+1 Recommend
0 collections
      • Record: found
      • Abstract: found
      • Article: found
      Is Open Access

      Machine learning for the prediction of volume responsiveness in patients with oliguric acute kidney injury in critical care

      Read this article at

          There is no author summary for this article yet. Authors can add summaries to their articles on ScienceOpen to make them more accessible to a non-specialist audience.


          Background and objectives

          Excess fluid balance in acute kidney injury (AKI) may be harmful, and conversely, some patients may respond to fluid challenges. This study aimed to develop a prediction model that can be used to differentiate between volume-responsive (VR) and volume-unresponsive (VU) AKI.


          AKI patients with urine output < 0.5 ml/kg/h for the first 6 h after ICU admission and fluid intake > 5 l in the following 6 h in the US-based critical care database (Medical Information Mart for Intensive Care (MIMIC-III)) were considered. Patients who received diuretics and renal replacement on day 1 were excluded. Two predictive models, using either machine learning extreme gradient boosting (XGBoost) or logistic regression, were developed to predict urine output > 0.65 ml/kg/h during 18 h succeeding the initial 6 h for assessing oliguria. Established models were assessed by using out-of-sample validation. The whole sample was split into training and testing samples by the ratio of 3:1.

          Main results

          Of the 6682 patients included in the analysis, 2456 (36.8%) patients were volume responsive with an increase in urine output after receiving > 5 l fluid. Urinary creatinine, blood urea nitrogen (BUN), age, and albumin were the important predictors of VR. The machine learning XGBoost model outperformed the traditional logistic regression model in differentiating between the VR and VU groups (AU-ROC, 0.860; 95% CI, 0.842 to 0.878 vs. 0.728; 95% CI 0.703 to 0.753, respectively).


          The XGBoost model was able to differentiate between patients who would and would not respond to fluid intake in urine output better than a traditional logistic regression model. This result suggests that machine learning techniques have the potential to improve the development and validation of predictive modeling in critical care research.

          Related collections

          Most cited references 20

          • Record: found
          • Abstract: found
          • Article: not found

          Acute kidney injury in the intensive care unit according to RIFLE.

          To apply the RIFLE criteria "risk," "injury," and "failure" for severity of acute kidney injury to patients admitted to the intensive care unit and to evaluate the significance of other prognostic factors. Retrospective analysis of the Riyadh Intensive Care Program database. Riyadh Intensive Care Unit Program database of 41,972 patients admitted to 22 intensive care units in the United Kingdom and Germany between 1989 and 1999. Acute kidney injury as defined by the RIFLE classification occurred in 15,019 (35.8%) patients; 7,207 (17.2%) patients were at risk, 4,613 (11%) had injury, and 3,199 (7.6%) had failure. It was found that 797 (2.3%) patients had end-stage dialysis-dependent renal failure when admitted to an intensive care unit. None. : Patients with risk, injury, and failure classifications had hospital mortality rates of 20.9%, 45.6%, and 56.8%, respectively, compared with 8.4% among patients without acute kidney injury. Independent risk factors for hospital mortality were age (odds ratio 1.02); Acute Physiology and Chronic Health Evaluation II score on admission to intensive care unit (odds ratio 1.10); presence of preexisting end-stage disease (odds ratio 1.17); mechanical ventilation (odds ratio 1.52); RIFLE categories risk (odds ratio 1.40), injury (odds ratio 1.96), and failure (odds ratio 1.59); maximum number of failed organs (odds ratio 2.13); admission after emergency surgery (odds ratio 3.08); and nonsurgical admission (odds ratio 3.92). Renal replacement therapy for acute kidney injury was not an independent risk factor for hospital mortality. The RIFLE classification was suitable for the definition of acute kidney injury in intensive care units. There was an association between acute kidney injury and hospital outcome, but associated organ failure, nonsurgical admission, and admission after emergency surgery had a greater impact on prognosis than severity of acute kidney injury.
            • Record: found
            • Abstract: found
            • Article: not found

            Fluid challenges in intensive care: the FENICE study

            Background Fluid challenges (FCs) are one of the most commonly used therapies in critically ill patients and represent the cornerstone of hemodynamic management in intensive care units. There are clear benefits and harms from fluid therapy. Limited data on the indication, type, amount and rate of an FC in critically ill patients exist in the literature. The primary aim was to evaluate how physicians conduct FCs in terms of type, volume, and rate of given fluid; the secondary aim was to evaluate variables used to trigger an FC and to compare the proportion of patients receiving further fluid administration based on the response to the FC. Methods This was an observational study conducted in ICUs around the world. Each participating unit entered a maximum of 20 patients with one FC. Results 2213 patients were enrolled and analyzed in the study. The median [interquartile range] amount of fluid given during an FC was 500 ml (500–1000). The median time was 24 min (40–60 min), and the median rate of FC was 1000 [500–1333] ml/h. The main indication for FC was hypotension in 1211 (59 %, CI 57–61 %). In 43 % (CI 41–45 %) of the cases no hemodynamic variable was used. Static markers of preload were used in 785 of 2213 cases (36 %, CI 34–37 %). Dynamic indices of preload responsiveness were used in 483 of 2213 cases (22 %, CI 20–24 %). No safety variable for the FC was used in 72 % (CI 70–74 %) of the cases. There was no statistically significant difference in the proportion of patients who received further fluids after the FC between those with a positive, with an uncertain or with a negatively judged response. Conclusions The current practice and evaluation of FC in critically ill patients are highly variable. Prediction of fluid responsiveness is not used routinely, safety limits are rarely used, and information from previous failed FCs is not always taken into account. Electronic supplementary material The online version of this article (doi:10.1007/s00134-015-3850-x) contains supplementary material, which is available to authorized users.
              • Record: found
              • Abstract: not found
              • Article: not found

              KDIGO clinical practice guideline for acute kidney injury


                Author and article information

                Crit Care
                Critical Care
                BioMed Central (London )
                8 April 2019
                8 April 2019
                : 23
                [1 ]ISNI 0000 0004 1759 700X, GRID grid.13402.34, Department of Emergency Medicine, Sir Run Run Shaw Hospital, , Zhejiang University School of Medicine, ; No. 3, East Qingchun Road, Hangzhou, 310016 Zhejiang Province China
                [2 ]ISNI 0000 0004 1936 7910, GRID grid.1012.2, School of Population and Global Health, , University of Western Australia, ; Perth, Australia
                © The Author(s). 2019

                Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License ( http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver ( http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

                Custom metadata
                © The Author(s) 2019


                Comment on this article