key: cord-1047378-qla23gq6
authors: Ejigu, B. A.; Asfaw, M. D.; Cavaleri, L.; Abebaw, T.; Nanyingi, M.; Baylis, M.
title: Assessing the impact of non-pharmaceutical interventions on the dynamics of COVID-19: A mathematical modelling study in the case of Ethiopia
date: 2020-11-18
journal: nan
DOI: 10.1101/2020.11.16.20231746
sha: a0b5822ba0f3ee8f5f47a1165aaba9660fd187d9
doc_id: 1047378
cord_uid: qla23gq6

The World Health Organisation (WHO) declared COVID-19 a pandemic on March 11, 2020 and by November 14, 2020 there were 53.3M confirmed cases and 1.3M reported deaths in the world. In the same period, Ethiopia reported 102K cases and 1.5K deaths. Effective public health preparedness and response to COVID-19 requires timely projections of the time and size of the peak of the outbreak. Currently, in the absence of vaccine or effective treatment, the implementation of NPIs (non-pharmaceutical interventions), like hand washing, wearing face coverings or social distancing, is recommended by WHO to bring the pandemic under control. This study proposes a modified Susceptible Exposed Infected and Recovered (SEIR) model to predict the number of COVID-19 cases at different stages of the disease under the implementation of NPIs with different adherence levels in both urban and rural settings of Ethiopia. To estimate the number of cases and their peak time, 30 different scenarios were simulated. The results reveal that the peak time of the pandemic is different in urban and rural populations of Ethiopia. In the urban population, under moderate implementation of three NPIs the pandemic will be expected to reach its peak in December, 2020 with 147,972 cases, of which 18,100 are symptomatic and 957 will require admission to an Intensive Care Unit (ICU). Among the implemented NPIs, increasing the coverage of wearing masks by 10% could reduce the number of new cases on average by one-fifth in urban-populations. Varying the coverage of wearing masks in rural populations minimally reduces the number of cases. In conclusion, the projection result reveals that the projected number of hospital cases is higher than the Ethiopian health system capacity during the peak time. To contain symptomatic and ICU cases within health system capacity, the government should give attention to the strict implementation of the existing NPIs or impose additional public health measures.

time, the most effective means to control the spread of COVID-19 remains the implementation of different NPIs to break chains of transmission Fong et al. (2020b) ; ; Lai et al. (2020) ; Flaxman et al. (2020) ; Ferguson et al. (2020) . The transmission pathways of COVID-19 from person to person are: i) close contact through respiratory droplet, ii) direct contact with infected persons, and iii) contact with contaminated formites (objects and surfaces) Kassa et al. (2020) . Public health measures are intended to diminish these transmission mechanisms.

Mathematical models have been previously used with success in understanding the transmission dynamics and control mechanisms of infectious diseases (Anderson and May (1979) ). To understand the early transmission dynamics of COVID-19 under different scenarios, a number of mathematical models have been proposed in the literature, Prem et al. (2020) ; Li et al. (2020) ; Walker et al. (2020) ; Ivorra et al. (2020) ; Ngonghala et al. (2020) ; Nicholas et al. (2020) . Those epidemiological and mathematical models contributed important insight for public health decisionmakers to enforce different mitigation strategies in different countries. As COVID-19 has spread worldwide, most countries are utilising mathematical models to inform on public health measures. Many countries (for example the UK, China, Germany, USA, Morocco) revised their public health measures based on COVID-19 modelling results. We believe that, due to the difference in the age-structure of the population, social interaction and life style in Ethiopia, mathematical models developed in other countries may not work to study the dynamics of disease in lower income settings. This study employs the use of mathematical models to simulate the spread and interruption of transmission of the disease in Ethiopia.

Assessment of implemented and proposed intervention strategies for combating , and predicting the number of new cases and deaths is a major challenge to both the public and scientific community. Banholzer et al.

