Is there an estimation bias in occupational health and safety surveys? The mode of administration and informants as a source of error