key: cord-0828630-uwzmvlr4 authors: Tuge Deressa, Chernet; File Duressa, Gemechis title: Modeling and Optimal Control Analysis of Transmission Dynamics of COVID-19: The Case of Ethiopia date: 2020-10-09 journal: nan DOI: 10.1016/j.aej.2020.10.004 sha: 529e701d64eb8709e3af6d25d7a0cfe3765ddd11 doc_id: 828630 cord_uid: uwzmvlr4 A mathematical model to estimate transmission dynamics of COVID-19 is developed. A real data of confirmed cases for Ethiopia is used for parameter estimation via model fitting. Results showed that, the diseases free and endemic equilibrium points are found to be locally and globally asymptotically stable for Ro<1 and Ro>1 respectively. The basic reproduction number is Ro=1.5085. Optimal control analysis also showed that, combination of optimal preventive strategies such as public health education, personal protective measures and treatment of hospitalized cases are effective to significantly decrease the number of COVID-19 cases in different compartments of the model. It is reported that [1] up to the present, the world has witnessed the happening of seven strains of human coronaviruses including SARS-CoV,MERS-CoV and the current coronavirus 2019-nCoV named COVID-19. COVID-19 was declared a pandemic by the World Health Organization(WHO) on 11 March,2020 [2] . As of July 27 th 2020, COVID-19 pandemic has claimed several hundreds of deaths globally and become the greatest threat of its kind in our generation [3] . It also posed a huge risk to public health and economics all over the world [4, 5] . The Ethiopian Federal Ministry of Health confirmed its first case of COVID-19 on 13 March, 2020, in Addis Ababa [6] . Despite the fact that the number of confirmed cases and deaths reported varies from time to time, the outbreak has taken the lives of several hundreds and affected different livelihoods in Ethiopia. Since the first case of COVID 19, various mathematicians around the world develop different mathematical models to understand the transmission dynamics of the virus, estimated the basic reproductive number and investigated effects of different intervention strategies via optimal control analysis. For instance, in [7] real data of confirmed cases of COVID-19 in Wuhan, China from January 21,2020 to January 28,2020 is used in a model developed to estimate the basic reproductive number The authors developed a 0 2.4829. R  fractional model and solved it numerically. A mathematical model for predicting transmission dynamics of the pandemic for Italy is considered in [8] .The result indicated that the combination of restrictive social distancing measures with widespread testing and contact tracing can result in ending the COVID-19 transmission in Italy. Analysis of Covid-19 infection, based on real data is made in [9] by developing a mathematical model for Wuhan, China. In [10] fractional order COVID-19 SIDARTHE model is developed to predict the evolution of the pandemic and investigated the impact of different plans to reduce the transmission of the disease with different values of fractional order. There are also many other mathematical models that reported on COVID-19 pandemics, refer [11] [12] [13] [14] . Mathematical models developed were mainly used to investigate the effects of different Non-pharmaceutical intervention strategies via simulation using different computing soft-wares [15] . Different scholars used different type of mathematical models in their analysis. For instance, an SEIR model was introduced by Wu et al. [16] for estimating the spread of the pandemic and obtained the basic reproductive number to . Tang et al. [17] used a deterministic 0 2.68 R  model to estimate the basic reproductive number to be as high as and concluded 0 6.47 R  that, contact tracing followed by quarantining and then isolation can mitigate the spread of the pandemic via reducing the basic reproductive number. Kiesha et al. [18] used SEIR model to simulate the outbreak of Coronavirus in Wuhan and indicated that imposing restriction on Wuhan people movement could help in delaying the peak time of the pandemic. Roda et al. [19] claimed that SIR model performs better than SEIR model and used it to predict COVID-19 epidemic in Wuhan and predicted the potential of a second outbreak after the return-towork in the city. Pang et al. used SEIHR mathematical model to investigate the effectiveness of quarantine measures applied in Wuhan city and factors affecting its effectiveness [20] . It is not only for COVID-19 that models were used to study the dynamics but also for several other infectious diseases. In [21] a mathematical model for the deadly disease named Ebola hemorrhage fever is developed and investigated the detail of endemic equilibrium points. SIR model with delay in the context of fractional derivative with Mittag-Leffler is considered in [22] . The author established the global and local stability of disease free and endemic equilibrium points using Lyapunov direct method. Mathematical models that investigated the dynamics of various infectious diseases are numerous in the literature [1, 23] and the references there in. Most of these studies have either restricted their domain of study to a particular country or they used data from a certain particular region. Some of the studies also overlooked the issue of optimal control analysis. Moreover, to the best of authors' knowledge, no study