Aspects of uncertainty and decision-making in fish disease control programs in Ontario, Canada
This thesis presents two projects that investigated two fish health programs in Ontario, Canada: federal fish health certification and provincial fish disease surveillance. One project addressed the accuracy and uncertainty of the test sensitivity and specificity of cell culture to detect virus in asymptomatic fish. Existent data were synthesized to obtain prevalence estimates for three pathogens. Subsequently, the probability of misclassifying a site in a zone-based program was assessed using, as examples, two pathogens found in Ontario. The other project assessed the suitability of several lake trout lakes in Ontario to receive fish infected with an indigenous pathogen. A Delphi survey was conducted for each of the projects under investigation. The Delphi experts believed that many factors considerably influence the ability to detect infectious pancreatic necrosis virus (IPNV) and infectious haematopoietic necrosis virus, and that relatively more of these factors are related to the laboratory procedure than to the host or environment. Panellists' estimates of the sensitivity of cell culture varied widely. The mean estimate for test sensitivity was less than 70%, whereas that for specificity was very high and precise. No virus other than IPNV was isolated in Ontario over all the years surveyed (1981 to 1995), and IPNV was only detected on some commercial farms. Mitigation hatcheries and wild sites had the highest percentages of infected sites for Renibacterium salmoninarum and Aeromonas salmonicida, respectively. Substantial uncertainty was shown to exist when sites are classified as IPNV or A. salmonicida-positive in Ontario (if test specificity is less than 100%), whereas negative site classifications were much more certain. According to the surveyed experts, several factors strongly influence the suitability of a site for receiving lake trout infected with A. salmonicida. Forty-three lakes were ranked according to their suitability using existent data, but a sensitivity analysis showed that lake pathogen status (an unknown) could significantly change this ranking. The present model, while a useful decision-making tool, has demonstrated the need to prioritize the collection of relevant and valid data.