Response bias in a study of general practice.

Response bias in quality of care research is an important, but largely neglected concern. Differences between health care professionals who consent to participate in research and those who do not may distort the conclusions and prevent the results being generalizable. This is particularly likely when response rates are low, as they often are in studies evaluating primary health care. The present study outlines a method for examining this important area. Fifty-six general practitioners who consented to participate in an observational study of general practice were compared with 52 doctors who declined to participate in the research. Comparisons were made of general characteristics including age, sex, practice size and postgraduate qualifications, as well as attitudes toward their role. This last analysis was deemed particularly important, as the attitudes expressed may have affected the behaviour of the doctor in the consultation, and therefore outcomes such as accurate diagnosis, compliance and satisfaction. Such outcomes are often the object of study in quality of care research. The only significant difference to emerge in the present study was that non-consenters were more strongly in favour of a medical system based on free enterprise and fee for service (t83 = 2.3P less than 0.05). No systematic differences were found on other general characteristics or attitudes relating to patient care. Response bias using the stated variables was therefore considered to be minimal. The results are discussed in terms of strategies aimed at increasing response rates in quality of care research.