ACADEMIA Letters
Lessons from Modelling COVID-19 Scenarios in Kenya
and Implications for Policy and Planning
Nashon Juma Adero, Taita Taveta University
Abstract
This study examined and modelled the cross-country spread of the novel coronavirus disease
of 2019 (COVID-19) in Africa and outside Africa, finally converging on Kenya as the country of interest. A review of the models widely used to calibrate policy and strategic responses
confirmed the suitability of statistical models for predicting the spread of COVID-19, within a
10% margin of error. The main objective was to provide insights into the spread of COVID-19
and present “what-if” scenarios in aid of policy simulations within a compact bandwidth of
future scenarios, as required for policy and planning. The models proved resourceful in predicting the end-month cases within 10% and the likelihood of subsequent waves, both in terms
of the magnitudes and the time of starting and flattening. In the case of Kenya, the period of
the waves tended to be approximately four months within the study period presented here,
from April 2020 to May 2021. The key lessons imply a greater role of modelling, digitalisation, geospatial and mapping technologies, and transdisciplinary research in the aspirational
future of disease and disaster governance.
Keywords: digitalisation, disease and disaster governance, geomedicine, Health 4.0, policy simulation
Academia Letters, July 2021
©2021 by the author — Open Access — Distributed under CC BY 4.0
Corresponding Author: Nashon Juma Adero, nashon.adero@ttu.ac.ke
Citation: Adero, N.J. (2021). Lessons from Modelling COVID-19 Scenarios in Kenya and Implications for
Policy and Planning. Academia Letters, Article 1862. https://doi.org/10.20935/AL1862.
1
Introduction
In recent years, disease and disaster governance has become a public policy concern of borderless geopolitical implications. The novel coronavirus disease (COVID-19), first reported
in China in December 2019, epitomises this scenario. These disruptive developments have
accelerated the pace of the Fourth Industrial Revolution (Industry 4.0), expected to be more
pronounced in the health and education sectors with increasing uptake of digitalisation and
immersive technologies.
Since January 2020, publicly available COVID-19 data has been increasing, from barely
a million cases at the beginning of April 2020 to more than 157 million cases by May 7, 2021
(Republic of Kenya [Ministry of Health, 2021]; WHO, 2020; Worldometer, 2020; Worldometer, 2021). The increasing COVID-19 data volume has enabled the development of multi-scale
data-driven models to inform critical trends and containment policies globally, across Africa,
and within Kenya. The World Health Organization (2020) has since confirmed community
transmission as the dominant form of spread of COVID-19, Africa not spared the brunt.
This paper illustrates the critical role of data-driven models in the science-policy interface, using mathematical models to calibrate policies and strategies towards influencing behaviour change and containment action. The rest of the paper details the recent developments
in COVID-19 containment, the modelling theory and methodology used in the study, the results obtained, and the implications of the findings for containing the pandemic in Kenya with
respect to policy and planning.
Recent developments in understanding and containing COVID-19
Chaari and Golubnitschaja (2020) confirmed from their research in Tunisia that reliable “realtime” monitoring based on randomised laboratory tests is the optimal predictive strategy for
evidence-based preventive measures. This approach should aid in calibrating policy decisions to avoid long-term economic recession from over-protection or eventually explosive
health risks that can arise from over-relaxation of containment measures and pandemic fatigue. Timeliness and testing are key to disaster and disease governance, as are tracing, transparency, training, transdisciplinary research, and trust-building among the public to own the
recommended containment policy and strategic measures (Adero, 2021).
The critical importance of timing has further been reinforced by Read et al. (2020), who
used a transmission model at the time the novel coronavirus began spreading fast in China to
estimate a basic reproductive number, R0, of 3.11 and further that 58-76% of transmissions
must be prevented to stop the surge. In terms of tracing, actionable, pin-pointed locationAcademia Letters, July 2021
©2021 by the author — Open Access — Distributed under CC BY 4.0
Corresponding Author: Nashon Juma Adero, nashon.adero@ttu.ac.ke
Citation: Adero, N.J. (2021). Lessons from Modelling COVID-19 Scenarios in Kenya and Implications for
Policy and Planning. Academia Letters, Article 1862. https://doi.org/10.20935/AL1862.
2
based results are critical to containing fast-spreading disasters, the case of COVID-19 further
aggravated by random mutations. Such an achievement was demonstrated using geospatial
technologies in Germany to trace the first COVID-19 case to a saltshaker in a Bavarian restaurant.
Africa’s first COVID-19 case was recorded in Egypt on February 14, 2020, making the
first of a series of imported cases through travellers from the hotspots in Europe, Asia, and
the USA. Recently, there have been active community transmissions in Africa, aggravated by
the high population densities and poor living conditions in most African towns. By March 17,
2020, the whole of Africa had confirmed a total of only 450 COVID-19 cases (Worldometer,
2020).
