An optimized classification system of acute kidney injury for predicting the short-term mortality after open heart surgery; comparison of current classification systems.

Abstract

Epidemiologic studies need a unique operational definition of acute kidney injury (AKI) to compare outcomes. We aimed to compare prognostic value of change in serum creatinine with classification systems of AKI to predict 30day mortality after heart surgery. From VA database, 27410 eligible patients with stable baseline kidney function who had heart surgery from 1999 to 2005 were selected. There was a graded increase in mortality from stage A to stage C of all systems. Adjusted 30day mortality odds ratio starts to increase significantly after an acute rise ≥ 0.3 mg/dL of creatinine in CKD stages 1 and 2, and after 0.6 mg/dL increase in CKD stage 3. Area under ROC curve of change of creatinine from baseline was significantly higher than those of classification systems (P<0.001). In conclusion, compared to continuous increase of creatinine, classification systems of AKI misestimate mortality risk by collapsing predictive, clinically important data into categories.