Adjusted projections for COVID-19 in Kenya

Adjusted projections for COVID-19 in Kenya

Update 9th July 2020 | Area: Nairobi County

Summary: Our adjusted model suggests 1.4m people in Nairobi County will be infected with COVID-19 during its life-cycle peaking in September 2020. According to the adjusted model, 102,742 will show symptoms. 8,755 people will need hospitalization over the next year, with 1,505 passing through ICU facilities. 928 will die.

Infections and immunity status

Infection and disease

Prevalence of infections by age groups


Throughout the COVID-19 pandemic, we have shared projections as to the impact and severity of the virus and gaps in healthcare capacity in Kenya. The goal of these models is to help prepare our fleet of first responders and help other frontline health workers prepare for potential outcomes and needed capacity.

One key piece of feedback we have received is that our previous model has not been localised to Kenya. This is definitely true as we used standard global COVID-19 models readily available at the time. Indeed the virus has not played out in Kenya in the way projected by global models and research based on its spread in China, Europe, and North America.

For our projections, we used open-sourced epi-models which we tailored with the best available Kenya-specific information. Three things have now changed that warrant revisiting this model and publishing an update: 

  1. There is now substantive peer-reviewed evidence as to the localised factors in Sub-Saharan Africa which have/will impact the severity and conditions of COVID-19 in countries like Kenya.
  2. On 7th July 2020, Kenya announced an end to the partial lockdown of the country and the ban on inter-county travel. This, coupled with the clear increase in confirmed cases of COVID-19 in recent days in the country suggests that the country may be entering a new phase of local transmission, it is the right time to revisit our model and address concerns. 
  3. There are now significantly better open-sourced models we can tailor to Kenya. These models take important localised information – like age breakdowns – into account before projecting expected infections.

In light of these three things, we are revisiting our COVID-19 projections. This time, we are including unique observations and conditions observed in Kenya specifically. These include:

  1. Age distribution: Our initial projection did not take into account demographics within Kenya, it has since become clear that age is a critical factor in determining mortality outcomes during CV-19. Kenya’s young population needs factoring into our model.
  2. Higher % of Asymptomatic patients: The Kenyan Ministry of Health has been reporting an average of 90% asymptomatic cases in recent weeks. This is significantly higher than what has been observed in other countries and clearly needs to be factored into our model. By comparison in an analysis of all confirmed cases across the EU, 25-50% of cases were found to be asymptomatic or minimally symptomatic.
  3. Lower CFR: The observed CFR (Case Fatality Rate, or the percentage of cases that lead to death) in Kenya is currently around 2%. This may partly be explained by the lower testing levels than in other countries, but the finding has been consistent over several weeks. It is also consistent with the high asymptomatic rate and younger population in Kenya. According to the WHO globally, about 3.4% of reported COVID-19 cases have died with others estimating the global CFR may be as high as 7%.

In order to make our findings as precise as possible, we will be limiting our analysis at this time to Nairobi County, where there are already medium-high levels of community transmission and where the majority (47%) of cases have been found.

Our scenario

With the lifting of lockdown measures, we expect inter-county transmission of the virus to increase but not to significantly increase the transmission of CV19 into Nairobi. This is because cases outside of Nairobi have remained low throughout the previous two weeks. Whilst many are expected to migrate into Nairobi over the coming weeks, it is assumed that few will be carriers of the virus.

We believe that the government will enforce continued curfews and/or restrictions on social contact for a further 100 days, before increased pressure and economic necessities reduce the measures. This is akin to what we have seen in a broad-mix of countries which have imposed lockdown restrictions on their populations.

Model implications

Kenya moved quickly to implement lockdown measures that significantly reduced local transmission of COVID-19. These measures slowed the spread of the virus, giving healthcare workers additional time to prepare for the outbreak. They are a testament to the country’s capacity to implement positive and far-reaching measures to combat the virus.