The spread of COVID-19 is a dynamic process, hence the justified application of dynamic
models. Model development is both art and science calibrated and matured over time through
passion and experience. Intervening variables are central to system functioning while moderating variables are mostly attitudinal or cultural factors, which usually end up modifying
system response. The uptake of modelling in the science-society-policy nexus has surged
during the COVID-19 pandemic.
In Kenya, a raft of containment measures since March 2020 included partial lockdowns,
dusk-to-dawn curfew, and restrictions on mass gatherings in public places, restaurants, and
places of worship. Political rallies, known to be among the super-spreader activities, continued
in several places on several occasions despite the ban on gatherings. This realisation was a
weak link in the war against the pandemic.
In terms of data and technical details, the models applied can be mechanistic or statistical
(Bala, Arshad, & Noh, 2017; Bellinger & Fortmann-Roe, 2013; Brailsford & Hilton, 2001;
Kelton & Law, 2000). Based on the writings of these authors, it is evident that system dynamics models are mostly applied to handle interactions in complex adaptive systems towards
understanding the big picture at higher strategic and conceptual levels with policy-oriented
simulations, so as to understand long-term system behaviour over months or years. Discrete
event simulation (DES) models find application in addressing specific questions at the operational and tactical levels over short time scales such as hours or days. Agent-based models are
useful where there is a good understanding of how individual agents influence changes and
respond in a system. Statistical models are data-intensive; they thrive in data abundance to
serve the need to establish a model of dependencies. Again, the taxonomy could be based on
the purpose of the models, hence a further classification into narrative models that are meant
to create a convincing storyline for behaviour change agency, inferential models for hypothesis testing, or predictive models for projections within a specified error margin (Bellinger &
Fortmann-Roe, 2013).
Academia Letters, July 2021
©2021 by the author — Open Access — Distributed under CC BY 4.0
Corresponding Author: Nashon Juma Adero, nashon.adero@ttu.ac.ke
Citation: Adero, N.J. (2021). Lessons from Modelling COVID-19 Scenarios in Kenya and Implications for
Policy and Planning. Academia Letters, Article 1862. https://doi.org/10.20935/AL1862.
3
Methodology
Predictive modelling was chosen to fit the intended purpose of this study. Though the modelling exercise has covered thirty-two countries across the world for comparisons, the scope
of this paper has been limited to Kenya’s COVID-19 curves since March 2020. Informed by
intervening and moderating variables, this research reviewed the success of system dynamics
models, agent-based models, and statistical models in simulating COVID-19 scenarios. Reducing uncertainties into quantifiable risks to inform containment measures within a compact
bandwidth of future scenarios was the underpinning philosophy.
The abundance of cumulative COVID-19 data justified the choice of statistical modelling
based on data-driven mathematical simulations using the best-fitting equations — keeping the
coefficient of determination, R-squared, to at least 0.995. The trend in population-normalised
testing rates, sampling efficacy, community behaviour, infection rates, and the effect of lockdowns were used to generate assumptions for three model scenarios: optimistic, business-asusual (BAU), and pessimistic scenario.
Based on experience in the model development process, the suitable update period for the
models came to be every three to four weeks. Based on nationally reported data from Wordometer (2020), reference to eleven selected countries in Africa[1] revealed leading metrics
on the spread of COVID-19 across Africa by April 16, 2020. The metrics were compared to
the ones derived for 14 selected countries outside Africa with leading metrics by then.
Results and discussion
The key metrics from the country models showed that it was taking an average of 16 days to
start the fast-rising phase of the country COVID-19 curves. This figure was lower than the
average of 33 days this study established for the 14 countries with leading cases in Europe,
Asia, and North America by mid-April 2020. By April 16, 2020, six out of the eleven African
countries had displayed exponential growth phases. The highest exponential rates in the group
were in South Africa (26%), Cameroon (23%), Kenya (18%), and Egypt (14%). The rest of
the countries in the study group displayed simulation curves that fitted a rising quadratic trend
(Niger) or a rising linear trend (Rwanda, Nigeria, Burkina Faso, and Algeria — in the order
of increasing gradients).
The daily average population-normalised testing rate in Kenya, at only 74 tests/million
people/day by May 7, 2021, was still low even in Africa when compared to Ghana’s 83,
Rwanda’s 241, and ranging about 400–500 in Morocco, South Africa, India, and Brazil. The
global leaders on this normalised testing score by May 8, 2021, included the UK (about 5,000),
Academia Letters, July 2021
©2021 by the author — Open Access — Distributed under CC BY 4.0
Corresponding Author: Nashon Juma Adero, nashon.adero@ttu.ac.ke
Citation: Adero, N.J. (2021). Lessons from Modelling COVID-19 Scenarios in Kenya and Implications for
Policy and Planning. Academia Letters, Article 1862. https://doi.org/10.20935/AL1862.
