IDENTIFIABILITY OF THE SPATIAL SEIR-HCD MODEL OF COVID-19 PROPAGATION
* · 2024
Read IDENTIFIABILITY OF THE SPATIAL SEIR-HCD MODEL OF COVID-19 PROPAGATION by * (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-18858.txt.
Introduction
Olga Krivorotko, Tatiana Zvonareva and Andrei Neverov ask which parameters of a spatial COVID-19 model can truly be recovered from case data. Their SEIR-HCD model spreads people through space by diffusion while tracking infection, hospital, critical and death states. Global sensitivity checks with the Sobol method and a Bayesian approach narrowed the ranges of the unknown values.
Case, critical and death counts alone could not pin down the diffusion rates that set how fast people move. Those rates need extra information about the process itself. The study warns model users which knobs their data can and cannot set, a guard against false confidence in epidemic forecasts.
Catalog record
| CREATOR | * |
|---|---|
| YEAR | 2024 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 34 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2412-18858
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
IDENTIFIABILITY OF THE SPATIAL SEIR-HCD MODEL OF COVID-19 PROPAGATION * Abstract This paper investigates the identifiability of a spatial mathematical model of the spread of fast-moving epidemics based on the law of acting masses and diffusion processes. The research algorithm is based on global methods of Sobol sensitivity analysis and Bayesian approach, which together allowed to reduce the variation boundaries of unknown parameters for further solving the problem of parameter identification by measurements of the number of detected cases, critical and dead. It is shown that for identification of diffusion coefficients responsible for the rate of movement of individuals in space, it is necessary to use additional information about the process. Keywords: reaction-diffusion model, identifiability, sensitivity analysis, source problem, optimization. INTRODUCTION Mathematical modelling of the spread of infectious diseases has come a long way with the outbreak of a new coronavirus infection caused by SARS-CoV-2 virus in 2019 in Wuhan. From classical differential models of epidemiology, whose foundation was laid in the work of Kermack and McKendrick in 1927 [ 1 ] , to agent-based models and mean-field control models, as well as their combination, researchers are still faced with incomplete and inaccurate data to sustainably describe the spread of epidemics. A more comprehensive review of models of fast-moving epidemic spread is given in the paper [ 2 ] and summarized in the figure 1 . Figure 1: Brief review of mathematical models of epidemic spread and relationship to statistical approaches. This paper investigates the direct and inverse problem for a spatial SEIR-HCD model of COVID-19 propagation in the region [ 3 ] : \displaystyle\dfrac{\partial u}{\partial t}=\nabla(nv\nabla u)+f(u\mathchar 59\relax q)\mathord{\mathchar 59\relax}\quad u(x\mathord{\mathchar 59\relax}0)=u_{0}(x)\mathchar 59\relax\quad t\in(0\mathchar 59\relax T). (1) Here u(x\mathord{\mathchar 59\relax}t)=(s\mathchar 59\relax e\mathchar 59\relax i\mathchar 59\relax r\mathchar 59\relax h\mathchar 59\relax c\mathchar 59\relax d)(x\mathord{\mathchar 59\relax}t) is the vector of system states characterizing the densities of susceptible ( s ), asymptomatic infected ( e ), COVID-19 patients ( i ), recovered ( r ), hospitalized ( h ), critical cases requiring ALV device connection ( c ) and died ( d ) as a result of COVID-19. The parameters v=(v_{s}\mathchar 59\relax v_{e}\mathchar 59\relax v_{i}\mathchar 59\relax v_{r}) and q(x)\in C(\mathbb{R}^{m}) are continuous spatial functions that describe the rate of movement of the respective populations in space and the patterns of infection in the population, respectively. Without limiting generality, we will consider a dimensionless one-dimensional space on x , i. e., x\in(0\mathord{\mathchar 59\relax}1) . The mathematical model ( 1 ) is based on a mass balance law in which the entire modelled population is constant in time, i. e. n(x\mathord{\mathchar 59\relax}t)=n(x) , where n(x\mathord{\mathchar 59\relax}t)=s(x\mathord{\mathchar 59\relax}t)+e(x\mathord{\mathchar 59\relax}t)+i(x\mathord{\mathchar 59\relax}t)+r(x\mathord{\mathchar 59\relax}t)+h(x\mathord{\mathchar 59\relax}t)+c(x\mathord{\mathchar 59\relax}t)+d(x\mathord{\mathchar 59\relax}t). For the first time in 1937, A. N. Kolmogorov, I. G. Petrovsky and N. S. Piskunov [ 4 ] proposed and justified the mathematical model ( 1 ), which later became known as the reaction-diffusion model, and also strictly proved that if the initial condition satisfies the following restrictions 0\leqslant u(x\mathord{\mathchar 59\relax}0)\leqslant 1\mathord{\mathchar 59\relax}\quad u(x\mathord{\mathchar 59\relax}0)=0\;\;\;\forall\,x<x_{1}\quad\text{and}\quad u(x\mathord{\mathchar 59\relax}0)=1\;\;\;\forall\,x>x_{2}\geqslant x_{1}\mathord{\mathchar 59\relax} then for f(u)=u(1-u) the population dynamics in time is described by the rate v^{*}=2\sqrt{f^{\prime}(0)nv} . It is interesting to note that similar wave solutions are found in the distributed epidemic propagation model [ 5 ] . In