To evaluate the management of severe dysentery cases in in-patient facilities during an epidemic of Shigella dysenteriae type 1 (Sd1), and to identify the factors associated with the risk of death, we conducted a prospective cohort study in 10 Rwandese hospitals between September and December 1994. Data were obtained from 849 cases admitted to hospitals with diarrhoea and visible blood in stools. The proportion of patients with persistent bloody diarrhoea was 51.0% at treatment day 3 and 27.9% at treatment day 5. At discharge, 79.9% had improved or were cured. The case fatality ratio was 13.2%, higher for patients treated with nalidixic acid than for those treated with ciprofloxacin (12.2% vs. 2.2%, RR = 5.80, 95% CI = 0.83-40.72). In a logistic regression model three risk factors were significantly associated with an increased risk of death during hospitalization: severe dehydration on admission (adjusted OR = 2.79, 95% CI = 1.46-5.33), age over 50 (adjusted OR vs. 5-49 age group = 3.22, 95% CI = 1.70-6.11) and prescription of nalidixic acid (adjusted OR vs. ciprofloxacin = 8.66, 95% CI = 1.08-69.67). Those results were consistent with reported high levels of resistance of Sd1 to the commonest antibiotics, including nalidixic acid. Patients belonging to groups with a higher risk of dying should be given special medical attention and supportive care. In areas of high resistance to nalidixic acid, severe cases of dysentery should be treated with fluoroquinolones in order to reduce the mortality associated with these epidemics.