Influence of clinical case definitions with differing levels of sensitivity and specificity on estimates of the relative and absolute health benefits of influenza vaccination among healthy working adults and implications for economic analyses

Abstract

Clinical illness case definitions for influenza and methods used to define influenza seasons can vary substantially from study to study. These differences often result in differing levels of sensitivity, specificity and positive predictive value for the case definitions used. We explored the implications of different case definitions and outcome periods on estimates of influenza vaccine effectiveness and cost benefit by conducting additional analyses of data collected from a randomized, double blind, placebo controlled trial of the trivalent, intranasal, live attenuated influenza virus vaccine in healthy working adults. Febrile upper respiratory tract illnesses occurring during the peak influenza period was identified as the most specific clinical case definition expected to have the highest positive predictive value for true influenza whereas events occurring on a day with any symptom occurring during the entire outcome period was identified as the most sensitive clinical case definition with the lowest positive predictive value for influenza. As expected, the former provided the highest estimates of vaccine effectiveness (28.4% reduction in work loss days, 24.6% reduction in days with impaired productivity and 40.9% reduction in days with health care provider visits) but the lowest estimates of absolute reductions in events (42.4 work loss days prevented per 1000, 79.0 impaired productivity days per 1000, and 16.5 days with health care provider visits per 1000). On the other hand, events on days with any symptom during the entire outcome period provided the lowest estimates of vaccine effectiveness (18% reduction in work loss days, 18% reduction in days with impaired productivity, and 13% reduction in days with health care provider visits) but the highest estimates of absolute reductions in events (186.4 work loss days per 1000, 271.5 impaired productivity days per 1000, and 44.8 days with health care provider visits per 1000). When applied to a cost benefit analysis, the more specific case definition provided a break even cost for vaccination of US$ 6.58 whereas the more sensitive case definition provided a break even cost for vaccination of US$ 43.07. Clearly the latter combination is most appropriate when trying to assess the total population level impact of a vaccine preventable disease and the potential cost effectiveness of vaccination whereas the former may be most appropriate for assessing whether the vaccine actually works. The definitions of clinical influenza illness and outcome periods should be selected to match the study question.