Remodeling the model for end‐stage liver disease for predicting mortality risk in critically ill patients with cirrhosis and acute kidney injury

Abstract

Serum creatinine measurement demonstrates a poor specificity and sensitivity for the early diagnosis of acute kidney injury (AKI) in patients with cirrhosis. The existing model for end‐stage liver disease (MELD) score reveals multiple pitfalls in critically ill patients with cirrhosis and acute kidney injury (CAKI). The aim of this study was to re‐evaluate the role of creatinine values in the existing MELD score and to develop a novel score for CAKI, named the “acute kidney injury–model for end‐stage liver disease score” (AKI‐MELD score). We extracted 651 CAKI from the Multiparameter Intelligent Monitoring in Intensive Care database. A time‐dependent Cox regression analysis was performed for developing remodeled MELD scores (Reweight‐MELD score, Del‐Cr‐MELD score, and AKI‐MELD score). The area under the receiver operating characteristic curve provided the discriminative power of scoring models related to outcome. The hazard ratio of creatinine was 1.104 (95% confidence interval [CI], 0.945‐1.290; P = 0.211). Reweight‐MELD score and Del‐Cr‐MELD score (decreasing the weight of creatinine) were superior to the original MELD score (all P < 0.001). The new AKI‐MELD score consists of bilirubin, the international normalized ratio, and the ratio of creatinine in 48 hours to creatinine at admission. It had competitive discriminative ability for predicting mortality (area under the receiver operating characteristic curve, 0.720 [95% CI, 0.653‐0.762] at 30 days, 0.688 [95% CI, 0.630‐0.742] at 90 days, and 0.671 [95% CI, 0.612‐0.725] at 1 year). Further, AKI‐MELD score had significantly higher predictive ability in comparison with MELD score, MELD‐Na score, and Updated MELD score (all P < 0.001). Conclusion: The predictive value of creatinine for CAKI should be re‐evaluated. AKI‐MELD score is a potentially reliable tool to determine the prognosis for mortality of CAKI. (Hepatology Communications 2017;1:748–756)

Application of interpretable machine learning for early prediction of prognosis in acute kidney injury

“…Early identification of AKI patients at high risk for clinical deterioration is of great importance and may help to deliver proper care and optimize the use of limited resources. Considering the potential benefits of electronic alerts in AKI, many researchers have developed a variety of machine learning-based models to predict mortality for AKI patients [5] , [6] , [7] , [8] . However, these previously established models were limited to clinical implementation due to the minimal interpretability and black box nature of the algorithms [9] .…”