Predicting the risk of dialysis and transplant among patients with CKD: a retrospective cohort study

• Published in: American journal of kidney diseases Vol. 52; no. 4; p. 653
• Main Authors: Johnson, Eric S, Thorp, Micah L, Platt, Robert W, Smith, David H
• Format: Journal Article
• Language: English
• Published: United States 01.10.2008
• Subjects: Aged; Chronic Disease; Cohort Studies; Disease Progression; Female; Humans; Kidney Diseases - diagnosis; Kidney Diseases - therapy; Kidney Transplantation; Male; Middle Aged; Outcome Assessment (Health Care) - methods; Prognosis; Proportional Hazards Models; Renal Dialysis; Renal Replacement Therapy - methods; Retrospective Studies; Risk Factors
• ISSN: 1523-6838
• DOI: 10.1053/j.ajkd.2008.04.026

Providers need a reliable way to identify patients with chronic kidney disease (CKD) at the highest risk of progression to end-stage renal disease so they can intervene to slow progression and refer patients to nephrology for comanagement. We developed a risk score to predict the 5-year risk of renal replacement therapy (RRT) in patients with stage 3 or 4 CKD. Retrospective cohort study. Participants were members of a health maintenance organization and met Kidney Disease Outcomes Quality Initiative criteria for stage 3 or 4 CKD during 1999 or 2000: two estimated glomerular filtration rate values of 15 to 59 mL/min/1.73 m(2). Characteristics collected during routine clinical practice. We ascertained the onset of RRT (dialysis or kidney transplantation) using the health maintenance organization databases. Cox regression predicted patient risk of RRT and generated a risk scoring system. 9,782 patients experienced a 3.3% five-year progression to RRT (95% confidence interval, 2.9 to 3.7). Using 6 characteristics (age, sex, estimated glomerular filtration rate, diabetes, anemia, and hypertension), the risk score discriminated the highest risk patients effectively: 19.0% of patients in the highest risk quintile experienced progression, and 0.2% of patients in the lowest risk quintile experienced progression. The c statistic also showed effective discrimination: 0.89 on a scale of 0.5 to 1.0. Predicted and observed risks agreed within 1.0%--effective calibration. We present a range of predicted risk cutoff values from 1% to 20% and their test properties for decision makers' consideration. Characteristics were measured without a protocol. The risk score can help providers identify patients with CKD at the highest risk of progression to improve referral to nephrology for comanagement. A separate risk score for mortality also is needed.