A deep-learning model to continuously predict severe acute kidney injury based on urine output changes in critically ill patients

Francesca Alfieri,Riccardo Maria Fagugli,Dario Crosetto,Eros Pasero,Luigi Vecchi,Giovanni Tripepi,Vincenzo Randazzo,Andrea Ancona,Valentina Cauda,Annunziata Paviglianiti

doi:10.1007/s40620-021-01046-6

Abstract

BackgroundAcute Kidney Injury (AKI), a frequent complication of pateints in the Intensive Care Unit (ICU), is associated with a high mortality rate. Early prediction of AKI is essential in order to trigger the use of preventive care actions.MethodsThe aim of this study was to ascertain the accuracy of two mathematical analysis models in obtaining a predictive score for AKI development. A deep learning model based on a urine output trends was compared with a logistic regression analysis for AKI prediction in stages 2 and 3 (defined as the simultaneous increase of serum creatinine and decrease of urine output, according to the Acute Kidney Injury Network (AKIN) guidelines). Two retrospective datasets including 35,573 ICU patients were analyzed. Urine output data were used to train and test the logistic regression and the deep learning model.ResultsThe deep learning model defined an area under the curve (AUC) of 0.89 (± 0.01), sensitivity = 0.8 and specificity = 0.84, which was higher than the logistic regression analysis. The deep learning model was able to predict 88% of AKI cases more than 12 h before their onset: for every 6 patients identified as being at risk of AKI by the deep learning model, 5 experienced the event. On the contrary, for every 12 patients not considered to be at risk by the model, 2 developed AKI.ConclusionIn conclusion, by using urine output trends, deep learning analysis was able to predict AKI episodes more than 12 h in advance, and with a higher accuracy than the classical urine output thresholds. We suggest that this algorithm could be integrated in the ICU setting to better manage, and potentially prevent, AKI episodes.Graphic abstract

Highlights

Acute kidney injury (AKI) is a global public health concern due to increasing patient complexities and aging populations
The present study focuses on the use of deep structural learning, a part of machine learning based on artificial neural networks (ANN), which is believed to improve the care of patients and the individual health outcome [8]
We retrospectively analyzed two large databases of critically ill patients, the electronic Intensive Care Unit [9] and the Medical Information Mart for Intensive Care (MIMIC-III) [10], and we investigated the accuracy of logistic regression and of a deep learning model to predict the risk of Acute Kidney Injury (AKI) development with mathematical models based on urine output

Summary

Introduction

Acute kidney injury (AKI) is a global public health concern due to increasing patient complexities and aging populations. The “Recognition” is a target of the care process, with both “Diagnosis” and “Staging” based on urine output (UO), serum creatinine (sCr), and new biomarkers Both the Acute Kidney Injury Network (AKIN) and the Kidney Disease: Improving Global Outcome (KDIGO) guidelines define oliguria as a reduction of urine output to < 0.5 ml/kg/hour [5]. A deep learning model based on a urine output trends was compared with a logistic regression analysis for AKI prediction in stages 2 and 3 (defined as the simultaneous increase of serum creatinine and decrease of urine output, according to the Acute Kidney Injury Network (AKIN) guidelines). Conclusion In conclusion, by using urine output trends, deep learning analysis was able to predict AKI episodes more than 12 h in advance, and with a higher accuracy than the classical urine output thresholds We suggest that this algorithm could be integrated in the ICU setting to better manage, and potentially prevent, AKI episodes

Objectives

Methods

Results

Conclusion