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted November 18, 2020. ; https://doi.org/10.1101 https://doi.org/10. /2020 https://doi.org/10. .11.16.20231746 doi: medRxiv preprint (2020 and Flaxman et al. (2020) estimated the impact of NPIs based on confirmed cases of COVID-19 in several countries. Similarly, Ferguson et al. (2020) studied the impact of NPIs to reduce COVID-19 mortality and healthcare demand. To our knowledge, there has been very limited study of the impact of NPIs on the disease dynamics in the local context of Ethiopia. The only published study, done by Siraj et al. (2020) , focused on small, medium and large clusters of the population with social distancing face masking and contact tracing implementation at different proportion. This study did not quantify the impact of hand washing measures on transmission dynamics. Further, their study did not provide the estimated number of cases under different stages of the disease. Takele (2020) implements the Autoregressive Integrated Moving Average (ARIMA) modeling to project COVID-19 prevalence patterns in East Africa Countries, mainly Ethiopia, Djibouti, Sudan and Somalia. While, her projection model did not take into account the impact of NPIs with different adherence level on the predicted number of cases.

Our study proposes a modified mathematical model of the classic SEIR model (Anderson and May (1979) ) and this modified model is used to compare the effect of different NPIs individually and in combination. The proposed model classifies the human population into eight non-overlapping stages of infection: Susceptible (S), Exposed (E), Asymptomatic infectious (Ia), Symptomatic infectious (Is), Isolation of cases with mild/moderate health condition (at home or hospital) (HIm), Hospitalized with critical health condition (Hc), Recovered from the disease (R), and Death due to COVID-19 (D).

Our proposed model (1) has a number of advantages: i) provide estimated time and size of the peak under the implementation of NPIs with different adherence level in urban and rural population settings, ii) proposed model differentiates asymptomatic and symptomatic infectious which influences the number of ICU cases and death due to the disease, iii) the effect of indirect transmission of the disease through contaminated environment is taken into account, and iv) provide the estimated impact of individual and synergy public health measures on the dynamics of the disease. The projected number of people in each stage of the disease at the peak period and the time of the peak in urban and rural areas of Ethiopia helps the government to choose and enforce better intervention mechanisms.

In our modeling framework, the population is divided into different compartments according to the infection status of individuals: susceptible (S), exposed (E), asymptomatic infected (Ia), symptomatic infected (Is), isolated at home or hospital with moderate health condition (IHm), hospitalized with severe health condition (Hc), recovered from the disease (R), and death (D) due to the disease. Contaminated environment is also a means of transmission for COVID-19 as the virus can stay up to several days on different surfaces van Doremalen et al. (2020) . To account for the impact of contaminated objects in the transmission, an additional compartment for contaminated environment is included. Paramter description with these assumed initial values are presented in Table 1 . Those values were extracted from a number of key papers and reports of the situation in Ethiopia (Institute (2020); Goshu et al. (2020) ; Baye (2020) ; Kebede et al. (2020) ; Siraj et al. (2020) ) and epidemiological scientific facts of COVID-19.

By assuming that the total population is N = S + E + Ia + Is + IHm + HC + R + D, the corresponding system 3 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted November 18, 2020. ; https://doi.org/10.1101 https://doi.org/10. /2020 

where the force of infection, γ, is obtained using the following formula:

In Equation (3, β1 and β2 are effective contact rates leading to COVID-19, ηa is relative infectiousness per contact for asymptomatic patients, r4 = (1−cF M ) * (1−SD) * (1−HW ), where SD represents the proportion of the population who practice physical distancing, HW represents the proportion of the population who implements hand washing, F M is coverage of wearing masks, c is efficacy of face mask wearing at reducing transmission, and K is the virus concentration in the environment that yields 50% of chance for a susceptible individual to catch the viral infection 4 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted November 18, 2020. ; https://doi.org/10.1101 https://doi.org/10. /2020 from that source, Kassa et al. (2020) . The model accounts for a distinction between non-diagnosed individuals Ia and Is, who can readily spread the infection because they are not in isolation, and hospitalized individuals HIm and Hc, who transmit the disease less thanks to isolation and complying with strict rules, either in hospital or at home.

To check the mathematical and biological meaningfulness of the model, both positivity and boundedness properties of the proposed model are assessed as follows. Let the domain space be defined by Ω = (S, E, Is, Ia, Hc, IHm, R, D, Env) ∈ R 9 + Theorem 2.1. int(Ω) = {(S, E, Is, Ia, Hc, IHm, R, D, Env)|S > 0, E ≥ 0, Is ≥ 0, Ia ≥ 0,

The proof of theorem 2.1 is presented in the Appendix and the theorem implies well-posedness of the system.

In order to identify parameters which significantly influence the model system, sensitivity analysis was performed.

