Research & Policy Briefs From the World Bank Malaysia Hub No. 43 March 2, 2021 Five Ways that COVID-19 Diagnostics Can Save Lives: Prioritizing Uses of Tests to Maximize Cost-Effectiveness Tristan Reed, William Waites, David Manheim, Damien de Walque, Chiara Vallini, Roberta Gatti, and Timothy B. Hallett Supplies of tests to diagnose the virus that causes COVID-19 (SARS-CoV-2) are still limited in many countries, and there is uncertainty about how to allocate the scarce supply across alternative types of testing to support diagnosis and disease control in the most cost-effective way. This Research & Policy Brief quantifies the cost-effectiveness of five types of interventions in terms of tests required per death averted. Across the five types, a single death can be averted by administering 940 to 8,838 tests, implying a large and positive return on investment in all five types—even assuming a very low value for loss of life. That is, all five types pay for themselves many times over. When prevalence of SARS-CoV-2 is high, the most cost-effective uses of SARS-CoV-2 diagnostics seem to be clinical triage of patients, at-risk worker screening, and population surveillance. Test-trace-isolate programs and border screening are also worthwhile, although they are more resource intensive per death averted if done comprehensively. These latter two interventions become relatively more cost effective when prevalence is low, and can stop the virus from entering a community completely. While governments should seek widespread deployment of tests in all five types of interventions, prioritizing them in this way is likely to maximize the cost-effectiveness of their use. As more contagious strains emerge, each type will become more valuable than ever. Introduction first quarter of 2020. In settings where households lack buffer stocks of food and savings, and must leave their homes for their livelihoods SARS-CoV-2, like many infectious diseases, can be transmitted from in the absence of social protection, lockdowns may not be feasible persons who are not obviously ill. This presents a major challenge and there may be greater reliance on a set of “social distancing” because it means that a public health response cannot rely upon measures. Such measures may have an important effect, but the symptoms to track and control its spread (Fraser et al. 2004). Indeed, available evidence does not suggest they would be sufficient to avert even if everyone immediately self-isolated upon the onset of an epidemic (Flaxman et al. 2020), especially as contagiousness rises symptoms, and the self-isolation eliminated any risk, this would with new strains. reduce transmission by, at most, 50 percent (Grassly et al. 2020)—and that is not enough to avert an epidemic of SARS-CoV-2. Fortunately, diagnostics can be used to test persons for the presence of current infection. This opens new approaches to control Because vaccinating enough of the population to reach herd transmission and ease the trade-off between economic and health immunity will take time, in both high-income countries (where a concerns with targeted rather than general containment measures. In majority are hesitant to get the vaccine immediately) (Galewitz 2021) combination with a range of other concerted activities (self-isolation and in middle-income and low-income countries (where supplies and nonpharmaceutical interventions), they could contribute could be constrained until 2024, according to a risk-assessment by enormously to saving lives and minimizing costly lockdowns. COVAX (Beaumont 2020), a global facility to provide universal vaccine access), countries continue to implement nonpharmaceutical This Brief considers five types of intervention (among many) to interventions, including social distancing and full or partial lockdowns. pinpoint the means by which testing for SARS-CoV-2 can contribute to Lockdowns can arrest epidemic spread in many settings (Flaxman et saving lives during the COVID-19 pandemic (table 1). These cases are al. 2020) but carry enormous economic costs—perhaps 0.25 to 0.86 provided as quantitative illustrations of what may be possible. The percent of GDP per week (see Acemoglu et al. 2020; Alon et al. 2020; precise way in which tests can be best put to use in any particular de Walque et al. 2020; Eichenbaum, Rebelo, and Trabandt 2020). setting, and the actual benefits derived, are