Some of these measures are now coming to an end and we expect community transmission to increase throughout the country. Nairobi, as an urban area, is in a weak position to stop the spread of the virus: population density is high and economic necessity means that social distancing will become increasingly challenging over the course of the next few weeks. In light of this, our model suggests that community transmission will rapidly increase and that a large number of Nairobi County residents will be infected from now until November 2020. It is important to note that we do think conditions like seasonality, weather or temperature will only play a very marginal role in the virality of COVID-19 in Kenya. Evidence to support such theories is rudimentary and inconclusive.

Despite this, The young population of Nairobi County, high asymptomatic rate and low case-fatality rates suggest that the country may not see the same level of mortality and hospitalisations seen elsewhere. Nevertheless, we believe that Nairobi County hospitals should be prepared for a significant increase in hospitalisations over the coming weeks. At its peak there may be as many as 3,358 people hospitalised across Nairobi with COVID-19 and up to 404 ICU beds should be made available. Nairobi currently only has 278 beds. We believe there is an urgent need to increase the number of ICU beds in the city.

These projections estimate that 928 Nairobi County residents will die from COVID-19 over the next year with the majority of fatalities coming before the new year. This mortality rate is based on a number of assumptions, which still lack important Kenya-specific data points that we will continue to collect to refine our model.   

To access the full logs of this model, please click here

Model Documentation

Disease parameters

Population size

We shall be using a total population size for Nairobi of 4,397,037 people (2019 census).


We have applied specific Kenyan demographics to this projection. This is based on the 2019 census of counties in Kenya. We have broken age groups into 20 5-year groups with a 100+ age bracket.

Initial infections

We will be using the confirmed cases of COVID-19 in Kenya as reported by the Ministry of Health on July 4th 2020, 3,968

Infections from outside of the population (per day)

Since the first case of COVID-19 was confirmed in Kenya, the government restricted movement within the country, putting in place a lockdown on inter-county travel and on international-domestic flights a few weeks later. This should have effectively slowed the rate of inter-county transmission. As of July 6th 2020, only 1,800 of the total 8,000 cases were outside of Nairobi & Mombasa.

It is hard to speculate exactly how many infected people will travel from outside Nairobi into the County following the lowering of lock-down restrictions. According to the Surgo Foundation, all the surrounding Counties to Nairobi have a high mobility index and we can assume that a significant percentage of people will begin travelling into and out of Nairobi County over the next few weeks. At a conservative estimate, we have assumed that at least 10% of the people travelling into Nairobi each month will be infected, which at current rates equates to 360 cases per month.

Another important dimension is international arrivals into Nairobi, particularly from infected countries such as the UK and US. Such travel will be made available from August 2020. Whilst tourism figures are expected to be significantly lower than the 119,670 who travelled to the country in January, we can project based on other tourism markets which have reopened in recent weeks that tourism arrivals will be at 5% of pre-COVID levels: 5,983 tourist arriving into Nairobi each month. Based on randomised studies of infection rates in several of the countries who supply the majority of tourists to Kenya, we can assume that 15% of new arrivals will carry the virus. This equates to 897 new infections arriving each month from overseas.

Combined (domestic and international), our estimate is that 40 infected individuals will travel into Nairobi each day as a result of lifting the lockdown measures.

It is important to note that this will be significantly higher than the number of confirmed imported cases into Nairobi to date (as of 6th of July 2020), listed as 170, a rate of less than 2 per day.


Simulation duration

We have simulated for 365 days which is more than enough to cover the expected peak and gradual reduction of infection rates. Projections over 1 year have tended to rely on increasingly large assumptions and may be more inaccurate than those over shorter time periods.

Transmission, progression, symptom onset and recovery duration of the virus

To assess the progression durations of COVID-19 we used the analysis of available literature compiled in a widely cited study of the virus. This analysis collates research from a wide range of geographies and regions. There is no current evidence that COVID-19 in Kenya has different timings for transmission, progression, symptom onset and recovery than cases observed in these studies.

