Influenza causes substantial morbidity and mortality yearly in both temperate and tropical regions, as well as sporadic and potentially severe pandemics. Although vaccines for seasonal influenza exist, most options for controlling influenza outbreaks are reactive in nature. Sufficiently accurate and well-calibrated forecasts, on the other hand, could allow public health practitioners, medical professionals, and the public to respond to unfolding influenza outbreaks proactively. For example, hospitals could prepare additional beds for a predicted surge, and public health experts could redouble vaccination efforts. Recently, skillful forecasts have been developed for a range of infectious diseases, including influenza, but this work has been limited to only a few countries. In this dissertation, we explore the potential for generating accurate influenza forecasts using a publicly-available dataset of country-level epidemiologic and virologic surveillance data. In Chapter 2, we use a combined model-inference system to generate retrospective forecasts for 64 countries in both temperate and tropical climates. We show that forecast accuracy is significantly better in countries with temperate climates, and that inclusion of environmental forcing, specifically modulation of viral transmissibility due to variability of absolute humidity conditions, also improves forecast accuracy in temperate climates. In Chapter 3, we develop a metapopulation model of twelve European countries using data on international air travel and commuting. We find that this model is unable to produce more skillful forecasts than those produced for individual countries in isolation. We make recommendations for improvements in data collection and reporting that may increase the success of similar modeling efforts in the future. In Chapter 4, we assess the performance of real-time forecasts generated for 37 countries over two influenza seasons and discuss the potential for their use in public health decision making. Finally, in Chapter 5 we describe the results of a small survey of public health practitioners in the United States. We find that the majority of respondents desire more effective communication between modelers and public health practitioners, and we discuss the importance of regular and improved communication in advancing the practical use of forecasts as public health decision making tools. This dissertation advances the science of influenza forecasting by demonstrating that skillful retrospective and real-time forecasts can be generated for many countries where previous forecasting efforts are either minimal or absent. However, it is vital that data quality issues be addressed if further progress is to be made. Future work should focus in particular on climatic drivers of influenza in the tropics and subtropics, on the role of human travel at various spatial scales, and on the development of regional and local forecasting capacity. Additionally, dedicated collaboration between modelers and public health practitioners will be instrumental for motivating and informing the use of forecasts in combating influenza outbreaks.
| Identifer | oai:union.ndltd.org:columbia.edu/oai:academiccommons.columbia.edu:10.7916/d8-pm6m-tj30 |
| Date | January 2020 |
| Creators | Kramer, Sarah Corinne |
| Source Sets | Columbia University |
| Language | English |
| Detected Language | English |
| Type | Theses |
Page generated in 0.0024 seconds