highly sensitive to many Notably, China’s GDP declined by 0.86 percent per week in the factors specific to the setting and mode of use. The Brief considers Table 1. Summary of Diagnostic Cost-Effectiveness by Type of Intervention Intervention Tests per COVID-19 Tests Testing death averted deaths averted required population 1 Clinical triage and cohorting 940 106 100,000 100,000 patients upon admission 2 At-risk worker screening 1,042– 5,208 19–96 100,000 100,000 workers for one week 3 Population surveillance to 1,611 175 281,884 Regular samples per 100,000 for one year trigger or avoid lockdown 4 Test-trace-isolate 4,459 392 1,763,485 Regular samples per 100,000 for one year 5 Border screening 8,838 11+ 100,000 100,000 border crossers Source: Authors’ calculations. Note: Tests per death averted may not match ratio of tests required to COVID-19 deaths averted due to rounding. All scenarios consider the incremental value of testing, compared to a scenario where individuals are isolated based on symptoms alone. The major effect of border screening is in minimizing the introduction of virus to the country and so contributes to making all the other aspects of mitigating epidemic more likely to work. “ + ” indicates that the estimate of deaths averted refers only to the number of infections among those quarantined, who were not already infected. Affiliations: Tristan Reed, World Bank Development Research Group; William Waites, University of Edinburgh; David Manheim, University of Haifa; Damien de Walque, World Bank Development Research Group; Chiara Vallini, Boston Consulting Group; Roberta Gatti, World Bank Human Development Global Practice; and Timothy B. Hallett, Imperial College timothy.hallett@imperial.ac.uk and London and Modelling Guidance Group of The Global Fund for the ACT Covid-19 Diagnostics Accelerator. For correspondence: timothy.hallett@imperial.ac.uk treed@worldbank.org. treed@worldbank.org Acknowledgements: We are grateful to Norman Loayza, Young Eun Kim, and members of https://www.finddx.org/covid-19/act-accelerator-progress/ ACT Accelerator Diagnostics Pillar for helpful comments, and Nancy Morrison for excellent editorial assistant. Objective and disclaimer: Research & Policy Briefs synthesize existing research and data to shed light on a useful and interesting question for policy debate. Research & Policy Briefs carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions are entirely those of the authors. They do not necessarily represent the views of the World Bank Group, its Executive Directors, or the governments they represent. Five Ways that COVID-19 Diagnostics Can Save Lives: Prioritizing Uses of Tests to Maximize Cost-Effectiveness those tests that detect infection 90 percent of the time (test may not be reflected in these calculations: for instance, screening of sensitivity), and correctly identify negative cases 97 percent of the key workers may enable more business to continue operating, and time (test specificity), in line with available antigen-based rapid screening haulers at the border (rather than closing borders) helps diagnostic tests. The details of each type are considered in turn in the international trade to continue. discussion that follows, although it is important to note that these cases do in fact interlock and depend upon one another. Five Types of Interventions The analysis also computes the cost-effectiveness of each type, as 1. Clinical Triage and Cohorting measured by “tests per death averted” (see results in table 1). The incremental impacts of diagnostics are considered, over and above In a clinical setting, diagnostic tests may be used to confirm patients’ self-isolation and social distancing. Given the low cost of a single test, infection so that infected patients may be grouped together and the results suggest benefits that substantially exceed costs: a death isolated from other patients, a policy known as cohorting. Without a can be averted for the cost of $4,700 to $44,190, depending on type, test for SARS-CoV-2, this may be done instead based on symptoms, or the cost of 940 tests to 8,838 tests at $5 each. Following others whereby hospitals group all patients without symptoms together. As a who have quantified a return on investment (ROI) with respect to the result, those who are not infected may be grouped with those who are recommended value of a statistical life, the ROI of