Modelling COVID-19 in Newfoundland and Labrador

Diagram by Irena Papst

Our modelling approach is to fit a compartmental epidemic model to Newfoundland and Labrador COVID-19 data. This approach is used to model COVID-19 in Ontario, however, there are differences in the model formulation for Newfoundland and Labrador.

This modelling approach is mechanistic, or process-based. Process-based modelling contrasts with pattern-based, or phenomenological modelling, which aims to capture the shape of the data and has parameters that do not necessarily have epidemiological interpretations.

To understand what is meant by a process-based model, consider some of the model parameters that are fixed, and independent of the Newfoundland and Labrador COVID-19 data. Such parameters may describe epidemiological characteristics that are not region-specific, for example, the recovery rate from infection; or are estimated using other data, for example, the population size of Newfoundland and Labrador, which is estimated from Statistics Canada data, not the COIVD-19 hospitalization and death data.

Selected fixed parameter values (estimated independently of the Newfoundland and Labrador COVID-19 data)

• Population size: 522453

• Relative asymptomatic transmission (or contact): 0.67

• % of cases asymptomatic: 39%

• Time in exposed class: 3.3 days

• Time in pre-symptomatic class: 1.2 days

• Time for asymptomatic recovery: 7 days

• Time for severely symptomatic transition to hospital/death: 5.72 days

• Time in hospital (acute care): 10 days

• Time in ICU to death: 8 days

• % of ICU cases dying: 26%

Aside from the population size of Newfoundland and Labrador, the fixed parameter values are the PHAC.csv defaults for the McMasterPandemic software package.

Fitted parameter values (calibrated to match the Newfoundland and Labrador COVID-19 data)

• % of symptomatic cases that are mild: 99.7%

• The infection transmission rate is fitted and full details are provided here.

These are just some of the fixed and fitted parameters. When new data is provided, for example, the recent report of COVID-19 in hospital, long-term care, and private dwellings, I can provide more fixed parameters, which facilitates more accurate estimation of the fitted parameters.

How the parameters affect the relevant quantities, i.e. ICU occupancy, is shown in the flow diagram, and the model dynamics with only fixed parameters can be explored here.

The software we used is freely available and runs on R, a free programming language.