Background: Currently, most chronic kidney disease (CKD) classifications identify patients at different stages of CKD but do not identify risk ofprogression or adverse outcome. This analysis aims to describe associationsbetween baseline characteristics and the evolution of estimated glomerular filtration rate (eGFR) and identify threshold values for clinical parameters that maximally discriminate progression to renal replacement therapy (RRT) in a referred cohort of patients with CKD stages 3-5. Design and methods: A longitudinal mixed-effect model was used to determine annualized estimated change in eGFR and classification tree analysis to identify threshold values that maximally discriminate progression to RRT. Results: A total of 1316 patients were available for analysis with median follow-up of 33 months(interquartile range 20-60). Mixed model analysis suggested that the underlying diagnoses of autosomal dominant polycystic kidney disease and diabeticnephropathy exhibited on average a 2.7 (0.3) and 0.7 (0.3) ml/min/year faster rate of decline in eGFR, respectively, compared to those patients with biopsy-proven glomerulonephritis. In the regression tree analysis, we attempted to identify threshold values for clinical parameters that maximally discriminate progression to RRT. eGFR 424 ml/min was the first ranked discriminator, diastolic blood pressure appeared in the second and fourth rounds, eGFR appeared again in the third round together with cholesterol and systolic blood pressure, with basal metabolic index in the fourth. Conclusion: This analysis highlights risk factors for progressive kidney disease and demonstrates the variability in evolution of eGFR across the cohort as well as the importance of underlying renal disease type on the progression of CKD.