4
Israel (over 3,500), and the USA (about 2,900). Country comparisons are shown in Figure 1.
Figure 1: Daily average population-normalised testing rates across thirty countries between
July 2020 and May 2021.
A heat map of the shares of COVID-19 cases in May across the 47 counties in Kenya
showed the dominance of the caseload in the main urban centres along the major transport
corridors, the Nairobi City County maintaining a share of 46% of late followed by 7% in
Mombasa County (Figure 2). The metropolitan region deeply and functionally connected to
Nairobi through employment and commerce formed the “COVID-19 ring of fire” in Kenya,
together holding 58% of the country’s caseload. The share rose to 63% with the addition
of Nakuru, a key urban centre connecting the metropolitan region to the western ring. The
centres that hosted major public gatherings, mainly because of political rallies as the country
Academia Letters, July 2021
©2021 by the author — Open Access — Distributed under CC BY 4.0
Corresponding Author: Nashon Juma Adero, nashon.adero@ttu.ac.ke
Citation: Adero, N.J. (2021). Lessons from Modelling COVID-19 Scenarios in Kenya and Implications for
Policy and Planning. Academia Letters, Article 1862. https://doi.org/10.20935/AL1862.
5
was nearing the 2022 general elections, tended to experience a faster and remarkable increase
in the share of confirmed COVID-19 cases. Nakuru, Kisumu, and Kisii were particularly
noted for such remarkable changes.
Figure 2: Heat map of the relative shares of COVID-19 cases by county across Kenya as of
May 20, 2021.
Academia Letters, July 2021
©2021 by the author — Open Access — Distributed under CC BY 4.0
Corresponding Author: Nashon Juma Adero, nashon.adero@ttu.ac.ke
Citation: Adero, N.J. (2021). Lessons from Modelling COVID-19 Scenarios in Kenya and Implications for
Policy and Planning. Academia Letters, Article 1862. https://doi.org/10.20935/AL1862.
6
The models developed have been projecting Kenya’s end-month COVID-19 cases within
0.1%–10% from the BAU and optimistic scenarios for more than 90% of the instances. The
highest difference of 16.7% was realised at the end of August 2020 following a 50% reduction
in the population-normalised testing rates observed after August 16, 2020, and 11.1% at the
end of November 2020, after the surge in October following the reopening on September 28,
2020. Between January 1 and January 14, 2021, the simulation assuming the second wave
would peak on January 14 remained within 4.6–5.5% above the actual reported cases (Figure
3).
Figure 3: Simulated scenarios of COVID-19 cases in Kenya against actual reported cases
between October 1, 2020 and January 31, 2021.
This yearlong modelling and testing of the character of Kenya’s COVID-19 curves simulated the possibility of a fourth wave picking up the pace in July 2021 and the third wave ebbing
from May 26 after reaching a simulated figure of 166,775 cases (+/- 1%). The confirmation
of the first case of the Delta variant in Kisumu, Kenya, in May 2021, however, changed this
Academia Letters, July 2021
©2021 by the author — Open Access — Distributed under CC BY 4.0
Corresponding Author: Nashon Juma Adero, nashon.adero@ttu.ac.ke
Citation: Adero, N.J. (2021). Lessons from Modelling COVID-19 Scenarios in Kenya and Implications for
Policy and Planning. Academia Letters, Article 1862. https://doi.org/10.20935/AL1862.
7
optimism. The new surging trend in the actual curve became evident in the observed change
in the magnitude and sign of differences between the model results and the reported COVID19 cases. Table 1 presents the model equations and the accompanying scenario of differences
between the model and reported cases from April 30 to May 26, 2021. It is evident that the
trend started changing noticeably from May 14, only made more pronounced from May 19,
2021, as the more contagious Delta variant started spreading in Kenya. The displayed differences are between the COVID-19 model results and the actual reported cases before and after
the entry of the Delta variant in May 2021.
Generally, the period the COVID waves have taken to flatten in Kenya has tended to be
about four months within the study period. The entry of the Indian (Delta) variant reported in
Kenya in May, the recent easing of movement restrictions in the “disease-infected zone” on
May 1, 2021, giving way to more community interactions and school reopening, and the slow
pace of vaccination that cannot assure herd immunity any time soon, implied the possibility
of a more vicious fourth wave surpassing 300,000 cases by July 2021 in Kenya. Avoiding
this pessimistic scenario calls for heightened civic discipline for compliance with COVID-19
health protocols and strict policy enforcement.
