Our findings show that the risk of transmission per minute spent in congregate settings or on public transport varies greatly between different locations/transport types. High levels of variation were found both between different types of location and transport, and also between different buildings/transport of the same type. Nevertheless, our findings provide a clear indication, that in the community we studied, transmission risk is likely to be higher in minibus taxis and trains than in salons, bars, and shops. Public transport may therefore be promising targets for infection prevention and control interventions in this setting, both to reduce Mtb transmission, but also to reduce the transmission of other airborne pathogens such as measles and SAR-CoV-2.
The drivers of high relative rate of transmission varied between the two high risk transport types and suggest that different intervention strategies may be optimal in different types of location. In minibus taxis, transmission risk was predominantly driven by low ventilation levels. Guidelines and rules requiring some windows to remain open could greatly reduce risk, although they may be difficult to enforce. In trains, ventilation levels were more much more variable, ranging from 232 l− 1 – 2469 l− 1, with the number of people present being the main driver of transmission risk. Ensuring that windows are kept open may reduce risk, but even with adequate levels of ventilation, the very high numbers of people present in train carriages on some trips mean that risk is likely to remain high. Measures to reduce overcrowding on trains may be necessary to reduce risk to acceptable levels.
We estimated that ventilation rates were low in small, congregate buildings in the community such as hair salons and shops. As only small number of people were present at a time, however, the relative rate of potential transmission was low, and we do not consider these to be particularly high-risk locations. In contrast, while the ventilation rates in the (naturally ventilated) clinic were higher than in most other buildings (reflecting the larger size of the waiting area), estimated rates of potential transmission in the clinic were nevertheless high at times, due to high occupancy levels. While we only collected data from the single clinic that serves the study community, this finding can be cautiously generalised to many other clinics in South Africa, as overcrowding in public clinics is common . An increased prevalence of people with undiagnosed tuberculosis in clinics may further increase risk [24, 25]. Unless very high ventilation rates can be achieved in clinics, or crowding reduced, other measures such as GUV systems may be desirable to reduce risk.
WHO IPC guidelines recommend that ventilation systems are used to reduce Mtb transmission in settings with a high risk of transmission, but do not give an explicit recommendation for a minimum acceptable ventilation rate. Our findings support this, as they demonstrate that risk varies greatly with the number of people present in a room, in addition to ventilation rates. Far higher ventilation rates will be needed to reduce transmission risk to acceptable levels in crowded locations than in quiet ones.
We presented our main results as estimated rate of transmission relative to the median rate across the locations we studied. These estimated ‘rates’ are correlated with the observed CO2 concentrations (see supplemental data), which can be used directly as a proxy for transmission risk if estimates of ventilation rates are not required. We chose not to use the Wells-Riley equation  to estimate absolute risk. This is because the Wells-Riley approach relies on an estimate of the rate of quanta production, about which there is considerable uncertainty, with estimates for tuberculosis ranging from 0.62–8.2 or more [26, 27] – a greater than 10-fold difference. Estimates of risk by building type calculated using the Wells-Riley approach are given in the supporting material.
We recorded data on more than 1 day in more than one salon, bar, and shop. While the numbers of repeat visits were too small for a formal statistical comparison to be adequately powered, the data show that the variation in the ventilation rates between the same building on different days was as large as the variation between different buildings of the same type (Table S1). This finding is not unexpected. Factors such as changes in outdoor wind speed or windows being opened or closed can have large effects on ventilation rates . It presents challenges for data collection however, with recordings over a much larger number of days being needed to gain a full understanding of variation in ventilation rates. Ventilation rates may also have changed during data collection, particularly on transport where the speed changed and doors were opened and closed. Our estimated ventilation rates should be interpreted as average rates over the data collection period.
There are a number of limitations to our work. In estimating the relative rate of potential transmission in congregate locations, we assume that the prevalence of people with infectious tuberculosis is the same in all locations. This means that we may have underestimated the risk in the clinic, with the prevalence of tuberculosis in people attending clinics likely to be higher than the prevalence in many other locations . We also do not take into account the age distribution of people present, and the fact that the prevalence of infectious tuberculosis is low in children . We may therefore have underestimated the relative risk in bars, where the majority of people present were adults. The effect of varying prevalences of tuberculosis on estimated risk in other location types is unclear, however could be explored using mathematical modelling of social contact data. We may also have underestimated the relative rate of transmission in bars due to the timing of data collection, as safety concerns meant that we were unable to collect data at peak times on Friday and Saturday nights, when occupancy levels were likely to have been higher. Locations were selected by convenience sampling and therefore may not be fully representative of all locations, however they were chosen to cover a diverse range of locations (e.g. based on location size and structure). We sampled the only clinic in the community and approached owners or pastors of the other locations for permission to sample. Sampling was challenging: while most pastors and owners were happy for us to sample, and so a wide range of venues were sampled, counters often didn’t record or tampered with (batteries removed for example), and so the final sample represents those venues with viable data. Seasonality is likely to have an effect on ventilation rates, with rates likely to be lower in winter (and in summer in air conditioned buildings) when windows are more likely to be closed. The majority of our data collection in buildings and transport was conducted in summer and winter respectively (see supplementary data), and we may therefore have overestimated risk in transport relative to buildings. Finally, data on building/transport occupancy and ventilation levels should be combined with social contact or time use data to gain a more complete understanding of the likely contributions of different types of location/transport to overall Mtb transmission in the community.
We assumed that all people present in all settings had an activity level consistent with light slow walking on a level surface. If average activity levels were lower in a setting, for instance people sitting on a bus, then CO2 generation rates would have been lower, and we will have overestimated the ventilation rates. The reverse is also true, if activity levels were higher in a setting. Our estimates of relative risk by setting are robust to differences in CO2 generation levels however, as any under- or over-estimation of ventilation rates will be cancelled out by a corresponding over- or under-estimation of the risk associated with the number of other people present.
To conclude, we contribute knowledge to an important but neglected area of tuberculosis research, presenting estimates of ventilation rates and relative rates of potential transmission from a wide range of congregate settings in a high TB community in Cape Town, South Africa. We show that, in our setting, the rate of transmission is likely to be particularly high on public transport, making it a promising target for infection prevention and control interventions. Risk may also be high in clinics, even with high ventilation rates, due to high occupancy levels and an increased prevalence of people with infectious TB.