Validity of extrapolating nonresponse bias from mail questionnaire follow-ups.