[Paper Review] A numerical simulation of the COVID-19 epidemic in Argentina using the SEIR model

Juan E. Santos, José M. Carcione+4 more|arXiv (Cornell University)|May 12, 2020

COVID-19 epidemiological studies27 references4 citations

TL;DR

This study uses a calibrated SEIR model to simulate the COVID-19 epidemic in Buenos Aires, Argentina, leveraging reported fatalities to estimate key transmission parameters. Despite early lockdown success, the reproduction number $R_0$ increases over time, predicting up to 9 million infections and significantly higher casualties if control measures are not strengthened, highlighting the urgent need for sustained intervention strategies to reverse rising transmission trends.

ABSTRACT

A pandemic caused by a new coronavirus has spread worldwide, affecting Argentina. We implement an SEIR model to analyze the disease evolution in Buenos Aires and neighbouring cities. The model parameters are calibrated using the number of casualties officially reported. Since infinite solutions honour the data, we show different cases. In all of them the reproduction ratio $R_0$ decreases after early lockdown, but then raises, probably due to an increase in contagion in highly populated slums. Therefore it is mandatory to reverse this growing trend in $R_0$ by applying control strategies to avoid a high number of infectious and dead individuals. The model provides an effective procedure to estimate epidemic parameters (fatality rate, transmission probability, infection and incubation periods) and monitor control measures during the epidemic evolution.

Motivation & Objective

• To model the time evolution of the COVID-19 epidemic in the Buenos Aires metropolitan region using a compartmental SEIR model.
• To calibrate model parameters—particularly transmission rate, incubation and infectious periods, and infection fatality rate—using officially reported death counts, considered more reliable than case numbers.
• To assess the impact of early lockdown and subsequent trends in $R_0$ on epidemic outcomes, especially the number of infected and deceased individuals.
• To evaluate the effectiveness of control strategies in reversing the rising $R_0$ trend to prevent overwhelming healthcare systems.
• To provide a data-driven, computationally feasible framework for monitoring and forecasting epidemic dynamics under uncertainty in key virological parameters.

Proposed method

• The study employs a system of first-order ordinary differential equations (ODEs) to represent the SEIR compartments: Susceptible (S), Exposed (E), Infected (I), and Removed (R).
• The ODE system includes parameters for transmission rate $\beta$, incubation rate $\epsilon$, infectious rate $\gamma$, natural death rate $\mu$, and disease-induced fatality rate $\alpha$, all in units of 1/time.
• Model calibration is performed using cumulative reported fatalities in Buenos Aires, treating this data as more reliable than reported infections for parameter estimation.
• A forward Euler scheme is used to numerically solve the ODE system, simulating daily peaks in infections and deaths under varying parameter sets.
• Multiple scenarios are explored, including different incubation periods (e.g., 11 days) and fatality rates, to assess sensitivity and forecast outcomes.
• The model accounts for population balance (births and natural deaths), assuming they are equal, so total population remains approximately constant.

Experimental results

Research questions

• RQ1 How does the reproduction number $R_0$ evolve over time in Buenos Aires, and what factors drive its post-lockdown resurgence?
• RQ2 What are the most plausible estimates for the incubation and infectious periods, and how do they affect the epidemic’s trajectory?
• RQ3 How sensitive are the predicted infection and death tolls to variations in the infection fatality rate (IFR) and transmission probability?
• RQ4 To what extent can early lockdown measures delay or reduce the epidemic peak, and what are the risks if $R_0$ increases afterward?
• RQ5 What control strategies are most effective in reversing the rising $R_0$ trend to prevent catastrophic case and death numbers?

Key findings

• The reproduction number $R_0$ initially decreased after early lockdown but began to rise again, indicating a resurgence of transmission, likely due to increased spread in densely populated slums.
• An incubation period of 11 days leads to approximately three times more casualties by the end of the epidemic compared to shorter incubation periods.
• If the rising $R_0$ trend continues, the model predicts between 4 million and 9 million infected individuals in Buenos Aires, with a corresponding sharp increase in deaths.
• The infection fatality rate (IFR) is estimated between 0.5% and 2%, with a case-specific value of 0.84% derived from calibration to official fatality data.
• The model successfully reproduces the observed fatality curve using calibrated parameters, validating its use as a forecasting tool under data uncertainty.
• The study demonstrates that relying solely on early interventions is insufficient; sustained control measures are essential to reverse the increasing $R_0$ trend and prevent overwhelming outcomes.