375 Khunti.qxd

Background. Different methods have been used to determine the prevalence and treatment of diabetes. Despite the large number of studies, previous estimations of prevalence and treatment have been carried out on relatively small numbers of patients, and then in only a few practices in single geographical regions. Aim. To investigate the feasibility of collating data from multi-practice audits organized by primary care audit groups in order to estimate the prevalence and treatment of patients with known diabetes, and to discuss the methodological issues and reasons for variation. Method. A postal questionnaire survey of all primary care audit groups in England and Wales that had conducted a multi-practice audit of diabetes between 1993–1995. Prevalence rates and patterns of diabetic care were compared with other community-based surveys of known diabetes from 1986–1996 identified on MEDLINE. Results. Twenty-five (43%) audit groups supplied data from multi-practice audits of diabetes. Seven (28%) multi-practice audits involving 259 practices fulfilled the inclusion criteria for prevalence estimation. The overall prevalence of diabetes based on a population of 1 475 512 patients was 1.46% (range between audit groups = 1.18% to 1.66%; c2 = 308; df = 6; P<0.0001). Male to female ratio was 1.15:1. Treatment of diabetes could be ascertained for 10 (40%) audit groups comprising 319 practices. Of these, 23.4% (range = 16.5%–27.4%) were controlled by diet, 48.5% (range = 43.6%–55.8%) were prescribed oral hypoglycaemic drugs, and 28.2% (range = 25.0%–32.4%) were treated with insulin. There were significant variations between audit groups in treatment pattern (c2 = 250; df = 18; P<0.0001). Conclusion. Prevalence and treatment rates of diabetes and other chronic diseases can be assessed and compared using data from multi-practice audits. Collation of audit data could improve the precision of quantitative estimates of health status in populations. A standard method of data recording and collection may provide a new approach that could considerably improve our ability to monitor disease and its management.