key: cord-0762581-1jwrq6j9
authors: Reddy, Krishna P.; Shebl, Fatma M.; Foote, Julia H. A.; Harling, Guy; Scott, Justine A.; Panella, Christopher; Flanagan, Clare; Hyle, Emily P.; Neilan, Anne M.; Mohareb, Amir M.; Bekker, Linda-Gail; Lessells, Richard J.; Ciaranello, Andrea L.; Wood, Robin; Losina, Elena; Freedberg, Kenneth A.; Kazemian, Pooyan; Siedner, Mark J.
title: Cost-effectiveness of public health strategies for COVID-19 epidemic control in South Africa
date: 2020-07-01
journal: medRxiv
DOI: 10.1101/2020.06.29.20140111
sha: 3192e3140930df53d11889b5d3a26c0336495945
doc_id: 762581
cord_uid: 1jwrq6j9

BACKGROUND: Healthcare resource constraints in low and middle-income countries necessitate selection of cost-effective public health interventions to address COVID-19. METHODS: We developed a dynamic COVID-19 microsimulation model to evaluate clinical and economic outcomes and cost-effectiveness of epidemic control strategies in KwaZulu-Natal, South Africa. Interventions assessed were Healthcare Testing (HT), where diagnostic testing is performed only for those presenting to healthcare centres; Contact Tracing (CT) in households of cases; Isolation Centres (IC), for cases not requiring hospitalisation; community health worker-led Mass Symptom Screening and diagnostic testing for symptomatic individuals (MS); and Quarantine Centres (QC), for contacts who test negative. Given uncertainties about epidemic dynamics in South Africa, we evaluated two main epidemic scenarios over 360 days, with effective reproduction numbers (R(e)) of 1·5 and 1·2. We compared HT, HT+CT, HT+CT+IC, HT+CT+IC+MS, HT+CT+IC+QC, and HT+CT+IC+MS+QC, considering strategies with incremental cost-effectiveness ratio (ICER) <US$1,290/year-of-life saved (YLS) to be cost-effective. FINDINGS: With R(e) 1·5, HT resulted in the most COVID-19 deaths and lowest costs over 360 days. Compared with HT, HT+CT+IC+MS reduced mortality by 76%, increased costs by 16%, and was cost-effective (ICER $350/YLS). HT+CT+IC+MS+QC provided the greatest reduction in mortality, but increased costs by 95% compared with HT+CT+IC+MS and was not cost-effective (ICER $8,000/YLS). With R(e) 1·2, HT+CT+IC+MS was the least costly strategy, and HT+CT+IC+MS+QC was not cost-effective (ICER $294,320/YLS). INTERPRETATION: In South Africa, a strategy of household contact tracing, isolation, and mass symptom screening would substantially reduce COVID-19 mortality and be cost-effective. Adding quarantine centres for COVID-19 contacts is not cost-effective.

10 Table S2  11   Table S3  12   Table S4  13   Table S5  14   Table S6  15   Table S7  16   Table S8  17   Table S9  18   Table S10  19 Figure S1 

The Clinical and Economic Analysis of COVID Interventions (CEACOV) model consists of several modules that together determine individual health/disease trajectories and epidemic growth. These modules include natural history of disease, transmission, interventions including testing, and resource utilization.

There is much debate around appropriate cost-effectiveness thresholds, especially in low and middle-income countries. In this analysis, we applied an opportunity cost-based threshold for South Africa as reported by Woods et al. To be conservative, we applied the low end of their reported range ($1,175) and converted it from 2013 United States dollars (USD) to 2019 USD, yielding a threshold of $1,290 per year-of-life saved. 1, 2 Health States CEACOV simulates individuals transitioning between the states of susceptibility to severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), infection with SARS-CoV-2 and COVID-19 disease, recovery from COVID-19, and death. Susceptible individuals face a daily probability of exposure. After being infected with SARS-CoV-2, individuals may progress through the following health states ( Figure S2 ):