the above situation, such oscillatory solutions are interpreted as repeated outbreaks of epidemics. Taking spatial heterogeneity into account makes it possible to more accurately model an epidemic from a spreading center (a large city in a country, a capital city in a region, etc.) under known initial conditions. Thus, in [ 6 , 7 , 8 , 9 ] , estimates of the spread of COVID-19 in the first months of the epidemic, taking into account passenger traffic, were obtained. In [ 10 ] , the authors propose a multiscale spatial model based on a system of SEIR-type kinetic transport equations describing the passenger population moving at large scales (outside the city) and a system of diffusion equations describing the urban population acting at small scales. This modelling approach avoids unrealistic effects of traditional diffusion models in epidemiology, such as infinite diffusion rates at large scales and mass migration dynamics. The system parameters include both epidemiological characteristics (number of contacts, disease rate) and socioeconomic characteristics (social distance, vaccination). Given empirically defined parameters, the authors obtained the spatial distribution of the COVID-19 epidemic in Italy by solving systems of partial derivative equations using the finite volume method on unstructured meshes. In [ 11 ] , an integro-differential equation with delay describing the epidemic spreading process is proposed and studied, in which the solution has some biologically valid features. The feature of epidemiological models consists of the principle of transmission from an infected I to a susceptible S individual, which is characterized by the type of function f(u\mathchar 59\relax q) , i. e. the function f_{I}(S\mathord{\mathchar 59\relax}I\mathord{\mathchar 59\relax}q)=\dfrac{\beta(x)SI^{p}}{1+\kappa I} characterizes the non-linear transmission of viral infection with a number of contacts \beta(x) and a parameter \kappa>0 describing social distance or other socioeconomic constraints during the period of infection spread. The paper is organized as follows. In section 1 , the formulation of the SEIR-HCD model, its spatial interpretation, and methods for its numerical solution are introduced. In section 2 , the formulation of the inverse problem for the spatial SEIR-HCD model is presented. Sensitivity analyses by two global methods are carried out in section 2.2 . The algorithm for the numerical solution of the inverse problem, taking into account the identifiability results, is described in 3 . 1. DIRECT PROBLEM FORMULATION FOR THE SPATIAL SEIR-HCD MODEL In [ 12 ] , a mathematical model of the COVID-19 outbreak propagation, based on the mass balance law and described by a system of 7 ordinary differential equations, is formulated and analyzed: \displaystyle\left\{\begin{array}[]{ll}\dfrac{dS}{dt}=-\dfrac{\alpha_{I}S(t)I(t)}{N}-\dfrac{\alpha_{E}S(t)E(t)}{N}+\dfrac{1}{t_{imm}}R(t)\mathchar 59\relax\\ \dfrac{dE}{dt}=\dfrac{\alpha_{I}S(t)I(t)}{N}+\dfrac{\alpha_{E}S(t)E(t)}{N}-\dfrac{1}{t_{inc}}E(t)\mathchar 59\relax\\ \dfrac{dI}{dt}=\dfrac{1}{t_{inc}}E(t)-\dfrac{1}{t_{inf}}I(t)\mathchar 59\relax\\ \dfrac{dR}{dt}=\dfrac{\beta(t)}{t_{inf}}I(t)+\dfrac{1-\varepsilon_{HC}}{t_{hosp}}H(t)-\dfrac{1}{t_{imm}}R(t)\mathchar 59\relax\\ \dfrac{dH}{dt}=\dfrac{1-\beta(t)}{t_{inf}}I(t)+\dfrac{1-\mu}{t_{crit}}C(t)-\dfrac{1}{t_{hosp}}H(t)\mathchar 59\relax\\ \dfrac{dC}{dt}=\dfrac{\varepsilon_{HC}}{t_{hosp}}H(t)-\dfrac{1}{t_{crit}}C(t)\mathchar 59\relax\\ \dfrac{dD}{dt}=\dfrac{\mu}{t_{crit}}C(t).\end{array}\right. The model ( 1 ) does not take into account the self-isolation index from Yandex, as it has stopped being updated since 2021 due to the removal of restrictive and control measures. The model takes into account: • incubation period t_{inc} of asymptomatic infection, • the possibility of re-infection due to weakening of immunity during the time t_{imm} , • hospitalization time t_{hosp} and duration of artificial lung ventilation (ALV) device use t_{crit} regulated in the simulated region, • possible mortality after critical state. The description of parameters of the model ( 1 ) and initial conditions characterizing the spread of COVID-19 in the Novosibirsk region from 31.01.2022 (the period of the outbreak due to the appearance of a new Omicron strain in the region) is given in the table 1 . Table 1: SEIR-HCD model parameters characterizing the distribution of COVID-19 in the Novosibirsk region from 31.01.2022. Parameter Description Value \alpha_{E} Infection parameter for asymptomatic and susceptible groups 0.0922 \alpha_{I} Infection parameter for infected and susceptible groups 0.3856 \beta(t) Proportion of individuals with late IgG antibodies to SARS-CoV-2 obtained by Invitro Data [ 13 ] \varepsilon_{HC} Proportion of hospitalized cases with ALV support 0.0376 \mu Proportion of dead from COVID-19 0.4754 t_{inc} Incubation period duration (days) \mathrm{LogN}(4.6\mathord{\mathchar 59\relax}4.8) t_{inf} Infection period duration (days) \mathrm{LogN}(6.6\mathchar 59\relax 4.9) t_{hosp} Hospitalization period duration (days) \mathrm{LogN}(3\mathchar 59\relax 7.4) t_{crit} ALV support duration (days) \mathrm{LogN}(6.2\mathchar 59\relax 1.7) t_{imm} Average