COVID-19 cases in Kenya are rising fast and we don’t know why.
Data on population movements and a recent seroprevalence study may offer clues.
COVID-19 cases in Kenya are accelerating rapidly. New cases have increased 300% month-over-month since April of this year while global and regional media have reported on the economic toll of stringent lock-down measures and heavy-handed government practices.
In previous posts, we explored how Africa and the rest of the world have struggled to understand the full extent of the pandemic on the continent. To date, there is no comprehensive analysis of how the crisis is playing out in sub-Saharan Africa and until now, the global community has been guilty of using limited data (cases and deaths) in most analyses, presenting a lopsided and incomplete picture of the pandemic in Africa.
To address this gap we have developed a bespoke analysis that integrates typical epidemiological datasets with population mobility data from Google Mobility reports and policy data from Oxford. We explore these data sources separately and in combination with one another through in an illustrative use case for Kenya. We then go a step further by comparing these data sources to a recent pre-print seroprevalence study, which adds deeper dimension to testing positivity and antibody presence in Kenyan populations.
For the first part of this analysis, we used three publicly-available datasets to construct a composite snapshot of the COVID-19 pandemic within Kenya from February 10th — August 10th.
Comparing these data side by side, we found the following:
1. Government policies only restricted daily movements for a short time and Kenyans continued to go to work
When restrictions (e.g. public events, gatherings, educational institutions) were implemented in early March, data showed a decline in mobility to grocery stores and pharmacies, parks, retail and transit stations.
In June, there was a gradual return to baseline mobility while many restrictions were still in place. Comparable rise in mobility for residences is not observed around the same period (early March). Instead, the rise in mobility for places of residence is much more gradual. This may be due to non-residence locations (grocery stores, parks, etc.) being easier to identify using mobile phone data compared to an individual’s “home.”
There is also a cyclical pattern for workplace mobility, indicating that many Kenyans were not working from home, likely because their job was not conducive to remote work. Another possible explanation is that, while the Kenyan government provided some financial relief, it may not have been sufficient to fully supplement Kenyans’ financial requirements.
2. Daily new cases and deaths are rapidly rising — but so are the number of tests
Throughout March and April, daily new cases in Kenya remained extremely low until approximately mid- to late-May when they began to rise. COVID-19 deaths were also generally low until June when they began to rise. Notably, up until June, 60% of days did not have a single COVID death, but since June 1st, there has not been a single day without a COVID-19 death and August is seeing average daily deaths in the double digits.
What, however, was the testing capacity was during this time period and how could it have changed over the past few months? We see that the pattern of new daily COVID-19 tests follows a very similar pattern to that of both cases and deaths. While there is some missing testing data in the early months, there is a clear association between increasing case and death detection and tests done.
3. Despite stringent policies initially implemented by the Kenyan government, after a short lull, cases continue to rise
Another critical insight for understanding a nation’s COVID-19 response is the policy landscape. The Oxford COVID-19 Government Response Tracker systematically collects, analyzes and presents policy responses over time, in countries around the globe, across several indicators.
From these data, we can see that public information campaigns were the first policy tool implemented in Kenya in February, suggesting that the government acknowledged the threat of COVID-19 early on. The next wave of policies came into play on March 14th — the date of Kenya’s first confirmed COVID case. On that date, we see sweeping regulations on international travel, public events, gatherings, workplaces, and schools.
Viewing policy data compared against testing data, we see that more liberal testing policies were put in place the first week of May. In April, the average number of new daily tests was about 650 and after testing policy expanded, average new tests in May tripled to nearly 1,900. We can see an additional surge in new daily tests from June to July when contact tracing policies intensified, resulting in an increase from 3,000 to 4,300 new tests per day on average.
Finally, we draw attention to the stay-at-home policy, gradually enacted in late March with a full lockdown in place by early April — lasting through June 22nd. From Google Mobility data, this policy coincides with a nearly 50% reduction in mobility during roughly the same time period. Notably however, mobility began to decline well ahead of a stay-at-home order and began returning to normal levels before the lockdown was truly lifted. This indicates that public information campaigns and individual behaviors were critically important early in the epidemic, while the economic demand to return to work and normal life activities led to more mobility in June and July — despite policies recommending otherwise.