one test ranges infected with SARS-CoV-2 but are asymptomatic. between 3.5 (for border screening) and 41.6 (for clinical triage and Authors’ calculation shows that cohorting using a diagnostic test cohorting) when assuming a low value of statistical life of only rather than symptoms alone can reduce infections originating from $200,000— which is 2 percent (1/50) of the estimate for the United the hospital by 87 percent, resulting in 106 fewer deaths per 100,000 States derived by Robinson et al. (2019). Higher valuations would lead patients at the peak of the epidemic. It is assumed that 50 percent of to even higher returns. patients entering the hospital have SARS-CoV-2 at the peak of an epidemic. To account for risk of in-hospital infection, it is assumed In certain circumstances (described in the discussion that follows), that the potential for transmission (i.e., R0, the basic reproduction these returns are in addition to other major benefits, such as avoiding number) is 2.5 in the hospital setting (meaning that each infection lockdowns and increasing confidence among consumers and workers. leads to 2.5 more infections), but this does not affect the These estimates tend to be conservative, as they do not proportionate reduction in transmission risk that is provided by the incorporate several factors that are likely to be important. First, use of testing (see Excel spreadsheet for calculation). https://drive.google.com/file/d/1HJ04XH23Ak7h7ARvBEqVH-TtuGGbZA1e/view several of the calculations consider only the effect of the intervention on reducing transmission from a single person, and do not account for 2. At-Risk Worker Screening the effect that this may have in stemming a whole chain of transmission. The analysis focuses only on deaths, whereas The risk of any group of workers contributing to onward transmission SARS-CoV-2 also causes substantial illness, and long stays in hospital can be reduced by screening them for SARS-CoV-2. Screening enables for many patients are costly and may overwhelm the health system, those persons who are infected but not symptomatic to self-isolate (in increasing the risk of death overall. The benefits of controlling addition to those who have symptoms) (Grassly et al.2020). The hospital-acquired (nosocomial) infections— by placing patients impact depends on how often they are screened and the delay until known to be infected with SARS-CoV-2 in isolation wards together and they self-isolate (if the test is positive). If screening is done weekly and screening health care workers may have the additional benefit of test results are provided before they risk infecting others, this could lessening disruption of other services and freeing up staff to attend to reduce the risk of onward transmission by about 32 percent. other forms of care; this could help avoid knock-on consequences of The cost of at-risk worker screening depends on the number of the COVID-19 epidemic (Hogan et al. 2020). Other economic benefits tests needed to find one person who is currently infected who would not self-isolate without screening: that is, the prevalence of Figure 1. How the Prevalence of SARS-CoV-2 among Health Workers SARS-CoV-2 among those without symptoms in that group. Thus, Influences the Number of Tests Needed to Avert One Death screening is more cost effective when it is done on higher-risk groups, such as health care workers, and those in public-facing roles. This Testing is especially cost effective when prevalence is high. prevalence will depend on the epidemic stage and other factors: for example, it was found to be as high as 10 percent among health 12,000 workers in London at the peak of the epidemic (see, for example, Treibel et al. 2020), and 20 percent among those on COVID-19 wards 10,000 in an Oxford hospital (Eyre et al. 2020), but 2 percent in frontline Tests per death averted 8,000 workers in the United Kingdom overall some weeks later (see, for example, the 2020 ONS survey in the United Kingdom). Between 6,000 these high and low extremes, the number of tests per death averted ranges between 1042 and 5208, respectively (figure 1). Thus, when 4,000 deployed among a group at very high risk, this would appear to be one of the most cost-effective and widely applicable types of 2,000 interventions. - 0 5 10 15 20 25 30 3. Population Surveillance to Trigger or Avoid Lockdown Prevalence among health care workers, percent If the policy is to implement a lockdown in an area (such as a large city) if there is a chance that the prevalence of SARS-CoV-2 infection is Source: Authors’ calculations. Note: Calculation assumes 100,000 health care workers, a case fatality rate of 1 more than 2 percent, random samples could be drawn from the percent, and a basic reproduction number of 3, consistent with a high degree of population to inform that decision. This trigger value of prevalence is interaction in the clinical setting. 