has been conducted on transmission dynamics of COVID-19 for the case of Ethiopia. To this end, the main objective of this work is to develop a mathematical model for the transmission dynamics of COVID-19 followed by mathematical analysis in order to come up with evidences that may help for designing mitigation strategies relevant to the context of Ethiopian. In this paper a mathematical model for COVID-19 relevant to study the transmission dynamics of coronavirus in Ethiopia is developed. The existence and uniqueness of solution of the model is proved. Local and global stability analysis of the diseases free and endemic equilibrium points are established. Parameter estimation based on real data of confirmed cases is made. The basic reproductive number suggested by the model is calculated. The importance of the basic reproductive number in target setting to control the pandemic is discussed by calculating sensitivity indices of the parameters of the basic reproductive number. Optimal control analysis of the model with three control strategies namely: public health education, personal protective measures and treatment of hospitalized COVID-19 cases were investigated followed by numerical simulation. Discussions, conclusions and limitation of the model are well specified. In this work, a mathematical model for COVID-19 transmission dynamics is developed based on compartmental approach of six groups named, S-susceptible, E-exposed, Isymptomatically infected, A-asymptomatically infected, H-hospitalized and R-recovered/ immune compartments. The total population at any time is given by Humans are recruited into the compartment  considered in the model. The above mentioned is the scenario that governmet of Ethiopia is using in combating the deadly COVID-19 pandemic, in the sense that the data of confirmed cases of COVID-19 in Ethiopia is organized incorporating all the compartments we have used in developing the model of this study. It is these data that we translated into a mathematical model. It seems logical to divide infected cases of COVID-19 into symptomatic and asymptomatic cases, in agreement with the observed property of coronavirus infection (majority are asymptomatic and the remaining symptomatic) in Ethiopia and elsewhere in the world. Moreover, the isolation/hospitalization is considered to be effective in reducing the transmission of COVID-19 [24] . In addition, it is known that coronavirus infected people don't get infectious immediately after contraction of the virus and this ensures the importance of considering the compartment of exposed cases. As a result, the aforementioned scenario and modification of SEIR [18, 23] model are used to develop the present model. The model developed and the parameters used are indicated in system of differential equation (1) and Table 1 . . The biological validity of the mathematical model depends on the solution of the dynamic system being positive and bounded for all values of time . The boundedness and positivity of the solution is proved in the subsequent Lemmas. remain positive in + . Proof: It is assumed that all the parameters used in the model are positive and hence we can set a lower bound for each of the equations in (1) as follows: . Solving the above inequalities respectively leads to: . Thus, for all and are positive in , of (1) there applying Groonwall's inequality leads to, ( ) and are all bounded since, , Thus, the region 6 ( , , , , , ) is positively invariant region for model (1) . then either the solution of (1) enters in a finite time or asymptotically. Hence the region attracts all solution of (1) in  Since has a continuous first derivative in ,it is then locally Lipschitz. As a result, by the f ℝ 6 + well known fundamental existence and uniqueness theorem [25] and Lemma 1 and 2 proved above, there exists a unique, positive and bounded solution for the system of differential equation (1) in ℝ 6 + . and that satisfies the quadratic equation where, The basic reproductive number, denoted by , is defined as number of secondary infections 0 R appears from one infected individual. It provides a threshold condition for the stability of the system. Finding through Jacobian approach using linearization of the dynamic system 0 R often doesn't work for complex systems [26] . Thus, we obtained by establishing the next 0 R generation matrix, , and calculating a spectral radius of the matrix at [27] . The matrices and are obtained by linearizing the mathematical model (1) about DFE that T 1 V  resulted in the Jacobian matrix given as: After constructing a matrix (M) from infectious components of the and partitioning it into dfe J transmission (V) and transition (T) matrices, we have: The spectral radius of the next generation matrix is the basic reproductive number Note that, can also be written as where This section deals with the local stability analysis of the diseases free and the endemic equilibrium points. Necessary Condition: Since both and , then from the 1 2 and are positive R R , for The two negative eigenvalues of the Jacobian matrix are . The sign of the real part of the remaining four eigenvalues are determined from a characteristic equation (5) by Routh-Hurwitz stability criteria: Necessary Condition: The coefficient is positive and can be shown to be positive as follows, In this subsection we showed global asymptotical stability of the DFE and EEP using Lyapunov function method. i. Global Stability Analysis of DFE The locally asymptotically stability of the DFE proved earlier indicates that COVID-19 can be eliminated from Ethiopia when if the initial sizes of the cases in each of the 0 1 R  compartments or in the society is very close to the DFE; initial size is in the basin of attraction of DFE, . In order to show that the stable in the positively invariant region defined in section2.  