Objective:The aim of this study was to examine the sociodemographic data of diabetic patients referred to our clinic and to correlate these with characteristics of their individual general practitioners. How these factors affect the interaction between patients, general practitioners and a hospital diabetic clinic was evaluated.Design:Prospective recruitment of consecutive referrals.Setting:The diabetic clinic of a teaching hospital located in the inner city suburbs of Sydney.Patients:Two hundred and forty‐six patients with diabetes (10% insulin dependent), aged 20–86 years, participating in our Shared Care Project, a randomised controlled study on various methods of following up diabetic patients.Interventions:Data for the study were gathered by interview and questionnaire during first assessment at the clinic.Main outcome measures:Clinical and sociodemographic characteristics of patients; the location of general practitioners, their diagnostic equipment and type of practice; and the level of detail recorded in referral letters from general practitioners.Results:Diabetic patients of migrant background who cannot speak English are older and have less formal education. They see their doctors more often and seek out general practitioners who speak their language, even if it means travelling longer distances. Patients referred from 24 hour medical centres are younger, more educated and have less contact with their doctors both in duration and frequency of visits when compared with patients whose general practitioners operate in conventional sole or partnership practices. Many patients have more than one general practitioner, making communication with hospitals difficult. In their referral letters, general practitioners usually concentrate on hypertension, hypercholesterolaemia and metabolic aspects of diabetes, but under‐emphasise diabetic complications; they make insufficient use of measurement of the glycosylated haemoglobin level to assess diabetic control while over‐using glucose tolerance testing in making the diagnosis. The referral letter often does not contain enough information to help clinical decision making.Conclusions:Demography of patients and characteristics of general practitioners are important factors which can affect their interaction with public hospitals. Bearing in mind the work load of general practitioners and the diverse nature of patients, hospitals must implement systems which make it easier for general practitioners and patients to interact with them.