Despite the favourable response of thrombotic thrombocytopenic purpura/haemolytic uraemic syndrome (TTP/HUS) to plasma exchange, an early level of mortality persists. Non-response has been associated with a low frequency of exchange. The Rose index of TTP/HUS severity, occasionally used to predict the response of TTP/HUS to plasma exchange, remains unsatisfactory. The purpose of this study was to develop a new index predicting response of TTP/HUS to plasma exchange and to compare it with the Rose index. Retrospective analysis of 171 cases of TTP/HUS from 39 apheresis units across Canada between 1980 and 2001 was conducted. Logistic regression analysis was used to derive a model predicting 6-month mortality from presenting characteristics. The reduced model contained age >40 years, haemoglobin <9.0 g/dl and the presence of a fever at presentation. Gender, platelet count, creatinine and neurological signs were not part of the final model. This model predicted 13.4% of outcome variance. Predictive scores of 0, 2, 4 and 6 correlated with 6-month mortality rates of 12.5%, 14.0%, 31.3% and 61.5% respectively in our source population. This simple model may help identify those patients who would benefit from higher treatment intensity.