2 chosen somewhat arbitrarily. Smaller trigger values would require Research & Policy Brief No.45 larger sample sizes but lead to longer periods of lockdown. It is an open question (not addressed here) as to the optimal triggers to use Box 2. Test-Trace-Isolate in Practice in this regard. The Republic of Korea experienced a steep growth in COVID-19 cases As an example, a daily random sample of 772 persons (or an early in the pandemic, but it quickly reduced rates of infection and equivalent) could be used to check if it can be ruled out (with 99 maintained low numbers of daily new cases. Korea did not implement percent certainty) that prevalence exceeds 2 percent (see strict lockdown measures, but focused on case-based contact tracing drive.google.com/file/d/1Pj-gappendix https:/online A for JaoeH_QEhjiURfHZd6Me8NaS2-T7/view?usp=shar ing derivation) To have the same certainty that the and cluster testing and isolating. The country expanded testing prevalence is below 0.1 percent would require a random sample of capacity from 3000/day on February 7, 2020 to 15,000/day to 15,745 persons. The same-size sample is required to measure 20,000/day, with a turnaround time of 6 hours to 24 hours by the end prevalence (with a given level of certainty) in any population, if the of March. All suspected cases and patients under investigation are sampling is representative. tested. Contact-tracing is performed through a mix of patient Using the epidemiological model described inhttponline appendix s:/ drive.google.com/file/d/15ah- B, ing 8rDspbkNzvtYh9AeVX3JNIdjJ1zB/view?usp=shar interviews and analysis of mobile phone location, credit card transactions, and health data. this analysis calculates lives saved from a lockdown policy that triggers when more than 2 percent of tests are positive, and is released when Source: Dighe et al. 2020. less than 1 percent of tests are positive. The lockdown is calibrated to reduce interpersonal contact such that the Reproduction Number (R) falls below 1. If implemented for one year, the lockdown policy guided by random sampling is expected to avert 175 deaths per 100,000 The analysis focuses on a highly effective testing and tracing people at the cost 281,884 tests per year (that is, 772 tests per day), system whereby tests are provided quickly (90 percent within 48 leading to a cost-efficiency of 1,611 tests per death averted. hours) to those who are self-isolating; the message to self-isolate is conveyed to all contacts rapidly (90 percent within 48 hours); and the By the same token, because lockdown can be released quickly (or intervention begins from a point when incidence is low (0.01 percent avoided altogether when unnecessary), days of unnecessary of the population infected). It considers a city of 100,000 people in lockdown are avoided, and this would minimize disruptions to the which other available interventions have already been used. The economy. Box 1 describes how such an approach may have analysis assumes that the R0 is 1.5, having been reduced through contributed to imposing and relieving lockdowns in Italy. effective social distancing measures, and further assumes that 25 percent of persons self-isolate effectively upon the onset of symptoms that could be caused by COVID-19. In addition, persons may have Box 1. Population Surveillance to Guide a Response in symptoms that are caused by another pathogen but that are mistaken Practice as being those of COVID-19, and this leads to unnecessary periods of isolation for some. Italy was the country second worst affected by COVID-19 after China by March 2020. When the country suffered its first death on February Compared to a scenario without it, the test-trace-isolate 22 in the small town of Vo (3,000 inhabitants) in Veneto, the whole intervention could help avoid a wave of the epidemic, preventing 392 town was put in quarantine and every inhabitant