Academia Letters, July 2021
©2021 by the author — Open Access — Distributed under CC BY 4.0
Corresponding Author: Nashon Juma Adero, nashon.adero@ttu.ac.ke
Citation: Adero, N.J. (2021). Lessons from Modelling COVID-19 Scenarios in Kenya and Implications for
Policy and Planning. Academia Letters, Article 1862. https://doi.org/10.20935/AL1862.
8
Conclusion
Since the model projections have largely remained within a 10% margin of error, they are
within the boundary recommended for countrywide policy and planning purposes. To effectively contain the pandemic’s resurgent waves in Kenya and attain herd immunity, vaccination
rates, contact tracing and population-normalised testing rates need to be enhanced, together
with timely containment responses informed by transdisciplinary scientific research. Building public trust in all the containment strategies is also crucial, attainable through data transparency and overall data integrity with regular sharing of mapped visual evidence. Applying
geospatial technologies (GIS) promises to enhance visualisation for a shared understanding
and location-based intelligence that can easily be appreciated and acted upon collectively by
the media, decision-makers, health workers, and the general citizenry for effective containment. The disruption COVID-19 has caused goes beyond the health sector to include education and the economy. The main research and policy implications for the future of medical
diagnostics, clinical practice and public health management are: enhanced technology adoption to embrace digitalisation and spatial mapping for precise and effective tracing and disease
governance at the intersection of geography and health (geomedicine), developing integrated
databases of health informatics with spatially referenced variables for timely and sound decision support, enhanced application of predictive models, and generally embracing the rhythm
of Industry 4.0 in healthcare - hence developments in “Health 4.0” complete with increasing
uptake of telemedicine.
References
Adero, N.J. (2021). Multiscale Modelling of COVID-19 Curves: Insights and Implications
for Disaster Governance in Africa. In N. J. Adero, & J. Juma (Eds.), The Future of Africa in
the Post-COVID-19 World (pp. 3-10). Nairobi: Inter Region Economic Network.
Bala, K.N., Arshad, F.M., & Noh, K.M. (2017). System Dynamics – Modelling and Simulation. Singapore: Springer.
Bellinger, G., & Fortmann-Roe, S. (2013). Beyond Connecting the Dots: Modelling for
Meaningful Results. eBook available online.
Brailsford, S.C., & Hilton, N.A. (2001). A comparison of discrete event simulation and
system dynamics for modelling health care systems. In: Riley, J. (ed.) Planning for the
Future: Health Service Quality and Emergency Accessibility. Operational Research Applied
to Health Services (ORAHS). Glasgow Caledonian University.
Chaari, L., & Golubnitschaja, O. (2020). Covid-19 pandemic by the “real-time” moniAcademia Letters, July 2021
©2021 by the author — Open Access — Distributed under CC BY 4.0
Corresponding Author: Nashon Juma Adero, nashon.adero@ttu.ac.ke
Citation: Adero, N.J. (2021). Lessons from Modelling COVID-19 Scenarios in Kenya and Implications for
Policy and Planning. Academia Letters, Article 1862. https://doi.org/10.20935/AL1862.
9
toring: the Tunisian case and lessons for global epidemics in the context of 3PM strategies.
EPMA J. 25;11(2):1–6. doi: 10.1007/s13167–020–00207–0.
Kelton, W.D., & Law, A.M. (2000). Simulation Modeling and Analysis. Boston, MA,
USA: McGraw Hill.
Read, J.M., Bridgen, J. RE., Cummings, D. AT., Ho, A., & Jewell, C.P. (2020). Novel
coronavirus 2019-nCoV: early estimation of epidemiological parameters and epidemic predictions. COVID-19 SARS-CoV-2 preprints from medRxiv and bioRxiv. doi: https://doi.org/
10.1101/2020.01.23.20018549
Republic of Kenya — Ministry of Health (2021). Daily COVID-19 updates. https://www.health.go.ke/.
World Health Organization (WHO) 2020. Novel Coronavirus (2019-nCoV) situation reports January — July 2020. Accessible from: https://www.who.int/emergencies/diseases/
novel-coronavirus-2019/situation-reports/.
Worldometer (2020). Online portal for COVID Live Update. https://www.worldometers.info/coronavirus/.
Worldometer (2021). Online portal for COVID Live Update. https://www.worldometers.info/coronavirus/.
[1] Egypt, South Africa, Algeria, Cameroon, Ghana, Ivory Coast, Niger, Burkina Faso,
Nigeria, Kenya, and Rwanda. Data from Worldometer (online).
Academia Letters, July 2021
©2021 by the author — Open Access — Distributed under CC BY 4.0
Corresponding Author: Nashon Juma Adero, nashon.adero@ttu.ac.ke
Citation: Adero, N.J. (2021). Lessons from Modelling COVID-19 Scenarios in Kenya and Implications for
Policy and Planning. Academia Letters, Article 1862. https://doi.org/10.20935/AL1862.