The uncertainty and sensitivity analysis is done by using Partial Rank Correlation Coefficient (PRCC) analysis with N=10,000 samples for various input parameters Wu et al. (2013) ; Marino et al. (2008) . The partial derivative of the threshold value Ro with respect to the input parameters were computed by varying the parameters around normal values.

Equilibrium points are those points where each equation in the system in Equation (1) is equal to zero. Equating the right hand side of each equation in Equation (1) to zero will provide disease-free and endemic equilibrium points. In this subsection, the disease-free equilibrium point is described as follows.

When there is no COVID-19 in the population, the disease-free equilibrium point is given by E0 = (S, E, Ia, Is, IHm, Hc, R, D, Env) = ( π µ , 0, 0, 0, 0, 0, 0, 0, 0)

In this study, to compute the basic reproduction number, Diekmann et al. (1990) , next-generation method for the disease-free equilibrium is employed. The basic reproduction number is the average number of secondary infection cases produced in a completely susceptible population by a typical infectious individual. According to the definitions stated in Diekmann et al. (1990) and van den Driessche and Watmough (2002a) , in the next-generation method, R0

is the spectral radius of the next-generation matrix which is given by

5 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted November 18, 2020. ; https://doi.org/10. 1101 where F is the Jacobian of the rate of appearance of new infections in the infected compartment, denoted by f and V is the Jacobian of v = v − − v + that represents all infection transfer interactions into (v + ) and out (v − ) of these compartments.

The infection compartments of our model are (E, Ia, Is, IHm, Hc, Env)

Computing the Jacobian of f and evaluating it at the disease-free equilibrium point using the force of infection, we have,

and the Jacobian of v evaluated at the disease free equilibrium point using the force of infection is given by the matrix

Finding the inverse of V and computing the product F V −1 , the next generation matrix is given by 

6 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted November 18, 2020. ; https://doi.org/10. 1101 where for k1 = σ + µ, k2 = ρ + ε + y + µand , k3 = e + α1 + ω + µ.

(9)

The corresponding Spectral radius of the matrix F V −1 is given by

which can be expressed as A = R0 = R0a + R0s + R0Env and the quantity R0 is the basic reproduction number of the Model (2). The quantity R0 is the sum of the constituent reproduction numbers associated with the number of new COVID-19 cases generated by symptomatically infectious humans (R0s), asymptomatically infectious humans (R0a) and contaminated environment (R0E nv ).

Theorem 2.2. The disease-free equilibrium (DFE) of model 2 is locally asymptotically stable if R0 < 1, and unstable if R0 > 1.

Proof. It follows from Theorem 2 of van den Driessche and Watmough (2002b) The implication of Theorem 2.2 is that a small introduction of COVID-19 cases will not generate a COVID-19

outbreak if the basic reproduction number (R0) is less than unity.

Theorem 2.3. The disease-free equilibrium of model 2 is globally asymptotically stable if R0 ≤ 1.

The endemic equilibrium is attained in the presence of the disease, which is denoted by:

7 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted November 18, 2020. ; https://doi.org/10. 1101 and the values of E * , I * s , H * m , H * c , R * and E * nv can be computed from the equations:

From Equation (14), solving for E * , we have

from Equation (15) solving for I * s we have

from Equation (16) , solving for H * m we have

solving for H * c from Equation (17) we have

solving Equation (18) for R * we have

by solving Equation (20) for E * nv , we can get

The proposed modified SEIR model described in section 2.1 is used to project the number of COVID-19 related cases in Ethiopia. Projecting the number of active COVID-19 cases under the implementation of different NPIs with different adherence level is important for policy makers to be able to mitigate against the disease. Projecting the number of asymptomatic and symptomatic people is important, as these two numbers have huge roles in the spread of 8 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted November 18, 2020. ; https://doi.org/10.1101/2020.11.16.20231746 doi: medRxiv preprint the disease. Further, knowing the expected number of people who may require treatment under intensive care (ICU cases) will greatly help to guide policy makers in preparation of manpower and medical facilities such as ventilators.

Due to the very limited number of available ventilators in Ethiopia (below 1000 ventilators in total), knowing the projected number of individuals who need ICU under the implementation of different NPIs with varying adherence level of hygiene, physical distancing and mask wearing are important to save lives. Many studies showed that most of the critically ill patients with COVID-19 are of older age and have more co-morbidities than the non-critically ill patients. Hence, projection is done for all active case, symptomatic cases, asymptomatic cases, ICU cases, and death due to the disease using the proposed model.

