BACKGROUND: Since chronic kidney disease (CKD) affects 11% of the United States population, and its incidence is rising, experts recommend early referral to nephrologists in the hope that it may delay the onset of end-stage disease and improve survival. However, limitations in the capacity of currently practicing nephrologists may prevent widespread early referral. OBJECTIVE: To examine the relationship between disease progression and timing of nephrology referral. STUDY DESIGN AND DATA COLLECTION: We retrospectively identified 1,553 veterans at the Durham, North Carolina VA hospital between January 1998 and December 1999 who had CKD, defined as two outpatient serum creatinines > or = 1.4 mg/dL at least three months apart. Our endpoint was a composite of progression to the next CKD stage or death. We compared the time to the composite endpoint for each CKD stage and for early CKD (stages 1-3) to advanced CKD (stages 4 and 5) using a Cox proportional hazards model for two groups: those with primary care only (PCP-only) and those with primary and nephrology care (nephrology). RESULTS: Ninety-two percent had hypertension, 52% diabetes, 49% coronary artery disease, and 89% proteinuria. Angiotensin-converting enzyme inhibitors and anti-lipid medications were used by 52% and 39%, respectively. The median number of days spent in each CKD stage and the proportion of each groups reaching the composite endpoint are--stage 1: 1,149 days, 68% of the PCP-only group and 73% of the nephrology group; stage 2: 1,206 days, 60% and 65%; stage 3: 1,158 days, 69% and 63%; and stage 4: 794 days, 86% and 72%. Adjusted survival curves for the composite endpoint were similar between the two groups for CKD stages 1 (HR 1.08 for nephrology versus PCP-only) and 2 (HR 1.20); however for CKD stages 3 (HR 0.80, p < 0.05) and 4 (HR 0.75, p < 0.05), the nephrology group gained 316, 215, and 120 more days of progression-free survival, respectively. LIMITATIONS: The major limitation is difficulty accounting for unmeasured bias in specialty referrals. We were unable to analyze stage 5-to-dialysis due to the small number of individuals with the outcome. CONCLUSION: Our data suggest that an appropriate time for nephrology comanagement of patients with CKD may be stage 3; however, prospective studies are needed to clarify the role and timing of nephrology referral.