Consider a Lyapunov function candidate defined by, ( , , , , , ) Differentiating with respect to time in the direction of the solution of ( , , , , , , ) (1), and then substituting the appropriate values from (1) and leads to, Note that, by the inequality of arithmetic and geometric means we have   Differentiating in the direction of the solution of (1) leads to, L Replacing by their respective expression from model (1) we get, , , , , , We can now write Since all the parameters used in model (1) EEP is globally asymptotically stable. Note the implication of Theorem 4 is that, COVID-19 will institute itself in the society whenever regardless of the initial size of infectious individuals in the population. Parameter estimation is made based on real data of COVID-19 confirmed cases of Ethiopia for the first 140 days from 13 March to 31 July 2020 available online at [28] https://en.wikipedia.org/wiki/COVID-19_pandemic_in_Ethiopia. Some of the parameters are calculated from the data as follows: The total population of Ethiopia is approximately and the life expectancy of Ethiopians [29] one symptomatic case was confirmed, no individual was confirmed asymptomatic, no hospitalized and no recovery cases. Consequently, assuming initial number of exposed cases to be 200, the initial number of population is assumed where It is also learnt from the data that, in the last 140 days of the pandemic 5700 hospitalized or isolated cases were recovered from the diseases. Hence, the recovery rate of hospitalized  cases is calculated as, =6700/17530 =0.00273/day.  Some of remaining parameters are estimated from the mathematical model (1) via model fit using a computing soft-ware MatLab2018a by optimal estimation of parameters; least square curve fit along with Trust-Region-Reflective to find bounded estimated parameters. The result of model fit versus confirmed cases of COVID-19 is portrayed in Figure 1 . Thus, values of the parameters used in this work are obtained from assumption, calculation, model fitting and from the literature and are indicated in Table 2 . all negative and hence the endemic equilibrium point is locally asymptotically stable, and is in agreement with the analytical proof made in subsection 3.2. In order to determine the best control measures using the basic reproductive number, the normalized sensitivity index, which measures the the relative change of with respect to 0 R  denoted by ,and defined as is useful [24, 27] . The magnitude of sensitivity analysis of the reproductive number (3) shows that the most sensitive parameters in descending order are and so on. Increasing the , , , , , ,... It then follows that, if the transmission from asymptomatic cases to suspected cases is reduced by a fraction of at least ( i.e. ), then the transmission of COVID- . Now, the objective is to find an optimal control for the preventive strategies/measures for the relatively reduced coasts of the prevention strategies. The necessary 1 2 3 , and u u u and sufficient conditions for the existence of optimal control are established by Pontryagin maximum principle [34] . The function that minimizes the number of Exposed cases , number of symptomatically infected cases , number of asymptomatically infected cases and number of hospitalized cases over a time interval of can be defined as, where the control set is given by constraint given by system of differential equation (6). The necessary conditions that need to be satisfied by optimal control called Pontryagin's Maximum Principle converts equation (6) and (7) . Differentiating the Hamiltonian function with respect to the compartment variables gives the adjoint variables corresponding to the system given as follows: and are the adjoint variables , , , , , ( , , , , , ) , . Now, setting the transversality condition: we obtain the optimal controls and the optimality conditions respectively as: Note that, state equation (6), the adjoint equation (9) together with the characterization of the optimal control (13) and the transversality condition (11) are said to be Optimality system. The sections and subsection treated above describe analytical behaviors of the optimal control. The corresponding numerical simulation of the optimal control is treated in the next section. In this section numerical solution for the optimality system is presented. The initial conditions used in the simulation are As can be seen from Figure 2 above, the peak for the spread of the virus is observed around 238 th day of the pandemic. The result indicated that there are about 5095000 exposed, 603500 asymptotic, 479100 symptomatic and 1154000 hospitalized cases at the peak. Case II: In this case, all the three intervention strategies namely: educating the public , using of personal protective measures , and treating the hospitalized Case III: This case considered the combination of optimal personal protective measure and optimal treatment of