In our model, common COVID-19 related parameters were obtained from previous studies (Table 1) . 2

Further, based on the 2016 Ethiopian demographic and health survey, 27.4% and 7.8% of the urban and rural population wash hands by soap, respectively. Thus, in the simulation, we assumed improved percentages of hygiene due to the awareness created by COVID-19 in both population settings. Due to differences in access to sanitation materials, lifestyle, cultural norms and other factors, adherence levels to the recommended NPIs are also different in the urban and rural populations. As a result, projection on the number of COVID-19 related cases is done separately for urban and rural populations of Ethiopia.

In this study, data on daily number of COVID-19 cases, cumulative number of deaths, and number of critical patients was extracted from Ethiopian Public Health Institute website (www.ephi.gov.et/) and Ministry of Health official Twitter page(https://twitter.com/FMoHealth/) on daily basis. We used initial parameter (1) (2020); Kassa et al. (2020) . The projection model considers daily number of cases in Ethiopia up to 210 days (October 08, 2020) since the first case of the disease was recorded.

Based on the current evidence about COVID-19, implementing different NPIs are essential requirements for limiting the spread of the disease. The main objective of NPIs is to reduce the rate of transmission, thereby minimizing the size of the epidemic peak and delaying peak time, buying time for preparations in the healthcare system, and enabling the potential for vaccines and drugs to be developed, approved and used, Table 2 ). We assume that implementing NPIs alone or in combination affects the rate of contact between uninfected people and infected people or objects. As a result, transmission probability of the virus from infected individuals or objects to the susceptible population is reduced.

We evaluated the impact of the three NPIs, alone or in combination, on the time and size of the peak by varying the adherence level to each of them.

2 COVID-19 ICU:32.9% died, 44.1% recovered https://www.bbc.com/news/uk-scotland-52653192 9 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted November 18, 2020. ; https://doi.org/10.1101/2020.11.16.20231746 doi: medRxiv preprint 10 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted November 18, 2020. ; https://doi.org/10.1101/2020.11.16.20231746 doi: medRxiv preprint 3 Results: Projection of number of COVID-19 Cases in Ethiopia

By fixing the adherence levels to hygiene to 30% and 40% of the urban population based on the aforementioned reasons, the projected number of active COVID-19 cases was obtained by varying the percentages of social distancing and face mask coverage. Figure 2 presents the projected number of active COVID-19 cases with varying adherence levels of NPIs.

At 15% face mask coverage and with 10% of the population implementing social distancing measures, a 10%

increase in hygiene will decrease the number of cases by 200,000 and delay the peak time. For a given implementation of wearing a face mask and hygiene, improving the percentage of social distancing by 10% will reduce the number 11 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted November 18, 2020. ; https://doi.org/10.1101/2020.11.16.20231746 doi: medRxiv preprint CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted November 18, 2020. ; https://doi.org/10.1101/2020.11.16.20231746 doi: medRxiv preprint of new COVID-19 cases by one-fifth. Increasing the number of people who wear face mask a by 10% will save on average 25,000 people from being infected by the disease (Figure 2 ).

In the rural parts of Ethiopia, physical distancing is a custom and part of the life style. This custom greatly minimizes the spread of the disease that is caused by contact and contaminated environment. Furthermore, due to low coverage of road accessibility (accessibility index=22%), in our simulation we assumed 20% of the population are Keeping adherence to face mask wearing and physical distancing constant, a 5% increase in hygiene will decrease the number of cases by 200, 000 and shift the peak time to the future nearly by one month (Figure 3) . For a given implementation of face mask wearing and hygiene, improving the percentage of physical distancing by 10% will reduce the number of new COVID-19 cases by one-fourth.

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted November 18, 2020. ; https://doi.org/10.1101/2020.11.16.20231746 doi: medRxiv preprint

The following projected number of symptomatic (broken line) and asymptomatic (solid line) cases were simulated by assuming 30% and 40% of the urban population wash hands with soap, with different face mask coverage (5%, 15%and 25%) with 35% mask efficacy. When 40% of the urban population wash hands with soap or keep hands clean using sanitizers, 25% wear face masks, improving the implementation of physical distancing greatly reduces both symptomatic and asymptomatic infections. By increasing the percentage of physical distancing in the population from 10% to 20%, the peak size will decrease by half for both infection compartments (Figure 4) .

