Patients with Clostridium difficile-associated diarrhoea (CDAD) may initially develop symptoms in the community and be subsequently diagnosed at hospital admission. At the present time there is no national surveillance system and no standardized case definition of CDAD in the USA, and baseline data on the incidence and epidemiology of CDAD are scarce. The objective of this study was to report the incidence of CDAD at a tertiary care hospital, and to determine the epidemiology of cases diagnosed within 48h of hospital admission, compared with cases of nosocomial CDAD diagnosed 48h or more after hospitalization. The average incidence was 4.0 cases/10 000 patient-days for CDAD on admission and 7.0 cases/10 000 patient-days for nosocomial CDAD. A significant difference was observed in CDAD rates on admission compared with nosocomial CDAD rates (P=0.017), but no differences were observed over time for either rate. Overall, 44% of cases had CDAD on admission and 56% of cases had nosocomial CDAD. Fifty-six (62%) patients with CDAD on admission had been admitted to the same hospital and 24 (27%) had been admitted to another hospital within the previous 90 days. Only eight (9%) patients had not been exposed to any healthcare services in the 90 days preceding hospital admission. A standardized case definition of healthcare-associated CDAD should include previous hospitalizations. Admitting physicians should consider C. difficile in the differential diagnosis of patients admitted with diarrhoea, with or without a history of admission to healthcare facilities.