hospitals to more accessible district facilities was introduced in 2011, but to date implementation has been suboptimal, with variable pace, coverage, and models of care emerging. This study explored multilevel policy implementation of DR-TB decentralization in two provinces of South Africa, Western Cape and KwaZulu-Natal. Applying interpretive policy analysis, this paper describes how actors across health system levels and geographies made sense of the DR-TB policy and how this shaped implementation. In an embedded qualitative case study, districts of the two provinces were compared, through data collected in 94 in-depth interviews, and analysed using Vickers' framework of reality, value, and action judgements. Five district cases characterize variation in the pace of implementation and models of DR-TB care that emerged. Individual and collective attitudes for and against the policy were underpinned by different systems of meaning for interpreting policy problems and making decisions. These meaning systems were reflected in actor stances on whether DR-TB care needed to be specialized or generalized, nurse- or doctor-led, and institutionalized or ambulatory. Actors' stances influenced their actions and implementation strategies adopted. Resistance to decentralized DR-TB care related to perceived threats of budget cuts to and loss of authority of central facilities, and was often justified in fears of increased transmission, poor quality of care, and inadequate resources at lower levels. New advances in diagnosis and treatment to address the growing burden of DR-TB in South Africa will have little impact unless implementation dynamics are better understood, and attention paid to the mindsets, interests, and interpretations of policy by actors tasked with implementation. Deliberative policy implementation processes will enhance the quality of discourse, communication and cross-learning between policy actors, and critical for reaching synthesis of meaning systems.