10
ACADEMIA Letters
Lessons from Modelling COVID-19 Scenarios in Kenya
and Implications for Policy and Planning
Nashon Juma Adero, Taita Taveta University
Abstract
This study examined and modelled the cross-country spread of the novel coronavirus disease
of 2019 (COVID-19) in Africa and outside Africa, finally converging on Kenya as the country of interest. A review of the models widely used to calibrate policy and strategic responses
confirmed the suitability of statistical models for predicting the spread of COVID-19, within a
10% margin of error. The main objective was to provide insights into the spread of COVID-19
and present “what-if” scenarios in aid of policy simulations within a compact bandwidth of
future scenarios, as required for policy and planning. The models proved resourceful in predicting the end-month cases within 10% and the likelihood of subsequent waves, both in terms
of the magnitudes and the time of starting and flattening. In the case of Kenya, the period of
the waves tended to be approximately four months within the study period presented here,
from April 2020 to May 2021. The key lessons imply a greater role of modelling, digitalisation, geospatial and mapping technologies, and transdisciplinary research in the aspirational
future of disease and disaster governance.
Keywords: digitalisation, disease and disaster governance, geomedicine, Health 4.0, policy simulation
Academia Letters, July 2021
©2021 by the author — Open Access — Distributed under CC BY 4.0
Corresponding Author: Nashon Juma Adero, nashon.adero@ttu.ac.ke
Citation: Adero, N.J. (2021). Lessons from Modelling COVID-19 Scenarios in Kenya and Implications for
Policy and Planning. Academia Letters, Article 1862. https://doi.org/10.20935/AL1862.
1
Introduction
In recent years, disease and disaster governance has become a public policy concern of borderless geopolitical implications. The novel coronavirus disease (COVID-19), first reported
in China in December 2019, epitomises this scenario. These disruptive developments have
accelerated the pace of the Fourth Industrial Revolution (Industry 4.0), expected to be more
pronounced in the health and education sectors with increasing uptake of digitalisation and
immersive technologies.
Since January 2020, publicly available COVID-19 data has been increasing, from barely
a million cases at the beginning of April 2020 to more than 157 million cases by May 7, 2021
(Republic of Kenya [Ministry of Health, 2021]; WHO, 2020; Worldometer, 2020; Worldometer, 2021). The increasing COVID-19 data volume has enabled the development of multi-scale
data-driven models to inform critical trends and containment policies globally, across Africa,
and within Kenya. The World Health Organization (2020) has since confirmed community
transmission as the dominant form of spread of COVID-19, Africa not spared the brunt.
This paper illustrates the critical role of data-driven models in the science-policy interface, using mathematical models to calibrate policies and strategies towards influencing behaviour change and containment action. The rest of the paper details the recent developments
in COVID-19 containment, the modelling theory and methodology used in the study, the results obtained, and the implications of the findings for containing the pandemic in Kenya with
respect to policy and planning.
Recent developments in understanding and containing COVID-19
Chaari and Golubnitschaja (2020) confirmed from their research in Tunisia that reliable “realtime” monitoring based on randomised laboratory tests is the optimal predictive strategy for
evidence-based preventive measures. This approach should aid in calibrating policy decisions to avoid long-term economic recession from over-protection or eventually explosive
health risks that can arise from over-relaxation of containment measures and pandemic fatigue. Timeliness and testing are key to disaster and disease governance, as are tracing, transparency, training, transdisciplinary research, and trust-building among the public to own the
recommended containment policy and strategic measures (Adero, 2021).
The critical importance of timing has further been reinforced by Read et al. (2020), who
used a transmission model at the time the novel coronavirus began spreading fast in China to
estimate a basic reproductive number, R0, of 3.11 and further that 58-76% of transmissions
must be prevented to stop the surge. In terms of tracing, actionable, pin-pointed locationAcademia Letters, July 2021
©2021 by the author — Open Access — Distributed under CC BY 4.0
Corresponding Author: Nashon Juma Adero, nashon.adero@ttu.ac.ke
Citation: Adero, N.J. (2021). Lessons from Modelling COVID-19 Scenarios in Kenya and Implications for
Policy and Planning. Academia Letters, Article 1862. https://doi.org/10.20935/AL1862.
2
based results are critical to containing fast-spreading disasters, the case of COVID-19 further
aggravated by random mutations. Such an achievement was demonstrated using geospatial
technologies in Germany to trace the first COVID-19 case to a saltshaker in a Bavarian restaurant.
Africa’s first COVID-19 case was recorded in Egypt on February 14, 2020, making the
first of a series of imported cases through travellers from the hotspots in Europe, Asia, and
the USA. Recently, there have been active community transmissions in Africa, aggravated by
the high population densities and poor living conditions in most African towns. By March 17,
2020, the whole of Africa had confirmed a total of only 450 COVID-19 cases (Worldometer,
2020).
