Research II: Disease Modeling

This is the second of the two applied research-based sessions and focuses on how extracted information can be used to better understand disease spread. Much has been and continues to be published on disease modeling, ranging from the use of animal and human movement to predict disease importation and exportation to the most cost-efficient number of persons needed to be vaccinated to achieve herd immunity. Modeling can help assess current risk, forecast spread and future risk, and inform public health practice and response activities...but how confident are we in media signals and the models that are developed? This session will highlight two ongoing projects: comparing influenza signals in online medial to signals from laboratory reporting across different regions and developing epidemiological models and indicators for assessing infectious disease risks.

Infectious disease patterns in global online media data: detection, reasoning, and evaluation

David Buckeridge, Professor, McGill University

Advances in natural language processing are identifying more information from online media with greater accuracy. This increase in the amount and quality of information extracted creates new opportunities to support experts in reasoning about the relevance of infectious disease patterns. We will present plans for research in this area, including development of methods to link related media reports, to assess potential sources of media bias, and to incorporated extracted information into existing disease risk surfaces. We will also present initial results from ongoing research to evaluate the agreement of information extracted from online media with information from other surveillance systems.

Analytical tools providing epidemiological insight for event-based surveillance

Erin Reese, Senior Biostatistician Epidemiologist, Public Health Agency of Canada

One focus of the InSIGHT project is developing disease modelling tools to couple with event-based surveillance (EBS) systems to enhance epidemiological interpretation of detached events. EBS is challenged to provide early warning about emerging threats given noise in data and often little information about emerging threats. Therefore, we are developing disease modelling tools that require few input data to provide epidemiological context about detected events. We will report our progress using air traffic data to predict the number of travel-acquired cases and preview our ideas for developing tools using social media data, and automated prioritisation scoring of multiple detected threats.