To our knowledge, this is the first systematic assessment of adherence to recommended protective measures used by personnel tasked with bird collection during a large outbreak of HPAI A/H5N1 in wild birds.
The environmental conditions during wild bird collection differed considerably from culling of poultry during outbreaks of avian influenza. Study participants reported difficulties owing to a wet and cold environment during wild bird collection, and almost half the participants collected potentially infected birds that were still alive which resulted in a high risk of exposure. These environmental conditions were reported to be favourable for virus survival as the A/H5N1 viruses are more stable in wet and fresh feces of infected animals [15, 16].
In contrast to other studies assessing adherence to recommended preventive measures during outbreaks of HPAI, we not only measured the proportion of PPE always used, but constructed a score to summarize both the completeness and frequency of PPE use simultaneously considering differences in the protective effect of PPE-devices. Studies conducted after the HPAI A/H7N7 outbreak in poultry in the Netherlands in 2003 showed a low self-reported adherence in the consistent use of masks and protective goggles among poultry farmers (6%, 1%) and cullers (25%, 13%) . Based on the PPE score, our study showed that PPE adherence differed between occupational groups as well, and was highest in firemen who probably similar to cullers had more previous experience in the use of PPE owing to their occupation. However, this estimation requires confirmation by further investigations. Our study showed better adherence to using protective goggles among all personnel tasked with bird collection (37%). Compared with the results of the study after an A/H7N3 outbreak in poultry in Canada in 2004 (always use masks: 83%, gloves: 85% and protective goggles: 55%) the adherence in our study was poor (19%; 78%; 37%) . However, the better result for adherence measured using the PPE-score method (46%, 88%, 51%), suggests an underestimation of the true adherence when restricting the analysis to the consistent use of PPE.
Interference of PPE with the task of wild bird collection was reported in particular for protective goggles and masks, in keeping with the lowest adherence for these two PPE-devices. The use of mobile phones during bird collection could have reduced the adherence of mask use as well. To our knowledge, other studies have not specifically addressed the question of gaps or barriers reducing adherence.
Since 2003, the German Committee for Biological Agents has recommended seasonal influenza vaccination for people exposed to A/H5N1 infected birds or poultry . Even though seasonal vaccination does not protect against infection with avian influenza, it can potentially reduce opportunities for reassortment by avoiding the simultaneous infection of humans with avian and human influenza viruses. After an influenza vaccination, the development of an immune response takes about 2 weeks . In our study, all participants had been offered seasonal influenza vaccination. However, 53% of those were unvaccinated. Among all participants, 29% had received a seasonal influenza vaccination only shortly before bird collection in February 2006 who may not have developed immunity against seasonal influenza during the first few days of bird collection. The proportion of study participants with a seasonal influenza vaccination prior to the outbreak (18%) in our investigation was similar to the proportion in a study carried out after an outbreak of A/H7N3 in Canada (21%)  and in a study after an outbreak of HPAI A/H5N1 in England (16%) . However, this proportion is lower than influenza immunization coverage in the general population of Germany in the season 2005/06 (32.5%) .
In our study, differences were found between PN and MN results. Five sera reactive to A/H5N1 in PN assay could not be confirmed by MN assay. Discrepant results between serological assays have been shown for sera in the study in the Netherlands as well. A/H7N7-reactive sera initially tested by haemagglutination inhibition (HI) assay were all negative by an MN assay . Therefore, the MN assay showed higher specificity than PN and HI assays. The lack of concordance of results between GNRCI and WHO-CC might be also explained by the use of different reference strains, even though both belong to clade 2.2.
A possible cross-reaction of A/H1N1 antibodies in high concentrations with the A/H5N1 reference virus was found by MN assay at GNRCI. It has been postulated that seasonal influenza vaccination may boost the anti-N1 response and therefore could lead to false-positive results against A/H5N1 virus as well because of cross-reactions between A/H1N1 and A/H5N1 . As no association between human influenza vaccination status and PN assay result among study participants was found, it is unclear whether the cross-reaction between A/H5N1 and A/H1N1 could explain the reactivity by PN assay.
A limitation of our study is its conduct one year after the outbreak. Some study participants might have been unable to remember the details of their activities during the outbreak in 2006, which could reduce the validity of the study findings. Also the influenza antibody titres could decay over time , so the serological investigation might have failed to reveal seroconversion to H5 owing to the length of time elapsed since exposure.
Because soldiers of the German Federal Defence Force and professional firemen could not be included in this study, this investigation was based on participation of local personnel who performed initial response to the outbreak. As this study conducted on Ruegen was voluntary, not all individuals involved in bird collection were included. Participation among government workers and veterinarians was high at > 80%. Therefore, this study provides a good assessment for these two groups. The participation among firemen, however, was lower (55%) and might lead to selection bias as possibly firemen taking part in this study were more highly motivated and interested and used PPE more intensively than their non-responding colleagues. Therefore, the findings from this group could be less reliable. This could be another limitation of this study.
Evidence of the risk factors or protective effect of the protective measures could not be further analysed in our study because no A/H5N1-positive case or influenza-like illness was detected. Our findings should be completed by future studies under other operating conditions.