Background: Chronic kidney disease stage 5 (CKD5) with end-stage renal failure (ESRF) is not common but it is expensive to treat, despite advances in technology. Appropriate service provision requires good quality information on the population served and the services provided. In Romania this is in limited supply. Aim: To estimate current and future service needs for CKD5 in Romania. Methods: Desktop research, surveys of a sample of Romanian treatment centres, and mathematical modelling. The baseline renal replacement therapy (RRT) stock was calculated using the capture-recapture method (CRM). The reported % of cases with diabetic nephropathy was compared with the expected % based on population attributable risk (PAR%) and trends in disease precursors. The acceptance rate was estimated using the Impact Fraction method. Needs for numbers of treatment places were calculated from service activity and clinical parameters (stock, acceptance and mortality) in a spreadsheet model. Estimates were made under variant scenarios for two periods: calibration and validation (1997-2006) and projection (2007-2016). Results: In Romania in 1997, the prevalence of chronic kidney disease (CKD) stages 1 to 5 was estimated at 1,222.5 per million population (pmp). There was a strong association between CKD5 and diabetes plus hypertension (OR =7.73 [95% CI: 0.99 to 60.38]). Increasing trends in age, diabetes and hypertension suggest an increasing incidence of CKD5; but a downward trend in smoking will offset this. Reported national RRT stock in Romania was 139 pmp in 1997 and 250 pmp in 2003. The CRM suggested that the Centre Questionnaire and EDTA data covered 71% of the total stock at baseline (2,995 patients). The % of CKD5 on RRT reported to have diabetic nephropathy was 10%, much less than the PAR%-based estimate of 30.5%. Acceptance and stock varied between the 14 centres from 11 to 85 pmp, and 112 to 222 pmp respectively. In the calibration period, two of the scenarios tried gave figures for 2006 of 238 and 251, close to the observed figure of 250. (Other scenarios gave figures from 238 to 721). Projecting the two chosen scenarios to 2016 gave 239 and 276 pmp. (Other scenarios gave up to 1,940 pmp.) Conclusions: The information base for this modelling exercise was weak. However there appears to be under-provision of care for CKD5 in Romania, particularly for diabetic nephropathy.