Clinical presentation at initiation of maintenance dialysis and subsequent survival: A retrospective cohort study

Abstract

Introduction: Clinical practice guidelines increasingly favor a more symptom-driven approach to maintenance dialysis initiation. But, little is known about the clinical presentation at dialysis initiation, such as the different kinds of signs and symptoms present at dialysis initiation, illness acuity at dialysis initiation, and how these aspects of the clinical presentation relate to subsequent survival. Methods: This is a retrospective cohort study of a random sample of veterans who initiated dialysis between 2000 and 2009 at Veterans Affairs (VA) medical centers across the country (N = 1,691). We looked at associations between the clinical presentation (types of signs and/or symptoms and illness acuity) at the time of dialysis initiation and 1-year mortality. Findings: The most common types of signs and/or symptoms at initiation were gastrointestinal (50.9%) and cardiopulmonary (48.7%). The crude mortality rate was 174.2 deaths per 1000 patients per year (95% CI 154.7, 196.1). Among different categories of signs and/or symptoms, only cardiopulmonary signs and/or symptoms were associated with a higher 1-year risk of mortality after initiation (risk ratio (RR) 1.32, 95% CI 1.05, 1.69) in adjusted analyses. Patients who were acutely ill at initiation were more likely to die during the following year as compared with those who initiated dialysis in the outpatient setting (RR 1.57, 95% CI 1.15, 2.21). Discussion: With the exception of cardiopulmonary signs and/or symptoms, most signs and/or symptoms documented around the time of dialysis initiation were not associated with an increased risk of death during the year after initiation. In contrast, whether patients were acutely ill at the time of initiation was strongly associated with an increased risk of death after initiation regardless of the specific signs and/or symptoms present. Limitations of this study include retrospective study design, residual confounding, and lack of generalizability to non-VA settings.