The spread of COVID-19 is a dynamic process, hence the justified application of dynamic
models. Model development is both art and science calibrated and matured over time through
passion and experience. Intervening variables are central to system functioning while moderating variables are mostly attitudinal or cultural factors, which usually end up modifying
system response. The uptake of modelling in the science-society-policy nexus has surged
during the COVID-19 pandemic.
In Kenya, a raft of containment measures since March 2020 included partial lockdowns,
dusk-to-dawn curfew, and restrictions on mass gatherings in public places, restaurants, and
places of worship. Political rallies, known to be among the super-spreader activities, continued
in several places on several occasions despite the ban on gatherings. This realisation was a
weak link in the war against the pandemic.
In terms of data and technical details, the models applied can be mechanistic or statistical
(Bala, Arshad, & Noh, 2017; Bellinger & Fortmann-Roe, 2013; Brailsford & Hilton, 2001;
Kelton & Law, 2000). Based on the writings of these authors, it is evident that system dynamics models are mostly applied to handle interactions in complex adaptive systems towards
understanding the big picture at higher strategic and conceptual levels with policy-oriented
simulations, so as to understand long-term system behaviour over months or years. Discrete
event simulation (DES) models find application in addressing specific questions at the operational and tactical levels over short time scales such as hours or days. Agent-based models are
useful where there is a good understanding of how individual agents influence changes and
respond in a system. Statistical models are data-intensive; they thrive in data abundance to
serve the need to establish a model of dependencies. Again, the taxonomy could be based on
the purpose of the models, hence a further classification into narrative models that are meant
to create a convincing storyline for behaviour change agency, inferential models for hypothesis testing, or predictive models for projections within a specified error margin (Bellinger &
Fortmann-Roe, 2013).
Academia Letters, July 2021
©2021 by the author — Open Access — Distributed under CC BY 4.0
Corresponding Author: Nashon Juma Adero, nashon.adero@ttu.ac.ke
Citation: Adero, N.J. (2021). Lessons from Modelling COVID-19 Scenarios in Kenya and Implications for
Policy and Planning. Academia Letters, Article 1862. https://doi.org/10.20935/AL1862.
3
Methodology
Predictive modelling was chosen to fit the intended purpose of this study. Though the modelling exercise has covered thirty-two countries across the world for comparisons, the scope
of this paper has been limited to Kenya’s COVID-19 curves since March 2020. Informed by
intervening and moderating variables, this research reviewed the success of system dynamics
models, agent-based models, and statistical models in simulating COVID-19 scenarios. Reducing uncertainties into quantifiable risks to inform containment measures within a compact
bandwidth of future scenarios was the underpinning philosophy.
The abundance of cumulative COVID-19 data justified the choice of statistical modelling
based on data-driven mathematical simulations using the best-fitting equations — keeping the
coefficient of determination, R-squared, to at least 0.995. The trend in population-normalised
testing rates, sampling efficacy, community behaviour, infection rates, and the effect of lockdowns were used to generate assumptions for three model scenarios: optimistic, business-asusual (BAU), and pessimistic scenario.
Based on experience in the model development process, the suitable update period for the
models came to be every three to four weeks. Based on nationally reported data from Wordometer (2020), reference to eleven selected countries in Africa[1] revealed leading metrics
on the spread of COVID-19 across Africa by April 16, 2020. The metrics were compared to
the ones derived for 14 selected countries outside Africa with leading metrics by then.
Results and discussion
The key metrics from the country models showed that it was taking an average of 16 days to
start the fast-rising phase of the country COVID-19 curves. This figure was lower than the
average of 33 days this study established for the 14 countries with leading cases in Europe,
Asia, and North America by mid-April 2020. By April 16, 2020, six out of the eleven African
countries had displayed exponential growth phases. The highest exponential rates in the group
were in South Africa (26%), Cameroon (23%), Kenya (18%), and Egypt (14%). The rest of
the countries in the study group displayed simulation curves that fitted a rising quadratic trend
(Niger) or a rising linear trend (Rwanda, Nigeria, Burkina Faso, and Algeria — in the order
of increasing gradients).
The daily average population-normalised testing rate in Kenya, at only 74 tests/million
people/day by May 7, 2021, was still low even in Africa when compared to Ghana’s 83,
Rwanda’s 241, and ranging about 400–500 in Morocco, South Africa, India, and Brazil. The
global leaders on this normalised testing score by May 8, 2021, included the UK (about 5,000),
Academia Letters, July 2021
©2021 by the author — Open Access — Distributed under CC BY 4.0
Corresponding Author: Nashon Juma Adero, nashon.adero@ttu.ac.ke
Citation: Adero, N.J. (2021). Lessons from Modelling COVID-19 Scenarios in Kenya and Implications for
Policy and Planning. Academia Letters, Article 1862. https://doi.org/10.20935/AL1862.