For a given coverage of face mask wearing and implementing physical distancing, improving hygiene by 10%, highly reduces the number of active cases and delays the peak time of the pandemic. For both hygiene scenarios, for a given level of face mask wearing, increasing adherence to physical distancing profoundly reduces the number of symptomatic and asymptomatic cases. A maximum number of asymptomatic infections will be observed if there 14 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted November 18, 2020. ; https://doi.org/10.1101/2020.11.16.20231746 doi: medRxiv preprint is poor implementation of physical distancing and face mask wearing. Regardless of other NPIs, COVID-19 will be suppressed if 40% of the population implements physical distancing properly. Figure 5 presents the projected number of symptomatic (broken line) and asymptomatic (solid line) cases in rural Ethiopia by varying the level of face mask wearing and physical distancing while fixing hygiene practice in the community at 15% and 20%. Increasing proper hand washing behaviour from 15% to 20% could shift the peak by around 2 months. Implementing physical distancing, hygiene, and face mask wearing at 40%, 20% and 5% levels, respectively, helps to control the pandemic under the Ethiopian health system capacity and shifts the peak forward to 2021. At a given level of implementation of hand washing and wearing of face masks, increasing the physical distancing by 10% could shift the peak time by around 40 days ( Figure 5 ).

(a) 15% hygiene and nobody wear a face mask (b) 20% hygiene and nobody wear face mask (c) 15% hygiene and 2% wear a face mask (d) 20% hygiene and 2% wear face mask (e) 15% hygiene and 5% wear a face mask (f) 20% hygiene and 5% wear face mask is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted November 18, 2020. ; https://doi.org/10.1101/2020.11.16.20231746 doi: medRxiv preprint It is not too late to suppress the disease if strict implementation of NPIs is practiced (40% hand wash, 40% physical distancing and 25% wearing masks) in all urban populations of Ethiopia.

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted November 18, 2020. ; https://doi.org/10.1101 https://doi.org/10. /2020 In the rural population the projected number of critical cases could exceed the available number of ventilators in our health system (Figure 7) . To contain critical cases, people should strictly implement the three NPIs or the government should enforce additional public health measures. Since the time of the peak in rural populations is different from that of urban populations, the health system can make preparations for the transport system to bring these critical cases to heath facilities in urban centers. The two extreme scenarios show that improving hygiene by 5% and physical distancing by 40% could reduce the total?at peak time? number of critical cases from 2,361 to 1,569 (assuming 5% wear masks). Strict implementation of the three NPIs will shift the peak time forward by three months (Figure 7) . 17 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this this version posted November 18, 2020. ; https://doi.org/10.1101 https://doi.org/10. /2020 3.4 Projection of Cumulative Death Due to COVID-19 The number of deaths due to COVID-19 could be reduced if more than 40% of the populations practice physical distancing. The projected number of cumulative deaths under the implementation of different adherence levels of physical distancing and face mask wearing at fixed (15% and 20%) levels of hygiene are presented in Figure 9 .

Practicing recommended NPIs properly (40% physical distancing 5% face mask and 20% proper hand washing) could greatly reduce the number of death due to COVID-19. Table 3 and Table 4 presents summary of the projected number of active COVID-19 and ICU cases under different adherence levels of hygiene, wearing-masks and physicaldistancing in the urban and rural population of Ethiopia, respectively.

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted November 18, 2020. ; https://doi.org/10.1101/2020.11.16.20231746 doi: medRxiv preprint (a) 15% hygiene and nobody wear face mask (b) 20% hygiene and nobody wear face mask (c) 15% hygiene and 2% wear face mask (d) 20% hygiene and 15% wear face mask (e) 15% hygiene and 5% wear face mask (f) 20% hygiene and 5% wear face mask 19 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted November 18, 2020. ; https://doi.org/10. 1101 20 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted November 18, 2020. ; https://doi.org/10. 1101 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted November 18, 2020. ; https://doi.org/10.1101/2020.11.16.20231746 doi: medRxiv preprint