was tested. During deaths per 100,000. If a more general lockdown would have been the first round of testing, 89 people tested positive. During the second ordered (as per the same criteria), the intervention would also help round, 9 days later, only 6 were infected. Interestingly, the Italian avoid that lockdown. This would spare the economy a loss of 12 authorities found that at the time of the first symptomatic case, about percent of GDP, assuming productivity under lockdown is 70 percent 3 percent of the population had already been infected and most of lower than normal. them were completely asymptomatic. Through mass testing and isolation of those infected, the virus was eradicated from the town The distinctive contribution of the testing component (as opposed rapidly. At least 60 percent of all people infected by the virus were to the tracing and isolating components) is to allow persons who are asymptomatic. Mass testing can give a clear picture of how many not actually infected to avoid the period of isolation (which they people are carrying the virus and can transmit it to others. would endure if there were no testing and they had symptoms or were in recent contact with someone who did and were traced). An Source: Crisanti and Cassone 2020. intervention that includes testing leads to 29 percent fewer person-days spent in isolation than one that does not include testing. These calculations are sensitive to many assumptions but are 4. Test-Trace-Isolate broadly consistent with other independent analyses of the impact of Again using the epidemiological model outlined in https:/ drionline Appendix ve.google.com/file/d/15ah-8rDspbkNzvt Yh9AeVX3JNIdjJ1zB/viB, ew?usp=sharing contact tracing (Ferretti et al. 2020; Grassly et al. 2020; Kucharski et al. the analysis considers a “test-trace-isolate” (TTI) intervention being 2020). The assumption used here that half the people infected with rolled out, whereby an index person who has had symptoms (who is SARS-CoV-2 have no symptoms but have the same risk of transmission self-isolating) receives a test. If the test is positive, their recent as others makes our estimates of the impact of the “test-trace-isolate” contacts are instructed to self-isolate, and they, in turn, are tested. If intervention conservative. However, the intervention would be their test is negative, they will cease self-isolating. If the test is substantially less effective if it was slower to test and trace persons positive, a new round of contact tracing occurs whereby their own than assumed here, or started from a point of a higher number of recent contacts are instructed to self-isolate and then are tested. TTI cases. The intrinsic value of testing per se would also be lower if the is the most targeted form of lockdown; it is applied only to those who frequency of symptoms not caused by SARS-CoV-2 (but rather by are infectious. Box 2 presents a the case study of this intervention in influenza) was lower, or adherence to self-isolation did not increase practice. even if those tested received a positive test result. 3 Five Ways that COVID-19 Diagnostics Can Save Lives: Prioritizing Uses of Tests to Maximize Cost-Effectiveness 5. Border Screening To arrive at these results, the analysis has assumed that disease prevalence is very high—that is, the pandemic is out of control—as is The number of infected persons entering a country can be reduced the case in many countries. A caveat demonstrated in figure 1 is that through testing and quarantine of those testing positive. Again, this deaths averted per test falls as disease prevalence falls, because with approach is more effective than solely quarantining persons who have lower prevalence more tests will be used on those who are not symptoms. The proportional reduction in the number of persons with infected. For countries with very low prevalence and where very few SARS-CoV-2 who enter the country following screening is equal to the cases would be detected in a clinical setting, border screening may be sensitivity of the test: so, a 90 percent sensitivity test implies 90 more cost effective than either clinical triage and cohorting or percent fewer introductions of infection to the country, compared to screening of at-risk workers. Readers may use our https:/ drive.googlExcel tool e.com/file/d/1HJ04XH23Ak7h7ARvBEqVH-TtuGGbZA1e/ view to screening with symptoms alone. The