Generally, molecular testing has multiple roles in disease management, including diagnosis, monitoring response to treatment and guiding decisions on appropriate management (e.g. indications for HIV drug resistance tests) (9). Further, as compared to other diagnostic counterparts (like TB culture), molecular tests are frequently able to return accurate results more rapidly, and hence allow for action to be taken at an earlier stage (9). Thus, molecular testing, if well-implemented, has the potential to greatly improve patient outcomes. This has been shown by the public sector molecular diagnostics programme in South Africa that generated a significant amount of health gains and economic value associated with these health gains. The return on investment was ≥1.4 for all individual tests, as high as 64.8 and 9.9 on average-suggesting excellent value for money across all molecular tests. These levels of return on investments for some tests rival childhood vaccination programmes (ROI 16)-often considered one of the most valuable public health interventions (32,33). While the ROI value for each of the individual types of molecular tests varied greatly, each still resulted in a positive impact on the South African healthcare system given that all tests had at least a full ROI.

Comparing to other public health tools such as vaccines and therapeutics, for each specific disease area is useful to understand the relative role of diagnostics as part of a public health programme. For SARS-CoV-2, the ROI of vaccination has not been estimated for South Africa, but in New York City was estimated to be 10 to 1(34). If adjusting this estimate for the GDP per capita in South Africa, and the smaller proportion of people at risk for severe disease and death in South Africa (due to demographic differences), the ROI for SARS-CoV-2 vaccines in South Africa might approach the value of diagnostics in South Africa. ROIs for specific therapeutics for SARS-CoV-2 in South Africa have not been calculated. Therapeutics such as paxlovid, however, have not been available in South Africa. For both TB and HIV, the return on investment for investments into vaccines has been effectively 0 (or negative) given that there are no vaccines available for HIV, and no effective vaccines for adults for TB (with the BCG vaccine for TB protective for <5-year-olds only). For HIV, the UNAIDS “Fast-Track” strategy which focuses on investments in expanding HIV service coverage (as opposed to maintaining constant coverage) estimates an approximate return on investment of 6.46 to 1 for Southern Africa (35,36). This contextualizes that viral load testing is complementary to a broader package of HIV-related services, with an ROI of 2, and that early infant diagnosis remains to have an extraordinarily high ROI. With respect to TB, there also exists a set of services which together have resulted in an estimated drop of more than 50% in the incidence rate between 2011 and 2022 (37). While it is difficult to quantify exactly how much each particular service (e.g. the increase in TB diagnostic capacity and accessibility, increased TB treatment options and the impact of improved HIV services) has contributed towards reduced incidence, the introduction of Xpert MTB/Rif and Ultra and the scale-up of the molecular testing programme has clearly contributed. Similarly in terms of investment, while there is no specific literature available on the return-on-investment of TB diagnostics, studies have shown that TB prevention and care yields a ROI of 43-implicitly including diagnosis-whereas we have estimated the ROI of diagnosis aspect to prevention and care to be 19.0 (38).

These values represent one consistent methodological way to calculate impact, economic value, and ROI. There are several changes in the underlying assumptions that might influence the direction of these values in both directions. Firstly, the average GDP per capita in South Africa may not reflect the burden of disease where the health gains occurred for each respective disease (e.g. slightly lower burden of disease among higher household social economic index score for tuberculosis infection) (39). If including these differences, the economic gains might be slightly overstated where the benefits occurred amongst individuals with lower annual GDP. We did not, however, want to weight the economic value of health differently among the South African population. In the other direction, within the scope of this analysis, we have only calculated the economic value of the health gains associated with molecular diagnostics rather than the total economic benefit. The total economic benefit likely far exceeds what we have estimated here-for example, the benefits associated with being able to return to work after a negative SARS-CoV-2 test during the pandemic or benefits associated with the reduced frequency of healthcare clinic visits required after a suppressed viral load test-reducing costs to individuals on antiretroviral treatment and possibly increasing economic activity due to fewer days spent seeking care. Estimating the total value was outside the scope of this analysis.

There were several limitations to our analysis. First, we did not include the value related to drug susceptibility testing (DST) for tuberculosis given the paucity of literature on the DALYs associated with TB resistance testing. If we had included that, the likely value and impact of the molecular diagnostic programme would have only increased, especially since the diagnosis of rifampicin resistant TB is provided simultaneously with the molecular diagnosis of TB. Second, for the incremental impact of the TB molecular diagnostics programme, we assumed that the size of the smear microscopy programme would have held constant in size from 2010 onwards. An expansion of this programme would have generated more health effects and related economic value of health gains, resulting in a diminished incremental benefit of the molecular programme. We made this assumption, however, given that the microscopy programme was already at capacity in 2010. Third, for the HIV viral load DALY calculations, it was assumed that individuals would be on a dolutegravir-based first line antiretroviral treatment regimen. The DALYs per viral load would have likely been greater in pre-dolutegravir era, given that previous regimens were slightly less effective. This would have resulted in a slight underestimate of the benefit of viral load testing before the introduction of dolutegravir-based first-line antiretroviral treatment regimens-but do reflect the likely benefit moving forward. Fourth, while we have presented here the net present benefits of the molecular diagnostics programme, the time horizon of the economic benefit associated with health gains differ between diseases. Finally, we did not consider the impact of potential additional costs incurred as a result of test sensitivity given the uncertainty around the costs that may occur as a consequence (e.g. false negative tests). Individuals who receive a false negative result may need to seek care again (or multiple times) to receive a correct diagnosis, incurring additional costs (most likely with TB testing, less so for SARS-CoV-2 given the short duration of infection). These individuals may also transmit infection if they are diagnosed negative (for TB, SARS-CoV-2, possibly for viral load if individuals think they are suppressed; not likely for EID). It is unlikely, however, that these tests will influence our results significantly, and the relative test sensitivity of molecular diagnostics is relatively high compared to other available diagnostics.