The basic reproductive number (Ro) measures the degree of spread of SARS-CoV-2. If this number is greater than one, the disease will spread out if additional public health measures are not taken. Table 5 and 6 presents estimated Ro at different time points by taking into account enforced NPIs since the onset of the disease in urban and rural Ethiopia, respectively. This number was higher in the first and 10 th week of the pandemic in urban Ethiopia. Further, compared with the rural population, at the beginning of the pandemic Ro was higher in the urban areas of Ethiopia. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted November 18, 2020. ; https://doi.org/10.1101 https://doi.org/10. /2020 3.6 Sensitivity Analysis Figure 10 shows that the model is particularly sensitive to physical distancing (SD), hand wash (HW) and direct human to human disease transmission rate (β1). Hand wash and physical distancing were inversely proportional to Ro, while β1 has directly proportional to Ro. Parameters with relatively large PRCC values (> 0.5 or < −0.5) as To see whether one parameter depends on another, pairwise comparisons were carried out. The processes underlying the parameters physical distancing, hand washing and the proportion of the population who become symptomatic infectious, θ, have the greatest potential of containing the epidemic if increased, whereas processes described by β1 and and ηa have the greatest potential of making the epidemic worse when increased ( Figure 10 ).

In this respect, increasing physical distancing directly reduces β1 as this lowers the likelihood of a susceptible individual getting in contact with a potentially infected individual. In addition, practicing good hygiene (such as regularly washing hands, using sanitizers to disinfect the infected environment, ventilation of rooms and avoiding touching the T-zones of the face) is associated with lowering the chance of contracting the virus from infected surfaces.

Anything contrary to the above increases the likelihood of getting the infection through the two aforementioned routes.

Moreover, improving hand wash practices will reduce viral contamination of the environment by infected individuals.

On the other hand, the processes underlying the parameters with negative PRCC have a potential to contain the number of cases when enhanced.

To ascertain whether the process described by those parameters are different or not, a pairwise comparison to 23 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted November 18, 2020. ; https://doi.org/10.1101/2020.11.16.20231746 doi: medRxiv preprint significant parameters was undertaken. Table 8 presents the computed p-values for different pairs of significant parameters by accounting for the false discovery rate (FDR). The major question posed at this point is: Are the different pairs of significant parameters different after FDR adjustment? Based on the FDR adjusted p-values in Table 8 , the compared pairs of parameters are rendered to be different if their p-value is less than 0.05 and not different otherwise. TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE   FM  TRUE FALSE TRUE TRUE TRUE FALSE TRUE TRUE TRUE   HW  TRUE TRUE TRUE TRUE TRUE TRUE TRUE TRUE   c  TRUE TRUE TRUE FALSE TRUE TRUE TRUE   β 1  TRUE TRUE TRUE TRUE TRUE TRUE   ε  TRUE TRUE TRUE TRUE FALSE   θ  TRUE TRUE TRUE TRUE   y  TRUE TRUE TRUE   η a  TRUE TRUE   w  TRUE   e   Table 8 presents a summary of compared parameters, where "TRUE" indicates compared parameters are significantly different, and "FALSE" otherwise. The results show that more sensitive parameters are also significantly different (see Table 8 ) except for the pair rate of recovery of asymptomatic infectious individuals-appropriate use of face masking y − F M , recovery rate of symptomatic infectious individual-rate of going to isolation w − e, social distancing and hand washing SD − HW , respectively which may not necessarily be correlated.

In this study, a modified SEIR model was developed to project the number of COVID-19 cases with different stage 

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted November 18, 2020. ; https://doi.org/10.1101/2020.11.16.20231746 doi: medRxiv preprint of the population who wear face masks by 10% will protect around 100,000 more people from being infected by COVID-19 in the urban population. Regardless of face mask coverage and hygiene, the peak time of COVID-19 could be shifted to the future by two months if 40% of the population implements social distancing properly. Similar to our finding, the study by Siraj et al. (2020) demonstrated practicing physical distancing and wearing face masks delays the peak time of the pandemic.

In the rural population setting, implementing hygiene measures, physical distancing, and wearing face masks at 20%, 40%, and 5%, respectively, could shift the peak time into to 2021. The projected number of active cases during the peak time could reach around 360,000 if 5%, 40%, and 20% of the population wear masks, keep physical distancing and hand wash with soap, respectively. As social distancing is already a custom due to the life style in the rural population of Ethiopia, improving hygiene by 20% could help to decrease the number of cases by 2-3 fold.

