A clinical score to predict 6-month prognosis in elderly patients starting dialysis for end-stage renal disease

Abstract

Aim: The aim of this study was to develop and validate a prognostic score for 6-month mortality in elderly patients starting dialysis for end-stage renal disease.

Methods: Using data from the French Rein registry, we developed a prognostic score in a training sample of 2500 patients aged 75 years or older who started dialysis between 2002 and 2006, which we validated in a similar sample of 1642 patients. Multivariate logistic regression with 500 bootstrap samples allowed us to select risk factors from 19 demographic and baseline clinical variables.

Results: The overall 6-month mortality was 19%. Age was not associated with early mortality. Nine risk factors were selected and points assigned for the score were as follows: body mass index <18.5 kg/m2 (2 points), diabetes (1), congestive heart failure stages III to IV (2), peripheral vascular disease stages III to IV (2), dysrhythmia (1), active malignancy (1), severe behavioural disorder (2), total dependency for transfers (3) and unplanned dialysis (2). The median score was 2. Mortality rates ranged from 8% in the lowest risk group (0 point) to 70% in the highest risk group (> or =9 points) and 17% in the median group (2 points). Seventeen percent of all deaths occurred after withdrawal from dialysis, ranging from 0% for a score of 0-1 to 15% for a score of 7 or higher.

Conclusions: This simple clinical score effectively predicts short-term prognosis among elderly patients starting dialysis. It should help to illuminate clinical decision making, but cannot be used to withhold dialysis. It ought to only be used by nephrologists to facilitate the discussion with the patients and their families.

Figure 1

Calibration curve: observed and predicted* 6-month mortality in the validation sample by group of predicted risk (Hosmer-Lemeshow test)

Figure 2

Survival by point score groups in the training sample and the validation sample^o

Figure 2

Survival by point score groups in the training sample and the validation sample^o

Figure 3

Distribution of patients in the validation sample, by score strata, according to vital status and whether or not they were withdrawn from dialysis within 6 months after starting dialysis^{o o} “mean” frequency from 5 imputed datasets (multiple imputation for missing data)