Item – Thèses Canada

Numéro d'OCLC
1000103296
Lien(s) vers le texte intégral
Exemplaire de BAC
Auteur
Morrison, Kathryn,
Titre
A methodological framework for environmental public health surveillance with a practical example in wildfire smoke
Diplôme
Ph. D. -- McGill University, 2017
Éditeur
[Montreal] : McGill University Libraries, [2017]
Description
1 online resource
Notes
Thesis supervisor: David Buckeridge (Supervisor1).
Thesis supervisor: Sarah Henderson (Supervisor2).
Includes bibliographical references.
Résumé
"Wildfire smoke is considered a globally important cause of mortality by the World Health Organization, with an estimated 339,000 deaths from landscape fire smoke each year. The health effects from wildfire smoke are expected to intensify as changes in land use and climate are increasing the frequency and severity of wildfires, exposing more individuals to the harmful effects of smoke each year. The objective of an environmental public health surveillance (EPHS) system is to detect potential changes in the health status of the population, and to provide timely evidence for public health intervention during periods of hazardous exposure. However, the methodological and conceptual frameworks for surveillance are generally designed for infectious disease. EPHS poses some unique challenges, as exposures like air pollution are difficult to accurately measure and there are many indicators of health impact. Public health officials in environmental health have called for better decision-making surveillance tools. To address these challenges, in Chapter 3 I proposed a methodological framework for forecasting two health indicators against a common imperfectly measured exposure. My objectives were to identify a statistical approach that would be appropriate EPHS, be flexible so that changes to data characteristics could be accommodated (e.g., additional indicators or changes in exposure metrics), and address the computational constraints of requiring an "online" daily surveillance system. I demonstrated how the proposed model could be implemented using integrated nested Laplace approximations (INLA), a cutting-edge approach to approximate Bayesian inference. In Chapter 4, I explored the challenges of using the proposed EPHS system across an entire state or province with data aggregated by administrative boundaries, where some communities may have very small absolute populations. As an alternative to aggregating administrative units, I added spatial smoothing to the previously proposed model. The spatial smoothing stabilized the prediction variance in smaller regions, in exchange for a small loss of accuracy. In regions with larger populations, the smoothing was generally not found to be beneficial or necessary. The decision of whether to include spatial smoothing can be made within context of characteristics of the regions. Finally, I demonstrated how the model proposed in Chapter 3 could be used for surveillance. I used the severe wildfire smoke event in July 2015 in southern British Columbia as a case study, where smoky conditions resulted in PM2.5 concentrations up to ten times higher than baseline levels. Based on the theoretical intervention assessment, I found that simple, early interventions with lower effectiveness, such as public health messaging about steps to reduce exposure, were preferable to delayed interventions with higher effectiveness, such as evacuation. This work provides a flexible methodological approach that can be extended and improved as related research advances, such as improvements in exposure estimation and availability of additional real-time surveillance health data. The methods developed and evaluated in the first two manuscripts and implemented in the third are currently being integrated into the existing public health surveillance system at the BC Centre for Disease Control, and will be used during the wildfire season of 2017"--
Autre lien(s)
digitool.Library.McGill.CA
escholarship.mcgill.ca
escholarship.mcgill.ca
Sujet
Epidemiology and Biostatistics