In the urban population setting, if 20% of the population implements physical distancing and 30% adopt hygiene measures, and 25% wear face masks the peak time of the pandemic will happen on December, 2020 with 119,000 estimated cases (Table 3) . During the peak time of the pandemic, except for some scenarios (30% physical distancing and above ≥ 15% face mask coverage), the projected number of ICU case are above the capacity of the Ethiopian health system, which needs the government attention (Table 3) .

COVID-19 is a deadly virus for which vaccination or effective medical treatment is not yet available. Hence the government should focus on prevention of infection using NPI mechanisms. Low socioeconomic status and behavioral attitudes lead to closer interactions in congested households, hampering the successful implementation of social distancing measures. Lack or improper use of masks and use of low quality masks may exacerbate the spread of infection. It is therefore imperative that the prudent use of NPIs is implemented in Ethiopia to contain the pandemic. NPIs used in combination are able to decrease cases and fatalities due to Covid-19. Separate implementation of each of the NPIs shortens the peak time of the pandemic and the number of new cases will increase by two-fold. We considered the place of residence being urban and rural as a factor since adherence to the recommended NPIs and the life style and living conditions, varies greatly in urban and rural society of Ethiopia.

Similar to the urban population projection, projection of cases in the rural setting was done by modifying the model parameters in the context of rural population practices. Table 4 presented summary of projected active COVID-19 cases and ICU cases in rural population of Ethiopia under 30 scenarios. The results shows that, at the peak of the pandemic, the number of projected cases in rural population will be higher than the urban population.

The first COVID-19 case was detected on March 13, 2020 and since then the government of Ethiopia has applied different mitigation strategies. But the virus continues to spread. Since community transmission of the virus is established, temporally relaxing any of the three non-pharmaceutical interventions could cause a rebound of the number of new cases, potentially leading to a collapse of the health care system. Our findings confirm, in the context of Ethiopia, that suppression of COVID-19 could be achieved by the combined implementation of three public health measures: wearing of face masks, social distancing and hand hygiene.The most effective public health measures are 25 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted November 18, 2020. ; https://doi.org/10. 1101 found to be face mask wearing in urban population and social distancing in rural populations. To mitigate COVID-19, in the urban population of Ethiopia, strict implementation of all three NPIs are required. In addition to announcing public health measures, the government should work on proper enforcement of NPIs, and the public should properly practice public health measures for their own safety and well-being. Based on the projected results under different conditions, concerned stakeholders could recommend achievable NPIs for their implementation to policy makers.

Studies on COVID-19 showed the distribution of cases strongly depend on age. The resources available could not get data on the number of cases by age group from concerned organizations in order to get age-structured predictions and to assess the impact of school closure on the pandemic. Further, the projection to the rural population setting is less reliable as the analysis has been constrained by different factors (i.e. exact number of incidences were not known, adherence to the implemented NPIs greatly vary across different regions of the country, estimated number of protected individuals assumed 20%,and others). Further, the study did not account for under reporting of COVID-19 cases.

This work was supported by Addis Ababa University, Ethiopia, and as a sandpit project of the One Health Regional 

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint

The copyright holder for this this version posted November 18, 2020. ; https://doi.org/10. 1101 

Mathematical modeling and optimal control analysis of covid-19 in ethiopia

Population biology of infectious diseases: Part I

Impact of non-pharmaceutical interventions on documented cases of covid-19

Covid-19 prevention measures in ethiopia current realities and prospects. ppp

The effectiveness of hand hygiene procedures in reducing the risks of infections in home and community settings including hand washing and alcohol-based hand sanitizers

Strong physical distancing measures in the united states reduced the covid-19 growth rate

On the definition and the computation of the basic reproduction ratio in models for infectious diseases in heterogeneous populations

Impact of non-pharmaceutical interventions (npis) to reduce covid-19 mortality and healthcare demand,. Imperial College COVID-19 Response Team

Estimating the number of infections and the impact of non-pharmaceutical interventions on covid-19 in 11 european countries. Imperial College COVID-19 Response Team

Nonpharmaceutical measures for pandemic influenza in non health care settings-physical distancing measures

Nonpharmaceutical measures for pandemic influenza in non health care settings-physical distancing measures

Economic and welfare effects of covid-19 and responses in ethiopia: Initial insights

Hand-washing

Face masks against covid-19: An evidence review

Covid-19 pandemic preparedness and response in ethiopia

Mathematical modeling of the spread of the coronavirus disease 2019 (covid-19) considering its particular characteristics. the case of china

