Objective: To evaluate the influence of late referral to nephrology of the patients with chronic renal failure in the morbimortality of the patients who start hemodialysis. SUBJETS AND METHODS: There were included in the study the patients who started hemodialysis (HD) as first form of treatment, and that survived at least three months in both hospitals of reference of the province of Huesca from january 1990 to december 2001. Patients who started HD after acute renal failure were excluded. Clinical and analytical data were determined for each patient at the start of HD and during the follow-up. Early (ER) and late referral (LR) were defined by the time of first nephrology encounter greather than or less than 4 months respectively, before HD initiation. Morbidity analysis (using multiple linear regression with rate of days of hospitalization as dependent variable) and global and anual during the first three years of follow-up survival analysis (using Cox proportional hazards regression) were carried out., Results: A total of 139 patients (78%) started HD in the ER group and 39 (22%) in LR group. Mean follow-up was similar in both (ER = 34.43 +/- 25.5 months; LR = 34.42 +/- 28.37 months). At the start of dialysis LR was associated to higher proportion of temporary catheters, lower level of hematocrit and albumin, higher comorbidity and higher levels of urea and creatinine. Risk factors selected by the model in the morbidity analysis were index of comorbidity (CI), late referral, serum albumin, urea reduction ratio (URR) and hematocrit (R2 = 0.334, F = 16.97, p < 0.005). The final equation of regression was: Rate of hospitalization's days = 101.12 + (2.45 x CI) - (12. 11 x LR) - (11.57 x Alb.) - (0.43 x PRU) - (0.83 x Hto). Variables selected by Cox's regression model that were associated with survival throughout complete follow-up were hematocrit (RR = -0,207, CI 95% 0.726-0.910, p < 0.0005), index of comorbidity (RR = 0,265, CI 95% 1.066-1.594, p = 0.007), PRU (RR = - 0,059, CI 95% 0.893-0.996, p = 0.038) and type of dialysis membrane (RR = 0,771, Cl 95% 0.260-0.822, p = 0.007). Nevertheless, in successive models fitting after 12, 24 and 36 months of follow-up the variable LR influenced in an independent way survival first two years, losing his significance later., Conclusion: In our study patients of the group LR presented a worse clinical and metabolic situation at the beginning of the HD. Later there was demonstrated in this group a higher long-term morbidity and a lower survival the first two years.