• Pre-infectious latency • Asymptomatic (or presymptomatic) infection • Mild/moderate disease: symptomatic • Severe disease: dyspnoea and/or hypoxemia ideally managed in a hospital with standard supplemental oxygen but not requiring intensive care unit (ICU) • Critical disease: ideally managed in an ICU with high-flow supplemental oxygen, non-invasive positive pressure ventilation, or invasive mechanical ventilation • Recuperation: only for those recuperating from critical disease and improving while remaining in the hospital or other health care facility • Recovered Individuals in the asymptomatic infection, mild/moderate disease, and severe disease states can transition directly to the recovered state. Individuals in the critical disease state can eventually die, or transition to the recuperation state and then to the recovered state. The recovered state is an absorbing state, and recovered individuals are assumed to have full immunity to SARS-CoV-2 over the model time horizon.

After being infected with SARS-CoV-2, a susceptible individual first transitions to the pre-infectious latency stage. Then, the individual has an age-dependent probability of progressing along one of four "paths," culminating in either asymptomatic infection, mild/moderate disease, severe disease, or critical disease. Before reaching a more advanced disease state, individuals must first transition through intermediate states (e.g., those destined for severe disease must first pass through the asymptomatic/presymptomatic infection state and the mild/moderate disease state) ( Figure S2 ). Transmission multipliers are setting-specific, time-dependent adjusting factors. They roughly account for population density and interventions that can alter the number of contacts and infectivity in the setting being modelled.

We assumed that all susceptible persons have an equal probability of contacting infected individuals and acquiring the virus (i.e., homogenous mixing). As the epidemic grows, the number of susceptible persons declines. Thus, not all the daily contacts of infected individuals will be with susceptible persons. The daily infection rate for a susceptible person is equal to the sum of transmission rates from all infected persons across all infection states divided by the cohort size. This leads to an expected daily number of infections equal to the number of susceptible persons multiplied by the infection rate on that day.

In this analysis, testing is performed on a nasopharyngeal specimen by polymerase chain reaction (PCR) assay. We assumed that test characteristics including sensitivity and specificity are independent of disease statei.e., the sensitivity is the same for those in the mild/moderate disease state and those in the critical disease state. We assumed that, after providing a specimen for testing and while awaiting the test result, hospitalised individuals are isolated and non-hospitalised individuals are advised to self-isolate at home. In the model, test results are acted upon (an intervention is started) on the day that the result is delivered.

To calibrate our model output with the COVID-19 epidemic in South Africa, we adjusted the transmission multiplier to generate an effective reproduction number (Re) of 1.5, matching that published by South Africa's National Institute for Communicable Diseases (NICD) based on empirical data collected in the country up to 19 May 2020. 3 We also evaluated alternative epidemic growth scenarios with Re=1.2 or Re=2.6, reflecting low and high estimates from different periods and regions in the NICD report.

We calculated age-stratified disease path probabilities. We used the proportions of people with COVID-19 who were: (a) asymptomatic, 4,5 (b) admitted to the ICU, 6 (c) hospitalised, 6 and (d) undiagnosed, 7 and the age-stratified proportions of different disease severity states. 8 We used the following sources to derive the duration of time in each state: presymptomatic infectious time; 8, 9 duration of viral shedding based on PCR detectability (WHO-China CDC Report, 8 Hu et al., 10 Zhou et al. 11 ); time to development of pneumonia (Wang et al. 12 ); time to ICU admission (Zhou et al. 11 ); time spent in the ICU (Zhou et al. 11 ); and median time to death (Zhou et al. 11 ). We calculated the transition probabilities until recovery (defined as the end of viral shedding) and the transition probabilities between disease states including death. Subsequently, after determining the duration in each state, we estimated transition rates. We then calculated transition probabilities from the transition rates. 

We estimated the years-of-life saved (YLS) from each averted death from COVID-19 in KwaZulu-Natal, South Africa. To do this, we calculated years-of-life lost (YLL), defined as the average number of years a person would have lived had s/he not died from COVID-19. 13 The absolute number of YLL were: 13, 14 = ℎ * Where, Deathsage i is the number of deaths from COVID-19 in the age stratum, LEage i is the life expectancy in South Africa in the age stratum.

