An increased incidence of diarrhoea due to Clostridium difficile (CDD) at the Northern General Hospital, Sheffield, prompted an investigation into antibiotic use on the renal, medical and geriatric wards, those mostly affected. For the first half of 1997 affected patients on these wards were identified and data collected as to which antimicrobials they had taken between admission and diagnosis. Rates were then calculated of the number of affected patients on a drug over the quantity prescribed (N/1000 x defined daily doses). These were expressed as rate ratios (RR) over the rate of ampicillin/amoxycillin. A corrective procedure was also applied to the RR in an attempt to compensate for bias from the effect of drug combinations. Additionally, quarterly data were collected of numbers of patients with CDD and the amount of cefotaxime, ceftriaxone and cefuroxime issued between January 1996 and September 1997. Those drugs with the greatest association with CDD were cefotaxime (RR = 27.5, P < 0.000001), ceftriaxone (15.1, 0.00004), cefuroxime (8.6, < 0.000001) and ceftazidime (6.4, 0.00008). Several other drugs had increased uncorrected RR (P < 0.05). However, after correction for drug combinations, only the above four drugs remained with significantly increased RRs of 16.9, 8.6, 5.3 and 4.8, respectively. The time course of the outbreak showed a correlation with the use of cefotaxime and ceftriaxone. The incidence peaked in the first months of 1997, following greatest use of cefotaxime in the last quarter of 1996 and of ceftriaxone in the first quarter of 1997. This study confirms the association of cephalosporin and particularly cefotaxime usage with CDD incidence and shows the need to review the use of these drugs, especially in the treatment of respiratory infection. The method is useful for looking at local effects of prescribing with a view to controlling outbreaks of antibiotic resistant bacteria.