In these cases the substitutions may have introduced some bias
to the sample although the observable characteristics of the HCEs,
which refused to take part (apparent number of beds, locations
and type) were not otherwise out of line with the sampled population.