Therefore, age-stratified deaths and age-stratified life expectancy are needed. Because these data are not readily available, we estimated them from other sources, as follows.

1. Calculate age-stratified number of cases: We used the published South Africa NICD report of age-and sex-stratified COVID-19 cases 15 to estimate the proportion of cases in each age stratum, and then calculated the number of cases in each age stratum, given the total number of cases generated by the model. = ℎ * 2. Calculate age-stratified number of cases and deaths: The reported COVID-19 deaths in South Africa are not age-stratified. Therefore, to estimate the number of deaths in each age stratum, we used the agestratified case fatality ratio (CFR) from Massachusetts, USA, 6 the age distribution of the population of KwaZulu-Natal, 16 and the number of cases estimated from step 1 to estimate the number of deaths in each age stratum, given the total number of deaths generated by the model. 

Assuming that (a) R0 is 2.6 for individuals with asymptomatic and mild/moderate disease, (b) R0 is one-tenth of 2.6 for individuals with severe and critical disease, 19 and (c) viral shedding times are 9.5, 12, 19, and 24 days for individuals with asymptomatic, mild/moderate, severe, and critical disease, respectively, 8,10,11 we estimated the nominal transmission rate as described above in S1.1.

We applied costs from the health sector perspective. We adjusted costs to 2019 United States dollars, using South Africa-specific inflation and exchange rates. 11, 12 We obtained costs of clinical care from Mahomed et al. and Netcare Hospitals. 20, 21 We obtained the cost of PCR testing, including personnel and supplies, from the Africa Health Research Institute (personal communication).

We derived the number of ICU and non-ICU hospital beds available in KwaZulu-Natal (KZN) based on data reported by the South Africa Department of Health:

We derived the costs of additional intervention strategies from data supplied by the Africa Health Research Institute. The daily per-person costs of isolation and quarantine centre beds were based on the cost of a 500-person tent and personnel requirements.

To calculate the per-person cost of contact tracing and mass symptom screening, including personnel, supplies, and transportation, we assumed that community health workers could visit 30 households per day, with 5 individuals per house, on 20 days per month:

(the same per-person cost was applied for mass symptom screening)

We assumed that the per-unit costs of resources would be the same regardless of the total quantity. For example, per-test cost of performing a PCR assay was the same regardless of the number of PCR assays performed, and perperson daily cost of a stay at an isolation centre was the same regardless of the number of individuals housed at an isolation centre.

To calculate the increase in the cumulative probability of undergoing testing from mass symptom screening (MS) relative to contact tracing (CT), we assumed that MS would screen the population of 11 million twice per year. We assumed that individuals with mild/moderate symptoms are symptomatic for 10 days, on average: The ICER is the difference between two strategies in costs divided by the difference in life-years.

Strategies are listed in order of ascending costs, per convention of cost-effectiveness analysis.

In the base case, contact tracing and mass symptom screening cost $3/person. The ICER is the difference between two strategies in costs divided by the difference in life-years.

Strategies are listed in order of ascending costs, per convention of cost-effectiveness analysis.

In the base case, hospital beds cost $165/person/day and ICU beds cost $2,048/person/day. The ICER is the difference between two strategies in costs divided by the difference in life-years.

Strategies are listed in order of ascending costs, per convention of cost-effectiveness analysis.

In the base case, the PCR test has a 70% sensitivity and a 5-day result return time. The ICER is the difference between two strategies in costs divided by the difference in life-years.

Strategies are listed in order of ascending costs, per convention of cost-effectiveness analysis.

In the base case, the PCR test cost $27/test. DOMINATED: results in more life-years lost and higher costs than an alternative strategy.

The ICER is the difference between two strategies in discounted costs divided by the difference in life-years.

Strategies are listed in order of ascending costs, per convention of cost-effectiveness analysis.

In the base case, there is no restriction on the availability of isolation centre beds and quarantine centre beds.

In the sensitivity analysis, we reduced the available number of isolation centre beds and quarantine centre beds to 30,000 per 11 million people. The ICER is the difference between two strategies in costs divided by the difference in life-years.