hospitalized cases with no any public health education intervention . The simulation result of the effect of these two intervention 1 2 3 ( ( ) 0, ( ) 0, ( ) 0 ) u t u t u t    strategies is almost identically the same as in Figure 4 . in mitigating the transmission of COVID-19 i.e. case II. But as to case III, the full intensity of using personal protective measures has to be prolonged to the first 50 days before it slowly gets down to the lower bound at the end of the pandemic. Note that, it has been checked that cases where and As can be seen from Figure 6 , using the intervention strategy 'treating hospitalized COVID-19 cases' alone with the maximum effort couldn't reduce the number of exposed, symptomatic and asymptomatic cases as the number of cases in Figure 6 is the same as number of cases in Figure 2 for the compartments . Applying this control strategy , and E I A alone reduced the number of hospitalized cases from 1154000 ( Figure 2) to 27130 ( Figure 6 ) at the peak. Further, the control profile of this case is the same as case II for of Figure 3 ( ) u t Case V: In this case, optimal practice of public health education is considered without the intervention of personal protective measures and treatment of hospitalized cases . The simulation result for the effect of optimally practicing this intervention strategy is presented in Figure 7 for the optimal profile of and in 1 ( ) u t As can be seen from Figure 7 , the control profile that the optimal time range for public education is ranging from 176 th -186 th only for 10 days. Further, Figure 8 reveals that the peak of the pandemic for this case is around 176 th day but it is around 238 th day in Figure 2 . One can also anlyze from Figures 2 and 8 that using optimal public education alone could have reduced the duration of the pandemic, the peak period and number of cases in each of the compartments. Case VI: This case considered the optimal practice of personal protective measure without intervention of the other two control strategies: public health education and treatment of hospitalized cases . The simulation result of the control profile is depicted in Figure 9 where as its effect on the number of cases in the compartments identically the same as in case II. It can be observed from Figure 9 above, the usage of personal protective measures is optimal for the first 120 days of the duration of the pandemic. Comparing cases II and VI, one can learn that both cases lead to the required result of mitigating the transmission of the pandemic independently. However, in case VI the practices of personal protective measures need to be applied optimally for the first 129 days as compared to 8 days for case II. Despite the effort made by the Gvernment to teach the public and many Ethiopians effort to maintain personal protective measures and treatments of hospitalized cases being underway, the transmission of COVID-19 is increasing from time to time (see Figure1) . Consequently, it seems reasonable to make numerical investigation of nonoptimal intensity of practicing personal protective stratgies to appreciate their effect in decreasing the number of cases in the compartments of model (6) . Accordingly, we assumed and simulted the 1 3 ( ) 0 ( ) u t u t   model for different percentage values of pesonal protective measure 2 ( ) 1%, 5%, 9% u t  and the result is shown in Figure 10 below. Figure 10 it can be inferred that, if at least 1% from each of the symptomatic, asymptomatic and hospitalized cases maintained practicing personal protective measures at an optimal level from the onset of the pandemic, then the number of cases in the compartments would have reduced (black) as compared to the graph with no personal , , and E I A H protective (red) measures . The figure further reveals that a pretty improved result 2 ( ( ) 0) u t  would have been obtained for in the different compartments. 2 5% and 9% u  In this study, a mathematical model for transmission dynamics of COVID-19 for the case of Ethiopia is developed and its different properties including local stability analysis of the diseases free and endemic equilibrium points have been checked. Some of the parameter estimates were taken from Spencer et al., [12] and the remaining parameters were computed via model fitting based on real daily data of COVID-19 confirmed cases of Ethiopia from 13 March to 31 July, 2020. It is then calculated that the basic reproductive is . 0 1.5085 R  An optimal analysis of the model for the purpose of assessing the effect of public health education, effect of personal protective measures and effect of treating hospitalized or isolated cases in mitigating transmission of COVID-19 was conducted. The result showed that the optimal practice of combination of all the three intervention strategies significantly reduces the number of exposed, symptomatic, asymptomatic and hospitalized cases (see Figures 3 and 4) . Likewise, optimal usage of personal protective measures alone led to the required decreases in the number of cases in the compartments except that the optimal application of the control measure needs to be maintained relatively for a longer period of time (see Figure 9 ). It is also found that combining control strategies personal protective measures and treatment of hospitalized cases (Case III) is as good as combining the three strategies (case I) in combating the