Hospitalisation (days)

Studies of the total number of days that people need to be hospitalised for COVID-19 range significantly around the world. An aggregated view of 51 studies of hospitalisation found that the median length of stay ranged from 4 to 53 days within China, and 4 to 21 days outside of China.

No study of Kenyan hospital length of stay has yet been published so this information is not yet available to us. Therefore we have made a hypothesis. Due to the potentially low level of hospitalisation and the severity of cases that end up being hospitalised we expect the length of stays at a hospital to average towards the upper bound of this range and have settled on 16 days. We will continue to adjust these figures as new evidence becomes available.

ICU admission (days)

An analysis of ICU length of stay from the same analysis of global COVID-19 research  found a wide discrepancy. Length of stay was reported by eight studies – four within and four outside China – with median values ranging from 6 to 12 and 4 to 19 days, respectively. 

Again, due to the lack of analysis so far in Kenya we need to make a hypothesis. Due to the potentially low level of hospitalisation and the severity of cases that end up being hospitalised we expect ICU length of stays to average towards the upper bound of this range and have settled on 10 days.

Severity by age group

% of infections that will lead to sickness

This is a key input in our model and is significantly different in Kenya than in other countries. The % of COVID-19 tests that have come back positive and asymptomatic are extremely high in Kenya. In the 4 weeks proceeding this projection, the % of asymptomatic cases was 88%, 88%, 92% and 93% respectively. That is an average of 90% asymptomatic cases. By comparison, in an analysis of all confirmed cases across the EU, 25-50% of cases were found to be asymptomatic or minimally symptomatic.

It is important to note that reported asymptomatic cases in Kenya are only an analysis of tested cases where the patient submits themselves voluntarily for testing, rather than randomized sampling. This means that we actually expect the 90% of cases that are asymptomatic is an underestimate of the true relationship. There are likely many more asymptomatic cases that don’t get tested which would be shown in more randomised tests. 

Given this we have used a flat 10% rate for this input. We have kept this consistent as there is currently no evidence to suggest that asymptomatic rates vary between ages in Kenya.

Sick patients who seek medical help

This determines what % of sick cases go to the doctor to seek medical help. Amongst studies of influenza viruses in high-income countries, this has tended to range from 60-80%.

There are two key factors that suggest that this rate will be significantly lower in Kenya than in many observed countries:

  1. COVID-19 treatments are largely uninsured and there is a perception that healthcare is expensive and inaccessible. Average medical savings are $5 in Kenya and it is expected that many will forgo treatment unless their case is severe. According to a national household survey released by Kenya National Bureau of Statistics, 57% of the population seeks treatment from traditional healers and herbalists compared to 28% who access care from health facilities.
  2. There is widespread fear of testing positive for COVID-19 in Kenya as positive cases are isolated at Government facilities. Few patients want to be isolated in this manner and we expect a high level of avoidance.

An analysis of the distribution of presenting symptoms amongst COVID-19 patients shows that 28% exhibit difficulty in breathing with 42% showing signs of fever. We expect both these groups to potentially need/want to seek medical help. However, given the financial burden of COVID-19 treatment, we have set our expectation that 20%. This implies that only 20% of all sick CV-19 patients will seek out medical help if they are under 50 years old and 30% will seek out medical help if they are above 50 years old.

% of those who seek medical help who are hospitalised

A widely cited study of Hospitalisations and ICU admissions around the world by Imperial College London has looked at the percentages of symptomatic COVID-19 cases that actively seek medical help and who are subsequently hospitalised. Crucially they provide an age breakdown which helps us apply this to Kenya’s specific demographics: 

Age groupSymptomatic hospitalisation rate
0-9 years0.1%
10-19 years0.3%
20-29 years1.2%
30-39 years3.2%
40-49 years4.9%
50-59 years10.2%
60-69 years16.6%
70-79 years24.3%
80+ years27.3%

There is no evidence to suggest that hospitalisation rates are different in the UK to those seen in other countries. Therefore we have adopted the above in its entirety in our model.

