Jessica Meaghan Healy
Published: 2015
Total Pages:
Get eBook
West Nile virus (WNV) is a zoonotic, mosquito-borne pathogen that causes human disease when virus transmission spills over from natural mosquito-bird transmission cycles. Vaccines and antiviral therapies are not available for human use, so mosquito control is the primary means for reducing transmission risk to the human population. Surveillance of mosquitoes and birds aims to provide early warning of times and places of increased epizootic transmission, while passive surveillance of the human population aims to detect cases of disease as soon as possible after their onset. In order to maximize the value of surveillance information, the system must be timely, sensitive, and efficient. In Chapter 1, a comparison of three common enzootic surveillance methods was conducted after standardization of spatio-temporal sampling frequency and costs. Testing of publicly-reported dead birds provided timely and cost-effective detection of WNV where susceptible bird species were found, and collection and testing of mosquitoes provided early and sensitive detection of virus activity. Serological information from sentinel chickens typically lagged detection of WNV by other methods, and random subsetting of existing flocks suggested that flock sizes could be reduced to 6-7 chickens without substantial losses in sensitivity or delays in detection. In Chapter 2, the association between enzootic surveillance methods and the occurrence of West Nile neuroinvasive disease (WNND) cases in humans. Mixed-effects Poisson regression validated by multiple measures of agreement was used to quantify differences in predictive performance of the surveillance methods alone or in combination. Predictions of WNND occurrence was most reliable for mosquito-based surveillance at a relatively coarse sub-county scale, and higher estimated abundance of infected mosquitoes was associated with higher numbers of disease cases. Comparisons of predictive performance across study areas suggested that caution is warranted when generalizing surveillance guidelines across diverse ecological regions. In Chapter 3, hierarchical Bayesian modeling was used to identify possible causes of variation in the sensitivity of the passive surveillance system for human disease. Febrile disease attributed to WNV is widely underreported, and estimating the causes of variation in the probability of detecting such cases was of primary interest. Sensitivity varied over space and time as a function of incidence anomalies and internet search indices, which were used as proxies for media attention and public awareness. Sensitivity estimates were highest during spring and early summer and varied among years. An estimated 1% of all West Nile fever cases and 66% of WNND cases were detected in the study time period. The results of this dissertation research can guide vector control agencies in the optimization of surveillance resources for efficient and early WNV detection, and can inform public health agencies about the factors that affect disease detection. This research has provided evidence of the comparative value of each of the surveillance methods, the predictive capabilities of the information gained from these methods, and suggested levels of sampling effort to extend existing surveillance resources. In addition this research has identified several influential factors on case detection which can be used by public health agencies to guide outreach campaigns to improve sensitivity of the passive surveillance system in places and times where it is lacking.
