Limited COVID-19 testing threatens Africa as anniversary of pandemic approaches
While many more developed continents are buckling under the strain of COVID-19, African countries have been faring relatively well. To date, despite having 17% of the world’s population, Africa has reported less than 6% of the global COVID-19 cases and about 3% of the total COVID-19 deaths worldwide. Experts believe this is due, in part, to its younger population, warmer temperatures and cross-over immunity from other diseases. In addition, policymakers acted early — closing borders and enforcing stringent policies such as lockdowns and strict social-distancing measures.
However, it would be premature to say that Africa has escaped the worst of this global pandemic. A deeper look into its available testing data reveals troubling gaps.
First, less than half of all countries in the continent — 20 out of 54 — have total testing data represented on Our World In Data, which is one of the foremost data hubs for COVID-19. Johns Hopkins University reports cases and deaths for Africa, but not testing data; the same is true for the UN dashboard. The PERC dashboard, in collaboration with CDC Africa, does report test data, but it is not publicly accessible for analysis. Second, of the 20 countries that have testing data on Our World In Data, only four countries have a test positivity rate of 5% or lower, indicating insufficient testing. Such patchy data can obscure the true prevalence of COVID-19 and resulting fatalities in the region. We may be severely under-detecting the number of cases in the population, and a recent study by Imperial College suggested only 2% of deaths in mortality to COVID-19 were recorded in Khartoum, Sudan. Even though the continent may be truly experiencing a lower death rate than other parts of the world, we clearly do not know the true extent of fatalities. It could be the case that COVID-related deaths are being attributed to other health conditions. This incomplete picture limits the ability of governments to institute targeted and localized measures to curb the spread and mitigate its impact.
Decision makers need to look toward complementary metrics that do not rely on high-quality disease surveillance data to guide their response to COVID-19.
Confirmed cases and deaths have been concentrated in a few countries, but so has testing
As of January 6, 2021, there have been 2,914,291 confirmed COVID-19 cases and 69,858 confirmed COVID-19 deaths across 54 countries in Africa, according to Our World in Data (OWID). Figure 1 shows the total deaths and cases per hundred thousand people for 54 countries. Among this group, 65% of infections and 74% of reported deaths have been concentrated in four countries: South Africa, Morocco, Egypt, and Tunisia, even though they represent only 16% of the population in Africa. These numbers suggest that many African countries may have ducked the worst of COVID-19 pandemic.
Figure 1: Total (cumulative) Confirmed Cases and Deaths per 100k for each African country by January 6, 2021. Source: Our World in Data.
However, this might well be a result of testing rather than true impact. Testing across the continent remains low. Of the 20 countries where OWID cumulative testing data is available, 12 have performed fewer than 2,000 tests per hundred thousand individuals. Some of the better-performing countries, like Morocco (12,258 tests/100K), South Africa (11,513 tests/100K), Namibia (8,558 tests/100K), and Rwanda (5,754 tests/100K) still lag considerably when compared to the United States and the UK (which have administered 74,920 tests/100K and 79,991 tests/100K, respectively, as of January 6, 2021).
Since July 1, 2020, we’ve observed fairly steady increases in COVID-19 cases in countries such as Tunisia (+12,656%) and Morocco (+3,438%), while others have fallen and then risen again (South Africa, Algeria, Kenya, Zimbabwe, DRC), raising the question of whether testing is simply increasing or if a second wave is beginning (Fig. 2).
Figure 2: Weekly cases over time from July 1 to January 6, 2021. Source: Our World In Data.
Consequences of a testing deficit, and how to solve it
A lack of testing can lead to severe underestimation of actual COVID-19 prevalence, which can have multiple downsides. First, it may trap governments in a false sense of security where they relax restrictions around social distancing or mask mandates, contributing to a further increase in the spread of the virus. Individuals, too, may lower their guards and take less preventive health measures prematurely if they feel the virus has been contained. It may also lead to an underestimation of the need for increased medical resources, thus leading to more deaths. And it may also hinder efforts to mitigate the secondary consequences that tend to follow from an outbreak — such as food insecurity or job losses.
To understand the testing landscape in Africa, we use two key metrics: tests per hundred thousand people (tests/100K) and test-positivity rate (notwithstanding the challenges with the test-positivity rate). Test-positivity rate is defined as the ratio of the number of positive tests to the total tests performed. Test-positivity rate tends to be high when testing is insufficient or testing is only targeted towards key population groups or people with symptoms. The WHO recommends countries aim for a test positivity of 5% or lower.
Our analysis of OWID data indicates that testing in Africa does not fall below the WHO-recommended 5% test positivity in 15 out of 20 countries (Fig. 3). On the other hand, among the top 20 high-income countries with OWID testing data available, only 6 countries — the Netherlands (15.6%), Switzerland (13.7%), Qatar (11.5%), Austria (9.4%), Belgium (9.2%), and the United States (8.1%) — fail to meet this criteria as of January 6, 2021. This suggests that, in Africa, our perception of COVID-19 pandemic progression could be more optimistic than is justified by the numbers.
