<i>Background: </i>The annual gross mortality of end-stage renal failure patients remains very high (approximately 15-20%) leading some to question the wisdom of accepting patients with limited prognosis for dialysis. We have reviewed the demographic and clinical characteristics of patients who died within a year of commencing renal replacement therapy (RRT) over a 5-year period to establish whether these patients could be identified at the start of therapy. <i>Methods: </i>Case notes of patients who died within 1 year of commencing RRT between 1st April, 1991 and 31st March, 1996 were reviewed. Comorbidity at the start of dialysis was used to classify patients into high-, medium- and low-risk groups using two published scales to determine whether either graded a high proportion of deaths as high risk. Factors such as age, social circumstances, cause of death, renal diagnosis and mode of dialysis were also analysed. <i>Results: </i>17.5% of patients commencing RRT died in the first year. Not all of these patients could be identified as high risk by comorbidity assessment at dialysis initiation - 50% of patients who died were classified by one scale as medium risk. Age did not clearly predict outcome, as 42% of patients who died were less than 65 years old. <i>Conclusion: </i>These data suggest that it is difficult to use current risk stratifications to accurately predict outcome on RRT.