RT Journal Article SR Electronic T1 Surveying general practitioners: does a low response rate matter? JF British Journal of General Practice JO Br J Gen Pract FD British Journal of General Practice SP 91 OP 94 VO 47 IS 415 A1 L Templeton A1 A Deehan A1 C Taylor A1 C Drummond A1 J Strang YR 1997 UL http://bjgp.org/content/47/415/91.abstract AB BACKGROUND: Primary care has long been of interest to policy research. Recently, there is evidence to suggest that it is becoming more difficult to encourage GPs (general practitioners) to participate in surveys. As low response rates can introduce bias into survey results, it is important to study the effects of non-response. AIM: To assess the validity of a response rate of 44% obtained in a national postal study of GPs surveyed about their work with alcohol-misusing patients by assessing the extent of any non-response bias. METHOD: A telephone survey of 148 GPs who had not responded to repeated mailings of a postal questionnaire was undertaken. In addition to personal and practice structure characteristics, the GPs were asked three questions taken from the original questionnaire about their work with alcohol-misusing patients. RESULTS: Of the 148 GPs telephoned, 64 responded to the telephone questionnaire in full; all had previously failed to respond to the postal questionnaire. Younger GPs were more likely to respond to both the national postal and telephone surveys, but more so to the latter. Telephone responders were more likely to be GPs in a single-handed practice. The work of GPs with alcohol-misusing patients highlighted differences between the two response groups. Male telephone responders were found to be identifying a significantly higher average of alcohol misusers than male postal responders. Telephone responders were more likely to feel trained in treating alcohol misuse and to feel better supported to deal with this patient group. CONCLUSION: Some significant differences were identified, indicating the presence of non-response bias. A low response rate need not affect the validity of the data collected, but it is still necessary to test for non-response effects and make corrections to the original data in order to maximize validity.