Skip to main content

Death and recovery of kidney function among patients continued on dialysis after discharge from hospital stays complicated by acute kidney injury: a cohort study

Clark EG, Hiremath S, Sood MM, Wald R, Garg AX, Silver SA, van Walraven C. Am J Kidney Dis. 2019; Jun 17 [Epub ahead of print]. DOI: 10.1053/j.ajkd.2019.03.429.

Rationale and Objective — Some patients who survive an inpatient episode of acute kidney injury (AKI) treated with dialysis (AKI-D) will require long-term outpatient dialysis. Factors that predict whether patients with AKI-D die or recover kidney function sufficiently to stop dialysis during the following year are poorly described and were the focus of this study.

Study Design — Population-based retrospective cohort study.

Setting and Participants — 2,771 individuals, identified using population-based data sets, who required outpatient dialysis after an inpatient episode of AKI-D in Ontario, Canada, between January 1, 2008, and September 30, 2015.

Predictors — Age, comorbid conditions, discharge disposition, estimated glomerular filtration rate, and urinary albumin-creatinine ratio (UACR) before AKI.

Outcomes — Death and kidney function recovery sufficient to stop dialysis at 1 year after hospital discharge.

Analytical Approach — Death and recovery submodels were created using proportional hazards survival analysis. A model-based point system was to generate quartiles of risk. Models were internally validated.

Results — After 1 year, 521 of 2,771 (18.8%) patients were alive and no longer requiring dialysis, 736 (26.6%) had died, 14 (0.5%) had undergone kidney transplantation, and 1,500 (54%) were receiving ongoing maintenance dialysis. Factors independently associated with a higher 1-year probability of death were older age, higher modified Charlson comorbidity score, history of cancer, and discharge disposition. Higher baseline estimated glomerular filtration rate and lower baseline UACR were independently associated with a higher 1-year probability of no longer requiring dialysis. The combined model's point system classified patients into distinct groups according to their probability of death and need for dialysis with excellent calibration.

Limitations — Baseline UACR data were missing for nearly half of all patients. External validation is required to establish this model's clinical utility.

Conclusions — The death and recovery after AKI model categorizes patients into clinically distinct groups according to their future likelihood of death and need for long-term dialysis.

View full text