OBJECTIVE To determine the cause of an outbreak of acute illness and death in a long-term hemodialysis unit. METHODS A retrospective cohort and case-control study of patients receiving hemodialysis and a laboratory study of a model deionization system to purify water for hemodialysis. METHODS An outpatient hemodialysis unit of a university hospital. METHODS 12 patients who became severely ill after hemodialysis treatment and 20 patients who did not become ill after receiving hemodialysis treatment in the same unit. METHODS Medical and dialysis unit records were reviewed to identify and characterize cases. Fluids for dialysis were tested for toxic substances, and fluoride was measured in patients' serum. Resistivity and fluoride were measured in effluent from a model deionization system operated in the same way as the system associated with illness. RESULTS During five consecutive hemodialysis shifts, 12 of 15 patients receiving dialysis treatment in one room became acutely ill, with severe pruritus, multiple nonspecific symptoms, and/or fatal ventricular fibrillation (3 patients). None of 17 patients treated in the adjacent room became ill (P < 0.0001). Death was associated with longer hemodialysis time and increased age compared with other patients who became ill. Serum concentrations of fluoride in the sick patients were markedly increased to as high as 716 mumol/L, and the source of fluoride was the temporary deionization system used to purify water for hemodialysis only in the affected room. Operation of a model deionization system showed how fluoride was adsorbed and then displaced in a massive efflux. CONCLUSIONS Because deionization systems are used widely in hemodialysis and can cause fatal fluoride intoxication, careful design and monitoring are essential.