Peter Parashos1, Michael V Morgan, Harold H Messer. 1. The School of Dental Science, University of Melbourne, 711 Elizabeth Street, Melbourne, Victoria, 3000, Australia. parashos@iprimus.com.au
Abstract
OBJECTIVES: (a) To report on response rate and nonresponse bias of a questionnaire survey of dentists. (b) To make recommendations for future questionnaire survey research in dentistry. METHODS: A questionnaire was mailed to a stratified systematic sample of 908 Australian dentists. The strategy included three mailings, a final telephonic contact, university stationery, paid reply envelopes and personalized correspondence. Nonresponse bias was assessed by comparing responses to a simple 'yes/no' question from each contact (late responders), and by comparing demographic information (nonresponders). RESULTS: The response rate achieved was 87% and there was no evidence of nonresponse bias based on practice location or year of graduation. The cumulative proportions of 'yes/no' responses essentially remained constant after each contact, but significantly more late responders answered in the negative to the test question than did early responders. The telephonic contact aided in the identification of nonparticipants and ineligible units. CONCLUSIONS: The current survey indicates that differences in data between early and nonresponders can occur despite there being no demographic differences. Therefore, assessment of nonresponse bias based on demographic data alone would seem to be insufficient. Questionnaire survey research must first be based on sound sampling techniques, and then on achieving as high a response rate as possible using the many incentives available. Copyright Blackwell Munksgaard, 2005.
OBJECTIVES: (a) To report on response rate and nonresponse bias of a questionnaire survey of dentists. (b) To make recommendations for future questionnaire survey research in dentistry. METHODS: A questionnaire was mailed to a stratified systematic sample of 908 Australian dentists. The strategy included three mailings, a final telephonic contact, university stationery, paid reply envelopes and personalized correspondence. Nonresponse bias was assessed by comparing responses to a simple 'yes/no' question from each contact (late responders), and by comparing demographic information (nonresponders). RESULTS: The response rate achieved was 87% and there was no evidence of nonresponse bias based on practice location or year of graduation. The cumulative proportions of 'yes/no' responses essentially remained constant after each contact, but significantly more late responders answered in the negative to the test question than did early responders. The telephonic contact aided in the identification of nonparticipants and ineligible units. CONCLUSIONS: The current survey indicates that differences in data between early and nonresponders can occur despite there being no demographic differences. Therefore, assessment of nonresponse bias based on demographic data alone would seem to be insufficient. Questionnaire survey research must first be based on sound sampling techniques, and then on achieving as high a response rate as possible using the many incentives available. Copyright Blackwell Munksgaard, 2005.