BACKGROUND: There is now clear evidence that tight control of blood glucose and blood pressure significantly lowers the risk of complications in both type I and type II diabetes. Although there is evidence that primary care can be as effective as secondary care in delivering care for people with diabetes, standards in primary care are variable. Previous studies have shown that practice, patient or organisational factors may influence the level of care of patients with diabetes. However, these studies have been conducted in single geographical areas and involved only small numbers of practices. AIM: To determine the standard of diabetes care in general practice and to determine which features of practices are associated with delivering good quality care. DESIGN OF STUDY: A questionnaire survey and analysis of multi-practice audit data. SETTING: Three health authorities in England, comprising 169 general practices. METHOD: This study was conducted with a total population of 1,182,872 patients and 18,642 people with diabetes. Linkage analysis was carried out on data collected by a questionnaire, routinely collected health authority data, and multi-practice audit data collected by primary care audit groups. Practice annual compliance was measured with process and outcome measures of care, including the proportion of patients who had an examination of their fundi, feet, blood pressure, urine, glycated haemoglobin, and the proportion who had a normal glycated haemoglobin. RESULTS: Median compliance with process and outcome measures of care varied widely between practices: fundi were checked for 64.6% of patients (interquartile range [IQR] = 45.3-77.8%), urine was checked for 71.4% (IQR = 49.7-84.3%), feet were checked for 70.4% (IQR = 51.0-84.4%), blood pressure for 83.6% (IQR = 66.7-91.5%), and glycated haemoglobin was checked for 83.0% of patients (IQR = 69.4-92.0%). The glycated haemoglobin was normal in 42.9% of patients (IQR = 33.0-51.2%). In multiple regression analysis, compliance with measures of process of care were significantly associated with smaller practices, fundholding practices, and practices with a recall system. Practices with more socioeconomically deprived patients were associated with lower compliance with most process measures. Practices with a greater proportion of patients attending hospital clinics had lower compliance with process and outcome measures. Being a training practice, having a diabetes mini-clinic, having more nurses, personal care, and general practitioner or nurse interest in diabetes were not associated with compliance of process or outcome of care. CONCLUSIONS: Despite recent evidence that complications of diabetes may be delayed or prevented, this study has highlighted a number of deficiencies in the provision of diabetes care and variations in care between general practices. Provision of high quality diabetes care in the United Kingdom will present an organisational challenge to primary care groups and trusts, especially those in deprived areas.