This multidisciplinary project adopted a 'whole systems' approach using methods from epidemiology, anthropology, and health systems research (Systems dynamic modelling) to understand the context, practice, and the potential for effective implementation of IPC for TB in South Africa. This project was conducted over four years (2017–2021) and had three stages: 1) observe & measure (data collection), 2) combine & design (system dynamics workshops) 3) model & cost (mathematical and economic modelling). These three phases of the project addressed seven research question. Research question 1 described the policy and systems context by looking at how South African policies on IPC for TB have evolved and been implemented. We spoke with members of civil society, and policymakers. For Research question 2, which related to the epidemiological context, we estimated how much TB transmission happens in clinics compared to other community locations. We estimated how many adults attending clinics had active TB and/or TB symptoms. We also estimated the risk of contact between people with infectious TB and other clients within clinics, and separately estimated, among community members, the frequency of social contacts in clinics as compared to other settings where people meet. Research questions 3 and 4 examined the effect of clinic design and working practices on transmission and looked to understand healthcare workers perceptions of risk and responsibility. We used structured and in-depth qualitative methods to document IPC practice in health clinics considering the role of clinic design, organisation of care, work practices, as well as health care worker, manager, and patient ideas about risk and responsibility in IPC. We spoke to patients, health workers, as well as specialists in primary care, IPC, and the built environment. The collected data enabled us to calculate the ventilation of waiting areas and consultation rooms; and we examined how people moved around clinics and where they spent time. Research question 5 involved the designing of whole-systems interventions to improve TB infection prevention and control. We used system dynamics modelling (SDM) to bring our data together and design interventions. With researchers, patient and union representatives, practitioners from clinics and hospitals, and policymakers from District, Provincial, and National Departments of Health, we developed ‘models’ (diagrams) of the system and identified targets for interventions to reduce Mtb transmission. Our collaborators prioritised interventions based on how likely they were to be effective and how easily they could be implemented. Research questions 6 and 7 involved synthesis of all these data to develop a package of health systems interventions to reduce DR-TB transmission in clinics, adapted to the constraints and opportunities of the South African health system. We used mathematical and economic modelling to project the potential impact of interrupting clinic-based transmission on community-wide TB incidence, and the consequent economic benefits for health systems and households.

This study was conducted over four years (2017–2021) and had three stages: 1) observe & measure (data collection), 2) combine & design (system dynamics workshops), and 3) model & cost (mathematical and economic modelling). All data collection was done before the start of the COVID-19 pandemic. Data collection. For the policy setting we conducted in-depth interviews with policy actors (health system, researchers, activists) at various levels of the health system, from local clinics to global policymaking bodies as well as specialists in primary care, IPC, and the built environment. The prevalence of TB survey involved randomly selecting adults (≥18 years) attending 2 primary healthcare clinics who were interviewed and requested to give sputum for mycobacterial culture. (quantitative data) For the clinic setting we used structured and unstructured observations and formal interviews and focus group discussions and informal conversations with clinic managers, health care workers, and patients. Patient flow was mapped in the clinics - unique barcodes were used to track attendees’ movements in 11 clinics in two provinces, multiple imputation was used to estimate missing arrival and departure times, and mixed-effects linear regression to examine associations with visit duration . Clinic ventilation was measured in clinic spaces using a tracer-gas release method. We used three different types of modelling including systems dynamic modelling, mathematical and economic the latter two using data collected in the project to generate code. System dynamic modelling collected data during two one-day participatory group model-building workshops with policy- and decision-makers at national- and province-levels, and patient advocates and health professionals at clinic- and district-level. Causal loop diagrams were generated by participants and combined by investigators.