fewer persons who enter a compare cost-effectiveness under alternative levels of prevalence. country with infection, the less likely it is that the infection will spread widely and the more likely it is that effective measures can be put in Additional considerations concern how transmissibility and place to control it. Of course, the effectiveness of border screening virulence affect the calculations. New, more contagious strains imply also depends on what coverage of border-crossers can be achieved. that any single social interaction is more likely to lead to infection, which will only make testing more valuable. Differences in the age There is further risk that, without testing, quarantine of persons distribution and obesity rates across communities may affect the case with symptoms crossing the border can result in additional infections, fatality rate (Goldberg and Reed 2020) and thus deaths averted per if those quarantined are housed in the same location as those without test, though such differences will not materially affect the ranking of SARS-CoV-2 (for example, Zimbabwe reports new infections at the five types of interventions by cost-effectiveness within a quarantine facilities, ACF 2020). Authors’ calculations show that community. Moreover, even with a lower case fatality rate, each type “border screening” could achieve 11 fewer deaths per 100,000 will still have an ROI greater than 1, even for a very low value of a people crossing the border. Here, it is assumed that 1 percent of the statistical life. population crossing the border has SARS-CoV-2, and that R0 is 3.0, to account for the fact that social distancing may not be adhered to at The effectiveness of testing is linked to how much societies invest border crossings. A lower R0 would imply fewer deaths averted per in it. If countries invest more in the research and development of test (see Excel spreadsheet with the calculation). https://drive.google.com/file/d/1HJ04XH23Ak7h7ARvBEqVH-TtuGGbZA1e/view more accurate and rapid tests and in the complementary skills required for delivery and tracing, this investment can help save even Economic Considerations more lives and livelihoods. This Brief has illustrated five ways that SARS-CoV-2 diagnostics can Finally, social returns from testing are higher than individual save lives, and has estimated the cost-effectiveness of each type of returns. Indeed, contagion implies a negative externality (a cost to intervention using a standard epidemiological approach. There is a others that is not internalized by the contagious individual), while large and positive return on investment in diagnostics in all five types, testing implies a positive externality (a benefit that extends to society even for a very low value of a statistical life. Clinical triage and beyond the contagious individuals’ benefit). This wedge between cohorting; population surveillance to trigger or avoid lockdown; and private and social benefits is particularly high for asymptomatic but at-risk worker screening appear to be the most cost effective, followed infectious people because they have no incentive to get tested or to by test-trace-and-isolate programs and border screening. While isolate themselves. To solve this externality, governments should think governments should seek supplies to deploy diagnostics in all five about using incentives for people to get tested. Such incentives for types, prioritizing them in this order is likely to maximize the testing have been used in the past for tuberculosis (TB) and HIV cost-effectiveness of their use. (Geffen 2011; Keating 2013). References Acemoglu, D., V. Chernozhukov, I. Werning, and M. D. Whinston. 2020. “A Multi-Risk SIR Model with Fraser, C., S. Riley, R. M. Anderson, and N. M. Ferguson. 2004. “Factors that Make an Infectious Disease Optimally Targeted Lockdown.” NBER Working Paper 27102, National Bureau of Economic Research, Outbreak Controllable.” Proceedings of the National Academy of Sciences 101 (16): 6146–51. Cambridge, MA. Galewitz, P. 2021. “Poll: Nearly Half of American Adults Now Want the Covid Vaccine—ASAP” Kaiser Health Action Contre La Faim (ACF). 2020. “Zimbabwe: Under-equipped quarantine centers are increasing the News (KHN), January 27. spread of COVID-19” Press Release. 