Emergency Department Prediction Model for Acute Kidney Injury

San Diego—More than 50% of community-acquired acute kidney injury (AKI), an independent risk factor for mortality, are identified in the emergency department (ED). The ED is the primary route for acute hospital admissions. The facilitation of prompt medical assessment and treatment depends on accurate stratification of risk. David A. Foxwell, MD, and colleagues in Wales, United Kingdom, recently conducted a study designed to derive a hospital front-door model for predicting AKI in the ED (ED-AKI). Results of the study were reported during a poster session at the AKF Kidney Week 2018 in a poster titled Derivation of a Prediction Model for Emergency Department AKI.

There were 20,421 adult patients who visited the ED of a university teaching hospital in Wales between April and August 2016 and who had a serum creatinine measurement. Of those, 1119 cases were included in a retrospective analysis (548 incident ED-AKI patients and 571 randomly assigned non-ED-AKI patients). A pragmatic model for predicting and risk-stratifying AKI was derived using univariate and stepwise backwards removal of insignificant variables in multivariable analyses. The primary outcome of interest was ED-AKI.

The prediction model includes 18 points using four variables for assessing patients on presentation to the ED. Zero indicates a low chance of ED-AKI and 18 indicates a high chance. Adjusted odds ratios for AKI were age, 1.03; male sex, 1.45; known chronic kidney disease, 2.08; and number of comorbidities (from 0 to 10), 1.31.

At a score of >2, the sensitivity was 99.8% (95% confidence interval [CI], 99.0-1.00); specificity was 5.1% (95% CI, 3.4-7.2). At a score of >5.5, sensitivity was 85.8% (95% CI, 82.6-88.6); specificity was 52.7% (95% CI, 48.5-56.9). At a score of >11, sensitivity was 15.0% (95% CI, 12.1-18.2)and specificity was 96.5% (95% CI, 94.6-97.8). There was an increase in the probability of AKI with score groups. The area under the receiver operating characteristic curve  was 0.745 (95% CI, 0.720-0.772; P<.0001; R2 18%). There were positive correlations between the score and peak creatinine (r=0.415; P<.001) but not with AKI stage.

In conclusion, the researchers said, “A simple, pragmatic 18-point score for predicting probability of ED-AKI on ED arrival has been derived. This now requires refinement and prospective validation.”

Source: Foxwell DA, Pradham S, Zouwail S, Rainer TH, Phillips AO. Derivation of a prediction model for emergency department AKI. Abstract of a poster (TH-PO004) presented at the American Society of Nephrology Kidney Week 2018, October 25, 2018, San Diego, California.