SEAHIR: A Specialized Compartmental Model for COVID-19

Abstract

The SEIR (Susceptible-Exposed-Infected-Removed) model is widely used in epidemiology to mathematically model the spread of infectious diseases with incubation periods. However, the SEIR model prototype is generic and not able to capture the unique nature of a novel viral pandemic such as SARS-CoV-2. We have developed and tested a specialized version of the SEIR model, called SEAHIR (Susceptible-Exposed-Asymptomatic-Hospitalized-Isolated-Removed) model. This proposed model is able to capture the unique dynamics of the COVID-19 outbreak including further dividing the Infected compartment into: (1) "Asymptomatic", (2) "Isolated" and (3) "Hospitalized" to delineate the transmission specifics of each compartment and forecast healthcare requirements. The model also takes into consideration the impact of non-pharmaceutical interventions such as physical distancing and different testing strategies on the number of confirmed cases. We used a publicly available dataset from the United Arab Emirates (UAE) as a case study to optimize the main parameters of the model and benchmarked it against the historical number of cases. The SEAHIR model was used by decision-makers in Dubai's COVID-19 Command and Control Center to make timely decisions on developing testing strategies, increasing healthcare capacity, and implementing interventions to contain the spread of the virus. The novel six-compartment SEAHIR model could be utilized by decision-makers and researchers in other countries for current or future pandemics.

Keywords: COVID-19; Erlang distribution; SARS-CoV-2; SEIR; compartmental models; predictive modelling.

Figure 1

Model definition. The most infectious compartments are those of the Exposed and Asymptomatic carriers. Less infectious compartments are those of Isolated and tested people due to subsequent isolation. In this graph we also show that the average time of a person in the Exposed compartment is less than the average time of a person in the Asymptomatic compartment, which is in turn less than the average time of a person in the Isolated or Hospitalized compartments [9].

Figure 2

The Erlang distribution with parameters k = 5 and λ = 5/5.2 = 0.96, showing the time (in days) an exposed person spends in incubation period.

Figure 3

The optimization result on the logarithmic scale.

Figure 4

The extrapolation of the result of Figure 3. The assumed division of infected population is as follows: 30% of the Infected population would be asymptomatic, 55% mild cases, and 15% hospitalized (which are further divided into three sub-categories) according to Lauer et al. [9].