Hospitalised cases that need intensive care (ICU)

A widely cited study of Hospitalisations and ICU admissions around the world by Imperial College London shows that of those taken to hospital for COVID-19 only a certain percentage will require ICU critical care. These are provided on an age basis:

Age groupHospitalistion to ICU rate
0-9 years5%
10-19 years5%
20-29 years5%
30-39 years5%
40-49 years6.3%
50-59 years12.2%
60-69 years27.2%
70-79 years43.2%
80+ years70.9%

There is no evidence to suggest that hospitalisation rates are different in the UK to those seen in other countries. Therefore we have adopted the above in its entirety in our model.

Sick patients who die from COVID-19

The same UK study provides age breakdowns which are widely accepted in epi-modelling.

Age groupHospitalistion to ICU rate
0-9 years0.002%
10-19 years0.006%
20-29 years0.03%
30-39 years0.08%
40-49 years0.15%
50-59 years0.6%
60-69 years2.2%
70-79 years5.1%
80+ years9.3%

There is no evidence to suggest that hospitalisation rates are different in the UK to those seen in other countries. Therefore we have adopted the above in its entirety in our model.


Amplitude of the seasonal fluctuation of the R number

A study estimating the basic case reproduction number (R0, a measure of the rate of infection) across provinces in China found no correlation with temperature or humidity. This finding is consistent with a further study of 224 cities across China, which observed no clear link between ambient temperature and rates of new infections. However, there is some preliminary data that seasonality may have as much as a 15% amplitude effect on the R number of COVID-19 we have decided to implement a small seasonal fluctuation amplitude of 8%.

Case isolation

Probability that a sick patient is isolated

A key factor when determining local transmission is the probability that sick patients are able to remain isolated. At the moment 100% of patients confirmed to have COVID-19 with symptomatic symptoms are isolated in Government facilities and the remainder are encouraged to isolate at home. We expect this to continue for the duration of the virus.

Maximum capacity of isolation wards (per 10,000 people)

Our best estimation is that there are 278 ICU beds in Nairobi, suggesting that there are 0.6 ICU beds per 10,000 people in Nairobi. 

Contact reduction for cases in home isolation

According to the ministry of Health patients assessed by a Health Care Worker who has a confirmed COVID-19 case and are asymptomatic, free of co-morbidities and have access to a suitable space are encouraged to home isolate for 14 days. Based on analysis of contact reduction for people in home isolation 90% adhere to strict home isolation leading to a 90% reduction in contact reduction.

Contact reduction

Contact reduction duration (days)

This factor determines how long the contact reduction measures put in place in Kenya last. After the end of this period, contacts are set back to the original 100% value. As we have seen from other countries such as the US or UK, contact reduction is only viable for a limited period of time until economic activity is forced to increase. We have only seen a few examples where lockdown has lasted over 100 days and we have used this upper bound as our input for this measure.

Contact reduction in the home (%)

This measure determines what percentage of contacts at home are prevented by social distancing. Unlike in some high-income countries where homes are large enough to facilitate low social contact during an intra-household outbreak, this is not possible in the majority of Kenyan homes. Therefore we have set this reduction at 10%, this is a very loose assumption as there has not been any definitive studies of contact reduction.

Contact reduction in schools (%)

Schools are expected to have a phased reopening from January 2021 in Kenya. This means that over the course of the year we are modeling, we estimate a 100% contact reduction for the first 6 months followed by a gradual increase in contact for the 6 months thereafter.

Contact reduction at work (%)

Small and medium-sized enterprises are the lifeblood of Kenya’s economy, constituting about 98% of all the businesses in the country. Given this we expect low social contact at work to be extremely challenging in Kenya and have set this limit at 50%. Again, this is an assumption that would benefit from empirical analysis.

What do you think?


Your Cart