Abstract
BACKGROUND: Accurate prognostic models could inform treatment decisions for older adults with end-stage renal disease who are considering dialysis and might identify patients more appropriate for conservative care or hospice. METHODS: In a cohort of patients aged ≥ 67 years commencing dialysis in the United States between January 1, 2008 and June 30, 2009, we compared the discrimination of three existing instruments (the Liu index; the French Renal Epidemiology and Information Network score; and hospice eligibility criteria) for the prediction of 6-month mortality. We estimated the odds of death associated with each prognostic index using logistic regression with and without adjustment for age. Predictive indices were compared using the concordance ("c")-statistic. RESULTS: Of 44,109 eligible patients, 10,289 (23.3%) died within 6 months of dialysis initiation. The c-statistic for the Liu, Renal Epidemiology and Information Network, hospice eligibility criteria, and combined Liu/hospice eligibility criteria scores without and with age were 0.62/0.65, 0.63/0.66, 0.65/0.68, and 0.68/0.70, respectively. Discrimination was poorer at older ages, especially for the Liu and Renal Epidemiology and Information Network scores. Although sensitivity was poor, a Renal Epidemiology and Information Network score ≥ 9 or an hospice eligibility criteria ≥ 3 had relatively high specificity. CONCLUSIONS: Existing prognostic indices based on administrative data perform poorly with respect to prediction of 6-month mortality in older patients with end-stage renal disease commencing dialysis.