Three hundred and ninety patients between 62 and 101 years of age admitted to a geriatric unit in The Royal Victoria Hospital (RVH), Edinburgh were investigated for the presence of C. difficile, C. difficile was cultured from 100 (26%) patients, on pre-reduced cycloserine-cefoxitin egg-yolk agar. Toxin(s) were detected in the faeces of 34 of these patients with the Techlab^TM ELISA test kit for the detection of C. difficile toxins A and/or B. Toxin(s) were detected in a further 18 patients from whom no C. difficile was detected in culture. A number of possible risk factors associated with C. difficile disease, and relating to medication, antibiotic use and underlying disease were investigated by logistic regression modelling. A two-step predictive model for C. difficile disease was hypothesised. In this two-step disease model, patients made a transition from C. difficile negative (Cdc-) to C. difficile culture positive (Cdc+) and in some cases a further transition from Cdc+ to C. difficile culture and toxin positive (Cdt+) was made. The logistic regression modelling found that the factors significantly associated with the transition from Cdc- to Cdc+ were the origin of the admission (another hospital, nursing home or the community) and the use of cefotaxime (a third generation cephalosporin). Amoxycillin and cephalosporins (other than cefotaxime) were significantly associated with the transition from Cdc+ to Cdt+. Statistical analysis also showed that Cdt+ patients were significantly older than the Cdc+ patients. These findings support the proposed two step model for infection and indicate that different risk factors are significant in each of the two steps.