Antibiotic resistance is a growing problem in all countries. However, because the precise role of community antibiotic prescribing is undefined, practitioners attribute the problem to overuse of antibiotics by others. METHODS: Case-control design with incident urine samples (no sample in previous six months) from the population of Tayside from 1 July 1993 to 31 December 1995. Cases were subjects from whom trimethoprim resistant Gram-negative bacteria (TR) were isolated. Controls were subjects whose samples grew trimethoprim sensitive Gram-negative bacteria (TS) or no bacterial growth (NG). Exposure to risk factors in each group was determined from the MEMO and DARTS record-linkage databases. Logistic regression was used to identify independent risk factors. Demographic variables were age, sex, socio-economic status and diabetes mellitus. Exposure variables (in the previous six months) were hospitalisation and dispensing of the following drugs from community pharmacies: antibiotics, corticosteroids or oestrogens. RESULTS: 9,362 incident urine samples were used: 380 TR , 1073 TS and 7909 NG. The most important risk factors were exposure to trimethoprim (OR 3.08 95%CI 2.48-3.81) or other antibiotics (OR 1.43 CI 1.14-1.80). Additional independent risk factors were age (OR 1.28 for every 10 years after the first decade, CI 1.23-1.35) and prior hospitalisation (OR 1.38 CI 1.07-1.79). However, 35% of the TR group had not been exposed to antibiotics or been hospitalised in the previous six months. CONCLUSION: There was a strong link between trimethoprim resistance and community prescribing of trimethoprim or other antibiotics. However, prior hospitalisation was an independent risk factor and 35% of cases were not associated with either risk factor. These data suggest that antibiotic resistance is multi-factorial.