4
Israel (over 3,500), and the USA (about 2,900). Country comparisons are shown in Figure 1.
Figure 1: Daily average population-normalised testing rates across thirty countries between
July 2020 and May 2021.
A heat map of the shares of COVID-19 cases in May across the 47 counties in Kenya
showed the dominance of the caseload in the main urban centres along the major transport
corridors, the Nairobi City County maintaining a share of 46% of late followed by 7% in
Mombasa County (Figure 2). The metropolitan region deeply and functionally connected to
Nairobi through employment and commerce formed the “COVID-19 ring of fire” in Kenya,
together holding 58% of the country’s caseload. The share rose to 63% with the addition
of Nakuru, a key urban centre connecting the metropolitan region to the western ring. The
centres that hosted major public gatherings, mainly because of political rallies as the country
Academia Letters, July 2021
©2021 by the author — Open Access — Distributed under CC BY 4.0
Corresponding Author: Nashon Juma Adero, nashon.adero@ttu.ac.ke
Citation: Adero, N.J. (2021). Lessons from Modelling COVID-19 Scenarios in Kenya and Implications for
Policy and Planning. Academia Letters, Article 1862. https://doi.org/10.20935/AL1862.
5
was nearing the 2022 general elections, tended to experience a faster and remarkable increase
in the share of confirmed COVID-19 cases. Nakuru, Kisumu, and Kisii were particularly
noted for such remarkable changes.
Figure 2: Heat map of the relative shares of COVID-19 cases by county across Kenya as of
May 20, 2021.
Academia Letters, July 2021
©2021 by the author — Open Access — Distributed under CC BY 4.0
Corresponding Author: Nashon Juma Adero, nashon.adero@ttu.ac.ke
Citation: Adero, N.J. (2021). Lessons from Modelling COVID-19 Scenarios in Kenya and Implications for
Policy and Planning. Academia Letters, Article 1862. https://doi.org/10.20935/AL1862.
6
The models developed have been projecting Kenya’s end-month COVID-19 cases within
0.1%–10% from the BAU and optimistic scenarios for more than 90% of the instances. The
highest difference of 16.7% was realised at the end of August 2020 following a 50% reduction
in the population-normalised testing rates observed after August 16, 2020, and 11.1% at the
end of November 2020, after the surge in October following the reopening on September 28,
2020. Between January 1 and January 14, 2021, the simulation assuming the second wave
would peak on January 14 remained within 4.6–5.5% above the actual reported cases (Figure
3).
Figure 3: Simulated scenarios of COVID-19 cases in Kenya against actual reported cases
between October 1, 2020 and January 31, 2021.
This yearlong modelling and testing of the character of Kenya’s COVID-19 curves simulated the possibility of a fourth wave picking up the pace in July 2021 and the third wave ebbing
from May 26 after reaching a simulated figure of 166,775 cases (+/- 1%). The confirmation
of the first case of the Delta variant in Kisumu, Kenya, in May 2021, however, changed this
Academia Letters, July 2021
©2021 by the author — Open Access — Distributed under CC BY 4.0
Corresponding Author: Nashon Juma Adero, nashon.adero@ttu.ac.ke
Citation: Adero, N.J. (2021). Lessons from Modelling COVID-19 Scenarios in Kenya and Implications for
Policy and Planning. Academia Letters, Article 1862. https://doi.org/10.20935/AL1862.
7
optimism. The new surging trend in the actual curve became evident in the observed change
in the magnitude and sign of differences between the model results and the reported COVID19 cases. Table 1 presents the model equations and the accompanying scenario of differences
between the model and reported cases from April 30 to May 26, 2021. It is evident that the
trend started changing noticeably from May 14, only made more pronounced from May 19,
2021, as the more contagious Delta variant started spreading in Kenya. The displayed differences are between the COVID-19 model results and the actual reported cases before and after
the entry of the Delta variant in May 2021.
Generally, the period the COVID waves have taken to flatten in Kenya has tended to be
about four months within the study period. The entry of the Indian (Delta) variant reported in
Kenya in May, the recent easing of movement restrictions in the “disease-infected zone” on
May 1, 2021, giving way to more community interactions and school reopening, and the slow
pace of vaccination that cannot assure herd immunity any time soon, implied the possibility
of a more vicious fourth wave surpassing 300,000 cases by July 2021 in Kenya. Avoiding
this pessimistic scenario calls for heightened civic discipline for compliance with COVID-19
health protocols and strict policy enforcement.