Strategies are listed in order of ascending costs, per convention of cost-effectiveness analysis.

In the base case, the numbers of available hospital (non-ICU) beds and ICU beds are 26,220 and 744 per 11 million people, respectively.

*We changed the available number of hospital (non-ICU) and ICU beds to match the reported median numbers across countries in sub-Saharan Africa: 22,275 and 371 per 11 million people, respectively. The ICER is the difference between two strategies in costs divided by the difference in life-years.

Strategies are listed in order of ascending costs, per convention of cost-effectiveness analysis. The ICER is the difference between two strategies in costs divided by the difference in life-years.

Strategies are listed in order of ascending costs, per convention of cost-effectiveness analysis. Figure S1 . Model flowcharts of select COVID-19 epidemic control strategies in KwaZulu-Natal, South Africa.

After providing a specimen for testing and while awaiting the result, hospitalised individuals are isolated and non-hospitalised individuals are advised to selfisolate at home. Test results are acted upon (an intervention is started) on the day the result is delivered. "Contacts" can be symptomatic or asymptomatic. After providing a specimen for testing and while awaiting the result, hospitalised individuals are isolated and non-hospitalised individuals are advised to self-isolate at home. Test results are acted upon (an intervention is started) on the day the result is delivered. "Contacts" can be symptomatic or asymptomatic. After providing a specimen for testing and while awaiting the result, hospitalised individuals are isolated and non-hospitalised individuals are advised to self-isolate at home. Test results are acted upon (an intervention is started) on the day the result is delivered. "Contacts" can be symptomatic or asymptomatic. After providing a specimen for testing and while awaiting the result, hospitalised individuals are isolated and non-hospitalised individuals are advised to self-isolate at home. Test results are acted upon (an intervention is started) on the day the result is delivered. 

Country-level cost-effectiveness thresholds: initial estimates and the need for further research. Value Health

CPI Inflation Calculator

The Initial and Daily COVID-19 Effective Reproductive Number (R) in South Africa

Estimating the asymptomatic proportion of coronavirus disease 2019 (COVID-19) cases on board the Diamond Princess cruise ship

High incidence of asymptomatic SARS-CoV-2 infection

Massachusetts Department of Health

Estimating COVID-19 Prevalence in Symptomatic Americans

Report of the WHO-China Joint Mission on Coronavirus Diseaes

Temporal dynamics in viral shedding and transmissibility of COVID-19

Clinical characteristics of 24 asymptomatic infections with COVID-19 screened among close contacts in Nanjing

Clinical course and risk factors for mortality of adult inpatients with COVID-19 in Wuhan, China: a retrospective cohort study

Clinical characteristics of 138 hospitalized patients with 2019 novel coronavirus-infected pneumonia in Wuhan, China

Years of potential life lost (YPLL)--what does it measure? Epidemiology

Reflection on modern methods: years of life lost due to premature mortality-a versatile and comprehensive measure for monitoring non-communicable disease mortality

COVID-19 Update

Mid-year population estimates

Disease Burden and Mortality Estimates: Cause-Specific Mortality

World Population Prospects, South Africa

The reproductive number of COVID-19 is higher compared to SARS coronavirus

Cost of intensive care services at a central hospital in South Africa

Netcare Hospitals. Netcare Tariffs

COVID-19 Public Health Response

National estimates of critical care capacity in 54 African countries

The ICER is the difference between two strategies in discounted costs divided by the difference in life-years.Strategies are listed in order of ascending costs, per convention of cost-effectiveness analysis.In the base case, isolation centres cost $44/person/day and quarantine centres cost $37/person/day. ICERs for HT+CT+IC+MS are calculated compared with the next least costly strategy. In the base case, isolation centres and quarantine centres reduce transmission by 95%; isolation centre care costs $43·60/person/day and quarantine centre care costs $36·90/person/day.The ICERs indicated in the legend are opportunity cost-based thresholds for South Africa reported by Woods et al. 1 After converting to 2019 United States dollars, the lower end of the reported range is $1,290, and the midpoint of the range is $3,232.