deadly COVID-19 pandemic in Ethiopia. In this work, the parameters that need to be targeted for suppression of COVID-19 pandemic by decreasing the basic reproductive number, are identified. For instance, it is shown that if the target is to decreasing the rate of transmission from asymptotically infected to suspected individuals then reducing this parameter to at least can suppress coronavirus from , ,  0.47   Ethiopia. The same type of work can be conducted in different countries by adapting to their situation. Moreover, optimal combination of the three intervention strategies proposed in this work or optimal combination of any two of the strategies or optimal practice of personal protective measures alone reduced the number of COVID-19 cases in the compartments as shown in the simulation results. Hence, the result of this study can be used as a policy input for the government of Ethiopia and other countries. The government of Ethiopia has to take necessary measures to make personal protective measures a mandatory practice throughout the period of the pandemic. Personal protective measures such as wearing facemask, regular hand washing and social distancing if practiced with optimum effort, can significantly decrease the disturbing effect of COVID-19 and safeguarded the nation. The government can also target its control strategy on reducing the basic reproductive number, taking into account the sensitivity analysis results of this work while considering the limitations of this work given below. It is ethical to make clear that mathematical models are in general approximations of real phenomena and as such they are naturally inaccurate. Moreover, the parameters used in any mathematical models are determined based on observations and experimentations using different numerical methods of computing software and hence are uncertain. Therefore, readers should take into account the limitation of the models while interpreting the findings. The dynamics of COVID-19 with quarantined and isolation, Advances in Difference Equations Situation Report -51, Data as reported by 11 A mathematical model for simulating the phase-based transmissibility of a novel coronavirus Pneumonia of unknown aetiology in Wuhan, China: potential for international spread via commercial air travel Nowcasting and forecasting the potential domestic and international spread of the 2019-nCoV outbreak originating in Wuhan, China: a modeling study World Health Organization, Ethiopia, first case of covid-19 confirmed in Ethiopia Modeling the dynamics of novel coronavirus (2019-nCov) with fractional derivative Modelling the COVID-19 epidemic and implementation of population-wide interventions in Italy A conceptual model for the coronavirus disease 2019 (COVID-19) outbreak in Wuhan. China with individual reaction and governmental action Novel Fractional Order SIDARTHE Mathematical Model of The COVID-19 Pandemic Impact of non-pharmaceutical interventions (NPIs) to reduce COVID-19 mortality and healthcare demand Estimation of the transmission risk of the 2019-nCoV and its implication for public health interventions A mathematical model for the novel coronavirus epidemic in Wuhan, China Mathematical modeling and epidemic prediction of COVID-19 and its significance to epidemic prevention and control measures The effect of control strategies to reduce social mixing on outcomes of the COVID-19 epidemic in Wuhan, China: a modelling study. The Lancet Public Health Nowcasting and forecasting the potential domestic and international spread of the 2019-nCoV outbreak originating in Wuhan, China: a modelling study. The Lancet Estimation of the transmission risk of the 2019-nCoV and its implication for public health interventions The effect of control strategies to reduce social mixing on outcomes of the COVID-19 epidemic in Wuhan, China: a modelling study. The Lancet Public Health Why is it difficult to accurately predict the COVID-19 epidemic Transmission dynamics and control strategies of COVID-19 in Wuhan On the mathematical analysis of Ebola hemorrhagic fever: deathly infection disease in West African countries. BioMed research international SIR epidemic model with Mittag-Leffler fractional derivative Reproduction numbers of infectious disease models. Infectious Disease Modelling Modelling the spread of COVID-19 with new fractal-fractional operators: Can the lockdown save mankind before vaccination Differential equations and dynamical systems A nonlinear epidemiological model considering asymptotic and quarantine classes for SARS CoV-2 virus On the definition and the computation of the basic reproduction ratio R 0 in models for infectious diseases in heterogeneous populations Wikipedia, the free encyclopedia Epidemiological parameter review and comparative dynamics of influenza, respiratory syncytial virus, rhinovirus, human coronavirus, and adenovirus The reproductive number of COVID-19 is higher compared to SARS coronavirus High Contagiousness and Rapid Spread of Severe Acute Respiratory Syndrome Coronavirus 2 Extending the type reproduction number to infectious disease control targeting contacts between types Uncertainty and sensitivity analysis of the basic reproductive rate: tuberculosis as an example An introduction to mathematical epidemiology We, the authors of the manuscript, thank Jimma University College of Natural Sciences for all the supports.