Madeline Ward, PhD student – University of Calgary

Supervisor – Rob Deardon

Collaborators – Lorna Deeth (Math/Stats) University of Guelph and Caitlin Ward (Public Health) University of Minnesota and Vineet Saini, Alberta Health Services

Project Description: Epidemic models can help us better understand the transmission mechanisms of emerging and other diseases, optimize the introduction of new mitigation policy, evaluate the effectiveness of previous interventions, and plan for future outbreaks. However, the utility of a disease model depends on its ability to capture underlying transmission dynamics. In the COVID-19 pandemic, for example, we observed populations react by changing their behaviour over time on an unprecedented scale. Changes in population behaviour as exhibited through movement patterns, physical distancing, and hygiene practices resulted in observed case counts that often deviated substantially from those that were predicted by standard epidemic models (e.g., SIR models). Traditional disease transmission models are generally built under the assumptions that population behaviour will remain stable over time, and that the population behaves homogeneously. While this may be reasonable over short time frames, in many situations we need models that have less restrictive assumptions to get accurate estimates.
The applicants and other co-authors have recently devised a framework of Bayesian epidemic models that capture both disease dynamics and behavioural change in the population simultaneously (Ward et al. 2023, Ward et al. 2024). This is done through a so-called alarm function which allows for decreasing force of infection (transmission risk) in response to increases in the number of cases observed in the global and/or local population. Such models can allow for heterogeneity in transmission risk over time due to the behavioural change, as well as heterogeneity in the ways that populations come into contact (e.g., spatial or network-based models). However this new class of models has so far ignored the availability of secondary data that can inform given behaviour change effects for modelling diseases such as COVID-19; e.g., the known timing of public health interventions, or metadata such as that available via Google Community Mobility reports and Facebook Data for Good Colocation Maps (Google LLC 2022, Iyer et al. 2023).
This project will focus on developing flexible statistical models that utilize such covariate data and explicitly capture the effect of behavioural changes over time and between different population subgroups. The behaviour data will be used to help inform the “alarm” component to the model that describes the extent of behaviour change at a given time. We will also investigate different model structures that can allow for different subgroups to change their behaviours in different ways – for example, at various points in the pandemic essential workers and families with school-aged children likely had different abilities to maintain protective behaviours from those who could easily work from home. Through this work, we aim to provide a framework for models that can more accurately capture disease dynamics in situations where behaviour changes impact transmission, as well as to be able to better describe how people changed their behaviours in response the COVID-19 pandemic.
This project is a collaborative effort with researchers at the Universities of Guelph (Lorna Deeth, Math/Stats) and Minnesota (Caitlin Ward, Public Health), and Alberta Health Services (Vineet Saini). Our collaborators provide expertise in epidemiology, epidemic-related behavioural change and public health policy, helping to keep our model and software development in a space relevant and utilizable by public health officials.

References.

1. M. Ward, R. Deardon, L. Deeth (2024). A framework for incorporating behavioural change into individual-level spatial epidemic models. To appear in the Canadian Journal of Statistics. http://arxiv.org/abs/2308.00815
2. C. Ward, R. Deardon & A. Schmidt (2023) Bayesian modelling of dynamic behavioural change during an epidemic, Infectious Disease Modelling, 8(4), 947-963. https://doi.org/10.1016/j.idm.2023.08.002
3. Google LLC (2022). Google COVID-19 Community Mobility Reports. https://www.google.com/covid19/mobility/
4. S. Iyer, B. Karrer, D.T. Citron, F. Kooti, P. Maas, Z. Wang, E. Giraudy, A. Medhat, P.A. Dow, A. Pompe (2023). Large-scale measurement of aggregate human colocation patterns for epidemiological modeling, Epidemics, 42:100663. https://doi.org/10.1016/j.epidem.2022.100663