The frequency of antibiotic-associated diarrhoea (AAD) and Clostridium difficile-associated diarrhoea (CdAD) was prospectively determined in a population of 2462 patients recruited from five Swedish hospitals, including divisions for infectious diseases, orthopaedics, surgery, geriatrics, nephrology and internal medicine. AAD developed in 4.9% of the treated patients. Faecal samples were obtained from 69% of patients with AAD and 55.4% were positive for C. difficile cytotoxin B. The frequency of AAD varied from 1.8 to 6.9% at the participating centres (P < 0.001). The frequency of AAD also varied considerably between medical disciplines and wards within different hospitals and was highest in the nephrology and geriatric units (6.7 and 7.1%, respectively). There was no difference in frequency of AAD when analysed with respect to gender or age. Medical interventions (laxative treatment, endoscopy and abdominal surgery) or presence of one concomitant disease (diabetes, malignancy, chronic renal disease and inflammatory bowel disease) did not significantly affect the frequency of AAD, whereas patients suffering from two or more of these illnesses had significantly (P = 0.001) higher frequencies of AAD. Patients treated with antibiotics for 3 days had a significantly (P = 0.009) lower frequency of AAD than those treated for longer periods. Treatment with cephalosporins, clindamycin or broad-spectrum penicillins was associated with an increased risk of AAD. With specimens from one centre, 62.5% of tested patients with AAD and 33.8% of asymptomatic patients were positive for cytotoxin B. Although C. difficile cytotoxin B in stool samples was significantly associated with AAD (P = 0.003), the causal relationship with diarrhoea is not always evident.