Analysis of the mitigation strategies for covid-19: from a mathematical modelling perspective

Adoption of preventive practices for covid-19 in ethiopia. findings from a telephone-survey

Effect of non-pharmaceutical interventions to contain covid-19 in china

Early transmission dynamics in wuhan, china, of novel coronavirus-infected pneumonia

A methodology for performing global uncertainty and sensitivity analysis in systems biology

Mathematical assessment of the impact of non

Effects of non-pharmaceutical interventions on covid-19 cases, deaths, and demand for hospital services in the uk:a modelling study

Why is There Low Morbidity and Mortality of COVID-19 in Africa?

Monitoring health for the sdgs, sustainable development goals. geneva: World health organization. World health statistics overview

The effect of control strategies to reduce physical mixing on outcomes of the covid-19 epidemic in wuhan, china: a modelling study

Nonpharmaceutical measures for pandemic influenza in non health care settings-international travel-related measures

Early estimates of covid-19 infections in small, medium and large population clusters. s, 18

CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprint The copyright holder for this this version posted

Reproduction numbers and sub-threshold endemic equilibria for compartmental models of disease transmission

Aerosol and surface stability of sars-cov-2 as compared with sars-cov-1

The global impact of covid-19 and strategies for mitigation and suppression

Novel coronavirus (2019-ncov) situation reports. online

Face mask use in the general population and optimal resource allocation during the covid-19 pandemic

Measuring rural access: Using new technologies. washington dc

Sensitivity analysis of infectious disease models: methods, advances and their application

Nonpharmaceutical measures for pandemic influenza in non health care settings-personal protective and environmental measures

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprintThe copyright holder for this this version posted November 18, 2020. ; https://doi.org/10.1101 https://doi.org/10. /2020 Appendix Theorem 2.1Proof. Positivity of Ω To show Ω positive, first we show S of the model is positive for all t ≥ 0. To prove by contradiction: suppose,Then, using the firstwhich contradicts S (t) < 0. Thus, S(t) remains positive for t ≥ 0. For the rest of population variables we can show positivity using Gronwall's inequality as follows.29 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprintThe copyright holder for this this version posted November 18, 2020. ; https://doi.org/10. 1101 Since S(t) > 0 for t ≥ 0 we can use Gronwalls inequality and E ≥ E0e −(σ+µ)t > 0.since E(t) > 0 for t ≥ 0 we can use Gronwalls inequality and Ia ≥ Ia0e −(ρ+ε+y+µ)t ≥ 0since E(t) > 0, Ia(t) > 0 for t ≥ 0 we can use Gronwalls inequality and Is ≥ Is0e −(e+ω+α 1 +µ)t ≥ 0.Since Ia(t) > 0, Is(t) > 0 for t ≥ 0 we can use Gronwalls inequality and IHm ≥ IHm0e −(x+i+µ)t ≥ 0.since Is(t) > 0, IHm(t) > 0 for t ≥ 0 we can use Gronwalls inequality and Hc ≥ Hc0e −(g+α 2 +µ)t ≥ 0since Ia(t) > 0, Is(t) > 0, IHm(t) > 0, Hc(t) > 0 for t ≥ 0 and R(t) ≥ R0e −µt ≥ 0since Is(t) > 0 Hc(t) > 0 for t ≥ 0 and D ≥ D0 ≥ 0Ėnv = aIa + bIs − ϕEnv ≤ (a + b)N − ϕEnv. Since Ia(t)and Is(t) are less than the total population N given as π µ for all t ≥ 0. Applying again the Gronwall inequality, for 0 ≤ Env(0) ≤ a+b ϕ π µ leading to 0 ≤ Env(t) ≤ a + b ϕ π µHence all are non-negative for t ≥ 0. Finally, the total number of populationThus for initial data 0 < N (0) < π µ we have 0 ≤ N (t) ≤ π µ . Moreover, for the environment Env, we have E nv = aIa + bIs − ϕEnv ≤ (a + b)N − ϕEnv Since Ia(t) and Is(t) are less than N for all t ≥ 0 applying Gronwalls inequality for 0 ≤ Env(0) ≤ (a+b) ϕ π µ gives 0 ≤ Env(t) ≤ (a+b) ϕ π µ .

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) preprintThe copyright holder for this this version posted November 18, 2020. ; https://doi.org/10. 1101