Academia Letters, July 2021
©2021 by the author — Open Access — Distributed under CC BY 4.0
Corresponding Author: Nashon Juma Adero, nashon.adero@ttu.ac.ke
Citation: Adero, N.J. (2021). Lessons from Modelling COVID-19 Scenarios in Kenya and Implications for
Policy and Planning. Academia Letters, Article 1862. https://doi.org/10.20935/AL1862.
8
Conclusion
Since the model projections have largely remained within a 10% margin of error, they are
within the boundary recommended for countrywide policy and planning purposes. To effectively contain the pandemic’s resurgent waves in Kenya and attain herd immunity, vaccination
rates, contact tracing and population-normalised testing rates need to be enhanced, together
with timely containment responses informed by transdisciplinary scientific research. Building public trust in all the containment strategies is also crucial, attainable through data transparency and overall data integrity with regular sharing of mapped visual evidence. Applying
geospatial technologies (GIS) promises to enhance visualisation for a shared understanding
and location-based intelligence that can easily be appreciated and acted upon collectively by
the media, decision-makers, health workers, and the general citizenry for effective containment. The disruption COVID-19 has caused goes beyond the health sector to include education and the economy. The main research and policy implications for the future of medical
diagnostics, clinical practice and public health management are: enhanced technology adoption to embrace digitalisation and spatial mapping for precise and effective tracing and disease
governance at the intersection of geography and health (geomedicine), developing integrated
databases of health informatics with spatially referenced variables for timely and sound decision support, enhanced application of predictive models, and generally embracing the rhythm
of Industry 4.0 in healthcare - hence developments in “Health 4.0” complete with increasing
uptake of telemedicine.
References
Adero, N.J. (2021). Multiscale Modelling of COVID-19 Curves: Insights and Implications
for Disaster Governance in Africa. In N. J. Adero, & J. Juma (Eds.), The Future of Africa in
the Post-COVID-19 World (pp. 3-10). Nairobi: Inter Region Economic Network.
Bala, K.N., Arshad, F.M., & Noh, K.M. (2017). System Dynamics – Modelling and Simulation. Singapore: Springer.
Bellinger, G., & Fortmann-Roe, S. (2013). Beyond Connecting the Dots: Modelling for
Meaningful Results. eBook available online.
Brailsford, S.C., & Hilton, N.A. (2001). A comparison of discrete event simulation and
system dynamics for modelling health care systems. In: Riley, J. (ed.) Planning for the
Future: Health Service Quality and Emergency Accessibility. Operational Research Applied
to Health Services (ORAHS). Glasgow Caledonian University.
Chaari, L., & Golubnitschaja, O. (2020). Covid-19 pandemic by the “real-time” moniAcademia Letters, July 2021
©2021 by the author — Open Access — Distributed under CC BY 4.0
Corresponding Author: Nashon Juma Adero, nashon.adero@ttu.ac.ke
Citation: Adero, N.J. (2021). Lessons from Modelling COVID-19 Scenarios in Kenya and Implications for
Policy and Planning. Academia Letters, Article 1862. https://doi.org/10.20935/AL1862.
9
toring: the Tunisian case and lessons for global epidemics in the context of 3PM strategies.
EPMA J. 25;11(2):1–6. doi: 10.1007/s13167–020–00207–0.
Kelton, W.D., & Law, A.M. (2000). Simulation Modeling and Analysis. Boston, MA,
USA: McGraw Hill.
Read, J.M., Bridgen, J. RE., Cummings, D. AT., Ho, A., & Jewell, C.P. (2020). Novel
coronavirus 2019-nCoV: early estimation of epidemiological parameters and epidemic predictions. COVID-19 SARS-CoV-2 preprints from medRxiv and bioRxiv. doi: https://doi.org/
10.1101/2020.01.23.20018549
Republic of Kenya — Ministry of Health (2021). Daily COVID-19 updates. https://www.health.go.ke/.
World Health Organization (WHO) 2020. Novel Coronavirus (2019-nCoV) situation reports January — July 2020. Accessible from: https://www.who.int/emergencies/diseases/
novel-coronavirus-2019/situation-reports/.
Worldometer (2020). Online portal for COVID Live Update. https://www.worldometers.info/coronavirus/.
Worldometer (2021). Online portal for COVID Live Update. https://www.worldometers.info/coronavirus/.
[1] Egypt, South Africa, Algeria, Cameroon, Ghana, Ivory Coast, Niger, Burkina Faso,
Nigeria, Kenya, and Rwanda. Data from Worldometer (online).
Academia Letters, July 2021
©2021 by the author — Open Access — Distributed under CC BY 4.0
Corresponding Author: Nashon Juma Adero, nashon.adero@ttu.ac.ke
Citation: Adero, N.J. (2021). Lessons from Modelling COVID-19 Scenarios in Kenya and Implications for
Policy and Planning. Academia Letters, Article 1862. https://doi.org/10.20935/AL1862.
10