Figure 3: Scatterplot of Our World in Data cumulative test positivity. Dotted line marks 5% test positivity. Color denotes whether a country adheres to the WHO recommendation of 5% or lower test positivity. The best place to be in terms of test coverage is in the bottom right, as it means many tests are being done with low test positivity.
When we compare total deaths per 100,000 people with the total tests per 100,000, we see a generally positive relationship: countries that have performed more testing tend to report higher mortality (Fig.4). However, only four countries have performed greater than 5,000 total tests per 100,000. The rest have conducted far fewer tests — this suggests that we might be missing the complete picture in identifying the direct health impact of COVID-19 in these regions.
Figure 4: Scatterplot of Our World in Data Total Deaths per 100k versus Total Tests per 100k. Countries that perform greater than 5000 total tests per 100k are considered “High Testing,” and countries that report greater than 10 total deaths per 100k are considered “High Mortality.”
To address this underestimation of prevalence due to insufficient testing, organizations such as the Institute for Health Metrics and Evaluation (IHME) have generated forecasting models of COVID-19 infections, deaths, and tests for most African countries. However, their predictions rely on confirmed deaths instead of excess deaths, which is highly problematic given low testing rates across most of the continent. As a result, the IHME models most likely underestimate COVID-19 deaths and infections in most countries.
For example, the South African Medical Research Council reported 71,778 excess deaths in South Africa from May 6 to December 29, 2020. The IHME, on the other hand, projected the country to have a similar number of total deaths (27,610) by December 29 as the actual confirmed 27,568 deaths. By ignoring the fact that many COVID-19 deaths in Africa are never confirmed as such, we fear the model paints a more positive picture than is justified.
Providing alternate data sources to support vulnerable countries
In the Spring, we were concerned that countries and regions most vulnerable to the pandemic would be ignored, simply due to their lack of resources to detect and report cases and deaths. This concern was further elevated by seroprevalence studies conducted in Kenya, urban Malawi, and among frontline healthcare workers in Ibadan, Nigeria, which show much higher COVID-19 exposure than obtained from surveillance data. More novel approaches, like the use of satellite images, have documented the appearance of mass graves in countries like Iran, and Aden governorate in Yemen, indicating higher excess deaths than reported. Technology could also play a role in filling data gaps. Mobile apps have been used in the UK and US to enable people to record their symptoms over time. By tracking which users actually test positive for COVID-19, organizations have been able to estimate case prevalence in the broader population. Surgo is currently engaged in a similar effort for Africa. In partnership with VIAMO and the Clinton Health Access Initiative (CHAI), we are implementing a call-based symptom checker that allows users to report their symptoms for free, and receive a recommendation on what to do next.
Another way of dealing with spotty surveillance data is to avoid a strong reliance on it altogether and leverage alternate ways to measure vulnerability to COVID-19. This is why we decided to produce a subnational Africa COVID-19 Community Vulnerability Index (Africa CCVI) that could guide the COVID-19 response in the absence of high-quality, up-to-date data. This static index ranks 48 countries and more than 750 regions in Africa, by assessing vulnerability across seven themes: socio-economic, housing and transportation, population-density, healthcare, old age, fragility, and epidemiological factors.
Figure 5: Visualization of overall vulnerability scores covering 751 regions across 48 countries, accessible on Surgo Ventures’ Africa CCVI website.
Indeed, we can see that countries deemed more vulnerable by our Index are more likely to not report any testing data on OWID whatsoever. They simply fall off the radar. OWID provides data for only 20 countries, whereas Africa has over 50. The 20 countries we analyzed mainly consist of countries that are better off than those left out of the analysis. Our vulnerability index points to several countries considered to be highly vulnerable to the pandemic but not represented in the OWID total testing data: DRC, Niger, Sierra Leone, and Chad. Overall, out of the 16 countries with both OWID testing data and Socioeconomic Vulnerability values available, 14 are not considered socioeconomically vulnerable (relative to other African countries). This is an example of how data and insights can neglect those who are less wealthy and most vulnerable. We are sourcing data from the Africa CDC for every country in Africa, but in the meantime, this suggests we have very limited ability to monitor some of the most vulnerable countries.
How can decision makers, who are hamstrung by limited testing data, direct planning and pandemic response efforts towards communities that need them the most? In a forthcoming blog post, we will explain how the CCVI can help us scratch beneath the surface.
This work was made possible by everyone at Surgo Ventures, including but not limited to (in alphabetical order): Daniela Garcia, Rahul Joseph, Anubhuti Mishra, and Peter Smittenaar.