A predictor finding study found patient-reported outcomes improve the prediction of mortality of incident dialysis patients

Introduction:
Prognostic models for mortality in patients receiving dialysis primarily use clinical predictors like age, comorbidities and laboratory markers. Studies in other fields suggest that patient-reported outcomes (PROs), like pain and fatigue, can be predictors of survival. Therefore, we aimed to assess the added value of PROs to predict mortality in incident dialysis patients.

Methods:
Data from NECOSAD (1956 individuals) were used, and analyses were replicated in the EQUAL (415 individuals) and NECOSAD 65+ (862 individuals) studies. A base model for two-year mortality containing clinical predictors was extended using PROs (mental component score, physical component score, general health perception, depressive symptoms, number of symptoms, symptom burden, fatigue and pain). Logistic regression was used, and the added predictive performance of the PROs was evaluated using the area under the curve (AUC), measures of calibration, Brier score, likelihood ratio tests, reclassification tables, net reclassification indices, integrated discrimination improvements, and decision curve analyses. We also examined different combinations of predictors, and each PRO individually.

Results:
Within two years, mortality rates were 22.9%, 24.3%, and 37.1% in NECOSAD, EQUAL, and NECOSAD 65+, respectively. The base model yielded optimism-corrected AUCs of 0.806, 0.781 and 0.699, which improved to 0.826, 0.878 and 0.746 after adding the PROs. Improvement of the calibration, Brier scores, and comparative measures confirmed their predictive value. The mental and physical component score, and symptom burden had the most consistent strong performance across all cohorts.

Conclusions:
PROs improved prognostic models for mortality of patients receiving incident dialysis, even when added to an already well-performing model of clinical predictors.