29 July. https://khn.org/news/article/poll-nearly-half-of-american-adults-now-want-the-covid-vaccine-asap/ https://www.actioncontrelafaim.org/en/press/zimbabwe-under-equipped-quarantine-centers-are- https://www.actioncontrelafaim.org/en/press/zimbabwe-under-equipped-quarantine-centers-are-increasing-the-spread-of-covid-19/ increasing-the-spread-of-covid-19/ Geffen, N. 2011. “What Does Science Show about Incentives for HIV Testing?” Quackdown, December, 13. Alon, T. M., M. Kim, D. Lagakos, and M. VanVuren. 2020. “How Should Policy Responses to the COVID-19 https://www.quackdown.info/article/what-does-science-show-about-incentives-hiv-testing/. Pandemic Differ in the Developing World?” NBER Working Paper 27273, National Bureau of Economic Goldberg, P. K., and T. Reed. 2020. “The Effects of the Coronavirus Pandemic in Emerging Market and Research, Cambridge, MA. Developing Economies: An Optimistic Preliminary Account.” Prepared for the 50th Brookings Papers Beaumont, P. 2020. “Scheme to Get Covid Vaccine to Poorer Countries at ‘High risk’ of Failure.” The on Economic Activity Conference, June. Guardian, December 16. https://amp.theguardian.com/world/2020/dec/16/scheme-to-get-covid-vaccine-to-poorer-countries- Grassly, N., M. Pons Salort, E. Parker, et al. 2020. “Role of Testing in COVID-19 Control.” Imperial College https://amp.theguardian.com/world/2020/dec/16/scheme-to-get-covid-vaccine-to-poorer-countries-at-high-risk-of-failure. at-high-risk-of-failure. https://amp.theguardian.com/world/2020/dec/16/scheme-to-get-covid-vaccine-to-poorer-countries-at-high-risk-of-failure. London, COVID-19 Report 16, April, 23. Crisanti, A., and A Cassone, 2020. “In One Italian Town, We Showed Mass Testing Could Eradicate the Hogan, A. B., B. L. Jewell, E. Sherrard-Smith, et al. 2020. “Potential Impact of the COVID-19 Pandemic on HIV, Coronavirus.” The Guardian, March 20. Tuberculosis, and Malaria in Low-income and Middle-income Countries: A Modelling Study.” The https://www.theguardian.com/commentisfree/2020/mar/20/eradicated-coronavirus-mass- https://www.theguardian.com/commentisfree/2020/mar/20/eradicated-coronavirus-mass-testing-covid-19-italy-vo. Lancet Global Health 8 (9): e1132–e1141. testing-covid-19-italy-vo. Keating, J. 2013. “The Nudgy State.” Foreign Policy, January, 2. de Walque, D., J. Friedman, R. Gatti, and A. Mattoo. 2020. “How Two Tests Can Help Contain COVID-19 and https://foreignpolicy.com/2013/01/02/the-nudgy-state/ Revive the Economy.” Research & Policy Brief 29, World Bank Group, Washington, DC. Dighe, A., L. Cattarino, G. Cuomo-Dannenburg, et al. 2020. “Response to COVID-19 in South Korea and Kucharski, A. J., P. Klepac, A. Conlan, S. M. Kissler, M. Tang, H. Fry, J. Gog, J. Edmunds, and CMMID COVID-19 Implications for Lifting Stringent Interventions.” BMC Medicine 18 (321). Working Group. 2020. “Effectiveness of Isolation, Testing, Contact Tracing and Physical Distancing on https://doi.org/10.1186/s12916-020-01791-8. Reducing Transmission of SARS-CoV-2 in Different Settings: A Mathematical Modelling Study.” The Lancet Infectious Diseases 20 (10): 1151–60. Eichenbaum, M. S., S. Rebelo, and M. Trabandt. 2020. “The Macroeconomics of Epidemics.” NBER Working Paper 26882, National Bureau of Economic Research, Cambridge, MA. Robinson, L. A., J. K. Hammitt, and L. O’Keeffe. 2019. “Valuing Mortality Risk Reductions in Global Benefit-Cost Analysis.” Journal of Benefit-Cost Analysis 10 (S1): 15–50. Eyre, D. W., S. F. Lumley, D. O'Donnell, et al. 2020. “Differential Occupational Risks to Healthcare Workers from SARS-CoV-2: A Prospective Observational Study.” MedRxiv. doi: 10.7554/eLife.60675. Treibel, T. A., C. Manisty, M. Burton, et al. 2020. “COVID-19: PCR Screening of Asymptomatic Health-care Workers at London Hospital.” The Lancet 395 (10237): 1608–10. Ferretti, L., C. Wymant, M. Kendall, et al. 2020. “Quantifying SARS-CoV-2 Transmission Suggests Epidemic Control with Digital Contact Tracing.” Science 368 (6491): eabb6936. United Kingdom, ONS (Office for National Statistics). 2020. Coronavirus (COVID-19) Infection Survey. Flaxman, S., S. Mishra, A. Gandy, et al. 2020. “Estimating the Effects of Non-Pharmaceutical Interventions on https://www.ons.gov.uk/peoplepopulationandcommunity/healthandsocialcare/conditionsand https://www.ons.gov.uk/peoplepopulationandcommunity/healthandsocialcare/conditionsanddiseases/datasets/coronaviruscovid19infectionsurveydata. COVID-19 in Europe.” Nature 584 (7820): 257–61. diseases/datasets/coronaviruscovid19infectionsurveydata. 4