Validation of the AKIN criteria definition using high-resolution ICU data from the MIMIC-II database
- 2.2k Downloads
Recently the Acute Kidney Injury (AKI) Network proposed criteria for the definition of AKI in the critically ill. The minimum hourly urine output rate used to define oliguria (< 0.5 ml/kg/hour) is based exclusively on clinical experience and animal models, not on clinical investigation. Moreover, the minimum duration of oliguria (6 hours) is based on clinical experience and was never experimentally determined. We used a massive database of ICU patients (MIMIC) to continuously vary the observation period and threshold of urine output measurements to determine optimal AKI definitions for improved in-hospital mortality prediction.
After excluding end-stage renal disease, 14,536 adult patients were included. Various AKI thresholds corresponding to different observation periods and urine output measurement thresholds were analyzed using a multivariate logistic regression model for each choice of thresholds. A total of 470 regression models were plotted. We controlled for sex, age, SOFA and co-morbidities (ICD-9 codes). To visualize dependence of adjusted mortality rate and mortality predictive power on AKI definition, we generated 3 D and contour plots.
The current AKIN recommendation that uses a urine output of 0.5 ml/kg/hour is valid. Since AKIN's stages 1 and 2 were found to exhibit similar mortality rates, we propose a reduction in the AKI 2 threshold to 0.4 ml/kg/hour to better delineate among the three stages. We demonstrated that the mortality rate increases sharply during the first 5 hours of oliguria. Hence, the current used observation period (6 hours) seems to be valid.
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.