Send to

Choose Destination
See comment in PubMed Commons below
Clin J Am Soc Nephrol. 2014 Apr;9(4):673-81. doi: 10.2215/CJN.07630713. Epub 2014 Mar 20.

Dialysis versus nondialysis in patients with AKI: a propensity-matched cohort study.

Author information

Renal, Electrolyte, and Hypertension Division,, †Center for Clinical Epidemiology and Biostatistics, and, ‡Department of Biostatistics and Epidemiology, Perelman School of Medicine at the University of Pennsylvania, Philadelphia, Pennsylvania;, §Division of Nephrology, University of Michigan, Ann Arbor, Michigan, ‖Penn Data Store Team, Information Services at the University of Pennsylvania, Philadelphia, Pennsylvania.



The benefit of the initiation of dialysis for AKI may differ depending on patient factors, but, because of a lack of robust evidence, the decision to initiate dialysis for AKI remains subjective in many cases. Prior studies examining dialysis initiation for AKI have examined outcomes of dialyzed patients compared with other dialyzed patients with different characteristics. Without an adequate nondialyzed control group, these studies cannot provide information on the benefit of dialysis initiation. To determine which patients would benefit from initiation of dialysis for AKI, a propensity-matched cohort study was performed among a large population of patients with severe AKI.


Adults admitted to one of three acute care hospitals within the University of Pennsylvania Health System from January 1, 2004, to August 31, 2010, who subsequently developed severe AKI were included (n=6119). Of these, 602 received dialysis. Demographic, clinical, and laboratory variables were used to generate a time-varying propensity score representing the daily probability of initiation of dialysis for AKI. Not-yet-dialyzed patients were matched to each dialyzed patient according to day of AKI and propensity score. Proportional hazards analysis was used to compare time to all-cause mortality among dialyzed versus nondialyzed patients across a spectrum of prespecified variables.


After propensity score matching, covariates were well balanced between the groups, and the overall hazard ratio for death in dialyzed versus nondialyzed patients was 1.01 (95% confidence interval, 0.85 to 1.21; P=0.89). Serum creatinine concentration modified the association between dialysis and survival, with a 20% (95% confidence interval, 9% to 30%) greater survival benefit from dialysis for each 1-mg/dl increase in serum creatinine concentration (P=0.001). This finding persisted after adjustment for markers of disease severity. Dialysis initiation was associated with more benefit than harm at a creatinine concentration ≥ 3.8 mg/dl.


Dialysis was associated with increased survival when initiated in patients with AKI who have a more elevated creatinine level but was associated with increased mortality when initiated in patients with lower creatinine concentrations.

Comment in

[Indexed for MEDLINE]
Free PMC Article
PubMed Commons home

PubMed Commons

How to join PubMed Commons

    Supplemental Content

    Full text links

    Icon for HighWire Icon for PubMed Central
    Loading ...
    Support Center