Abstract: FR-PO051

Slope-Based Staging Outperforms KDIGO Staging for Assessing Inpatient Mortality Risk with AKI

Session Information

  • AKI Clinical: Predictors
    November 03, 2017 | Location: Hall H, Morial Convention Center
    Abstract Time: 10:00 AM - 10:00 AM

Category: Acute Kidney Injury

  • 003 AKI: Clinical and Translational

Authors

  • Warnock, David G., UAB, Birmingham, Alabama, United States
  • Bojan, Mirela, Necker-Enfants Malades University Hospital, Paris, France
  • Wanner, Christoph, University Hospital, Wuerzburg, Germany
  • Agarwal, Anupam, University of Alabama at Birmingham, Birmingham, Alabama, United States
  • Froissart, Marc, CHUV / UNIL, Lausanne, Switzerland
Background

The association between AKI severity and mortality is well described, but current staging criteria have not be optimized for risk assessment or developing prognostic models for specific outcomes

Methods

Confidence limits (CI) for individual patient serum creatinine (sCr) trajectories were developed using an adaptive Bayesian approach. AKI episodes were defined by sCr excursions above the CI, with a) ≥3 sCr values between baseline and peak; b) ≥6 hours between peak and baseline; and c) sCr increase ≥0.3 mg/dL. Specific baseline sCr values were defined for each as the minimal sCr value immediately preceding each AKI episode. Survivor functions, and relative Integrated Discrimination Improvement (rIDI) were done with Stata version 14.1.

Results

All adult 1st admissions were reviewed for FY2010-2013 at UAB Hospital, for patients with ≥3 sCr determinations, length of stay ≥1 & <31 days, excluding patients with ESRD before admission; 35, 079 patients with 1,544 inpatient deaths. AKI staging was done with KDIGO criteria, and slope-based staging for patients with AKI episodes. Survivor functions (Figure) show better discrimination between AKI stages with slope-based criteria compared to KDIGO criteria: rIDI=15.5% (95% CI: 13.5%–17.6%, P<0.001); C-statistics of 0.7694 vs. 0.7226, and AUC/ROC 0.8616 vs. 0.8551 (P=0.012).

Conclusion

Slope-based AKI staging outperforms KDIGO staging criteria for describing the association of AKI severity with inpatient mortality at UAB Hospital. Sloped-based staging is compatible with real-time, dynamic risk assessment than KDIGO staging, and also recognizes multiple AKI episodes during a single admission, with baseline sCr defined for each AKI episode. KDIGO staging is based on the absolute inclrease in sCr with reference to a prior baseline sCr, and does not recognize multiple AKI episodes.

Survivor Functions for AKI Staging by a) KDIGO criteria and b) Slope-Based criteria

Funding

  • NIDDK Support