In addition, our data set may have been biased toward
birds that were easily available to sample. Because active
cases of H5N1 are rare we queried the families whose
flocks we sampled for currently or recently sick poultry in
addition to sampling healthy appearing poultry. We also
sampled a single dead chicken found in a garbage pile.
This strategy could have resulted in sampling bias if
people at some sites failed to report sick poultry, if only
some sites had a practice of slaughtering sick poultry, or
the practice of slaughtering sick poultry was correlated
with disposal method of dead birds. We were able to test
for two of these potential sources of bias and found that
the practice of killing sick birds was the norm (95%) and
the tendency did not vary among sites (t = 0.1868, df = 15,
p = 0.854). People who killed sick birds did not dispose of
dead birds differently than people who did not (Fisher’s
exact test: p = 0.4527). The dead chicken in which H5N1
was detected was not included in analyses to avoid
potential bias