4. A new seroprevalence study shows rising levels of SARS-CoV-02 antibodies
Despite available data on mobility, new cases, deaths, tests and policy, understanding the true underlying epidemiology of COVID-19 remains elusive given severe limitations in testing and potential biases arising from who gets tested, has severe disease and dies.
New data from a preprint (not yet peer reviewed) study on the seroprevalence of anti-SARS-CoV-2 IgG antibodies in Kenyan blood donors may provide clues as to the true trends of incidence and prevalence of COVID-19 in Kenya.
A seroprevalence study uses serology tests to identify people in a population that have antibodies against an infectious disease in order to estimate the percentage of the population that may have been infected. In addition, it shows how an infection progresses through the population over time.
In Kenya, the study population’s seroprevalence hovered around 5% throughout May and has been slowly rising since early June.
Similarly, Kenya’s test positivity rate — the percentage of all tests conducted that come back positive — was also increasing in early June. The test positivity rate remained below 10% until early June and increased dramatically in early July.
Combining our four data sources including the seroprevalence data indicates that COVID-19 incidence has been rising in Kenya as early as May. However, stay-at-home orders were not lifted until the end of June. So why did COVID-19 begin spreading so early and so rapidly?
1. Stay-at-home orders and bans on social gatherings have an impact on mobility — but so do individual behaviors
Stay-at-home orders and restrictions on social events and gatherings likely had a significant impact on population mobility. There was also a clear return to more normal mobility before lockdown was lifted and even while social distancing recommendations remained in place.
The likely explanation is that people needed to meet their subsistence and household income needs and economic support from the government wasn’t enough. To sustain individual behavior change over periods longer than a month, nations must provide adequate and effective economic and psychosocial support if maintaining limited contact is a priority.
2. Government policies on testing are critically important
The expansion of testing and contact tracing policies resulted in drastic increases in average daily testing levels, increasing capacity by thousands of tests per day.
This highlights the importance of government action in procuring commodities and supporting supply chains to ensure the availability of treatments, hospital supplies and infrastructure and personal protective equipment.
3. Testing rates and test positivity increased.
Although testing rates did increase dramatically over the time period, we also saw that test positivity increased, suggesting that the increase in tests alone did not account for the increase in cases.
Along with the trending seroprevalence data, it is safe to assume that underlying incidence was (and is still) on the rise. Only by taking each data source into consideration simultaneously can we draw accurate conclusions about the epidemiology of COVID and inform policy accordingly.
Overall, we saw that increased mobility, despite stringent country policies or increased testing, is likely the primary reason for the increase of COVID cases in Kenya. Renewed focus on policies which mitigate economic and psychosocial harm, while enabling responsible and safe behavior, will be key to sustaining a long-term response to the crisis.
What’s next for Kenya?
Our goal with this analysis was to identify underlying epidemic patterns from atypical combinations of publicly available data sources. In doing so, we reveal a data-driven, multidimensional view of Kenya’s experience with the COVID-19 pandemic.
As a global community, we have a mandate to think of fresh questions, hypothesize answers and generate new leads that may help slow or stop the ongoing crisis. In a previous post, we outlined a balanced approach that may offer another way to slow the spread of COVID-19 across the African continent.
Kenya could benefit from implementing hyper-targeted, data-driven strategies that take into account their population’s response to top-down policies. Knowing the location and frequency of population movement, combined with regular review of seroprevalence rates and typical testing data, could lead to clearer informational campaigns, informed placement of testing sites or public-safety strategies in key areas.
Data is our most powerful tool against COVID-19. By leveraging its data, Kenya can implement better and more proactive policies and strategies to gain an edge against the pandemic.
This analysis was led by Cooper/Smith’s resident interns Nathali Gunawardena and Deborah Chan, Masters of Science in Public Health students at McGill University and supervised by Dylan Green, MPH, PhC in Epidemiology at University of Washington. Nathali holds a BS in Biomedical Science with a minor in Psychology with extensive research experience on maternal and child health in Africa. Deborah is joint researcher and clinician as a Nurse with quantitative research interests in global health. Together, they have been invaluable members of our team as we continue to attempt to make sense of clinical and epidemiological data during the COVID-19 pandemic.
