Predicting the Peak of Influenza Cases by Geograhic Zones in Alberta

Abstract

Influenza or the ‘flu’ can affect people from all walks of life. The burden from influenza epidemics puts tremendous pressure on health services and other resources during a flu season. To better prepare for an incoming flu season, clinicians, health services and policy makers have a great interest in developing the capacity to predict the timing of the peak of an influenza epidemic based on identified cases in the early phase of the season. The objective of this thesis is to formulate an influenza model that can predict the week when the laboratory confirmed influenza cases peak for different geographical zones in the Province of Alberta: Edmonton, Calgary, North, Central and South zones. A Kermarck-McKendrick type compartmental model that comprises of the susceptible-infected (SI) compartments for three age groups (0-18 years, 19-64 years, and 64 years and over) is proposed. Contact mixing matrix among the age groups is computed. Estimates of model parameters are obtained by fitting the model to past influenza data (Year 2014-2015) from Alberta Health, using the nonlinear least squares method and the Mathematica software. The 95% confidence intervals for model parameters are obtained using the Markov Chain Monte Carlo method and then used for uncertainty analysis of our model predictions. Our model predictions for the peak time have shown a good agreement with past data for all Alberta zones. The findings in this thesis provide the groundwork and insight valuable for further development of influenza forecasting models.