Abstract
This study presents a method for evaluation of surveillance for avian influenza (AI) in wild birds and compares surveillance activities before and after changes in surveillance strategy in Great Britain (GB). In October 2008 the AI Wild Bird Surveillance (AIWBS) system in GB was modified to focus on passive surveillance (birds found dead), including those found during warden patrols of wetlands and wildlife reserves, with less emphasis on public reporting of birds found dead. The number of birds sampled by active surveillance (birds live-trapped or shot) was also reduced. In the present study the impact of these changes was investigated by comparing the 12 mo prior to October 2008 with the subsequent 12 mo. Four factors were considered for each surveillance system component: 1) the number of wild birds tested; 2) whether the tested wild birds were considered "higher risk species" (HRS) for being infected with AI; 3) the location of the birds tested with respect to counties designated as a priority for surveillance; and 4) the probability that the birds tested might yield a positive AI virus result based on surveillance results in wild birds across Europe. The number of birds tested by both surveillance types was greatly reduced after the strategy change. The proportion of birds sampled in priority counties also significantly decreased in the second year for both active and passive surveillance. However, the proportion of HRS sampled by active surveillance significantly increased, while a significant decrease in these species was seen for passive surveillance in the second year. The derived probability scores for detecting AI based on European surveillance results indicated a reduction in sensitivity for H5N1 highly pathogenic AI detection by passive surveillance. The methods developed to evaluate AIWBS in GB may be applicable to other European Union countries. The results also reflect the complex issues associated with evaluation of disease surveillance in wildlife populations in which the disease ecology is only partially understood.