Awareness about the specific risk faced by poultry workers was uniformly high, and a majority of over 60% among respondents felt afraid of contracting AI. These findings are unsurprising given that the study was conducted shortly after the first AI outbreak in Nepal but might also to some extent reflect increased governmental campaign efforts at promoting awareness after the outbreak had occurred. Also, almost everyone knew about the importance of washing hands with soap and water, which had been the main message in the campaign. This finding is in line with studies on poultry workers in other countries which similarly found hand washing to be by far the best known practice [25, 26]. Assessment of knowledge in other areas, however, unveiled distinct gaps and deficits. While still almost 70% knew about the protective capacity of gloves, only half of the sample mentioned face masks as an option and only few knew about special boots or boot covers and body suits. Also, only about one fourth named a basic procedure such as washing and disinfecting surfaces and utensils. Another study, on poultry workers in Nigeria, also reported "low" levels of knowledge about preventive behaviours , others, however, found distinctly higher rates for knowledge about face masks, boots covers and cleaning procedures  than the present study did. One reason for this discrepancy might be that open-format questions like those used in the present study generate lower knowledge scores than identification tasks. Also, the finding about low knowledge about cleaning/disinfecting could be interpreted as a tendency to perceive such behaviours as routine everyday practices instead of as extraordinary precautionary measures against AI. Yet, these gaps in knowledge raise concern and suggest that future campaigns should make additional efforts to specifically target poultry workers and - beyond hand washing - focus also on the more specific behaviours which are relevant for prevention and containment of the virus at the source, i.e. on the poultry farms.
Analysis of the factors which were associated with knowledge about protection showed that TV and newspapers, which carried a substantial part of the campaign messages in Nepal, played an important role. Those who received information about AI via TV and newspapers were able to name more preventive behaviours than those without that kind of exposure - an effect also reported by other studies [26, 27]. This finding certainly suggests a beneficial effect of the Nepali mass media campaign but at the same time highlights deficits in reaching groups without access to these types of media - something to be considered for future health education efforts. Another relevant factor, which was negatively associated with level of knowledge, was fear. At first glance, this seems to suggest that a higher degree of fear leads to less knowledge due to defensive processes such as not wanting to deal with the threat and therefore searching for less information. While such an explanation cannot be excluded, another mechanism is more plausible. The focus in this case was specifically on knowledge about protective behaviours, not on knowledge about AI in general, its danger potential, transmission pathways etc. While the latter type of knowledge is likely to make people aware of risks and therefore also more concerned, knowledge about effective protective behaviours might rather reduce fear by creating expectancies about successful control.
The data on protective behaviours showed that washing hands with soap and water were fairly standard practice. High-frequency cleaning and disinfecting, however, was not and neither was habitual use of personal protective equipment. Low usage rates for protective clothing have recently also been reported by studies with Nigerian poultry farmers [25, 26], while findings from an Italian study registered considerably higher rates , which probably reflects different financial resources to fund such equipment on a regular basis.
The relevance of economic constraints was also indicated by the findings from the multivariable models. There was a substantial difference in usage rates of protective equipment between poultry farm owners and employees. The latter had higher odds to use personal protective equipment than owners of farms. Employed poultry workers in Nepal tend to work more often in larger-scale, economically better-off poultry businesses whereas many owners operate small-scale family businesses. Paid employees might thus more often have been provided with protective equipment by farm management while owners of small-scale family businesses were more likely to save on expenses, thereby trading off possible longer-term preventive gains against more immediate economic savings [26, 28].
The finding that those who had more knowledge were also those who actually acted more preventively is consistent with some other studies from the field [18–20] even if the overall evidence on this issue is still inconsistent . One possible explanation for such discrepancies is that effects might depend upon the specific type of knowledge measured. Knowledge about effective behaviours, which was the focus of the present study, is particularly likely to enable perceptions about efficacy of behaviours which have consistently been linked to precautionary practices . Nevertheless, knowledge about a threat and potential countermeasures alone will most often be insufficient to achieve behaviour change, as other factors such as economic concerns or social norms are essential for enabling or disabling such change. Yet, the findings emphasize the role of awareness-building about the availability of preventive options as a first step in generating preventive habits.
A result which raises concern is that despite a generally voiced agreement with governmental emergency control measures and bio-security regulations large parts of the respondents expressed doubts with regard to the sufficiency of such control measures and eventual compensation mechanisms. Also, in response to a question about what they habitually did in case of sudden chicken deaths only four per cent reported actually having notified authorities in case of sick/dead poultry. Similar findings have been published by other studies [29–34]. Anticipated financial losses due to culling without sufficient compensation, lack of knowledge about how to proceed in notifying authorities, but also social considerations, such as stigma and shame might play a role and need dealing with to overcome avoidance of timely reporting [35–37]. If early notification is a key component of prevention and rapid response, trust in government actions, including compensation measures, is crucial in order to enable pervasive compliance with drastic and economically threatening actions like mass culling in an outbreak situation .
Limitations of the study
A major limitation of the study lies in the small, non-random sample which restricts possibilities to generalize findings from the present data and also, due to lack of power, might have led to underestimation of potential effects. Another clear weakness is the cross-sectional study design which prohibits drawing causal conclusions about the relationships between some of the variables, such as fear and knowledge or knowledge and practices. Finally, self-report on practices are generally vulnerable to recall bias and social desirability tendencies. The face-to-face-interview situation, while enabling full response-rates on all variables as well as participation of poultry workers who lack reading or writing abilities, might have additionally heightened this type of bias in assessing attitudes and behaviours. As for attitudes, however, the very low percentage of respondents reporting compliance with notification procedures indicates that such tendencies were not pervasive.