HDCN Abstract:  ASN Annual Meeting 2020 -- Digital Meeting  

Calvert J, Allen AO, Le SH, et al.

Development and Validation of a Convolutional Neural Network Model for Intensive Care Unit AKI Prediction

ASN Annual Meeting 2020 -- Digital Meeting
J Am Soc Nephrol (Oct) 31:16A 2020

BACKGROUND

Acute kidney injury (AKI) is common among hospitalized patients and has a significant impact on morbidity and mortality. While early prediction of AKI has the potential to reduce adverse patient outcomes, it remains a difficult condition to predict and diagnose. The purpose of this study was to evaluate the ability of a machine learning algorithm to predict for AKI KDIGO Stage 2 or 3 up to 72 hours in advance of onset using convolutional recurrent neural nets (CNN) and patient Electronic Health Record (EHR) data.

METHODS

A CNN prediction system was developed to continuously and automatically monitor for incipient AKI. 7122 patient encounters were retrospectively analyzed from the Medical Information Mart for Intensive Care III (MIMIC-III) database. The CNN machine learning- based AKI prediction model was compared to an established XGBoost AKI prediction model and the Sequential Organ Failure Assessment (SOFA) scoring system. AKI onset was used for the outcome. The model was trained on routinely-collected patient EHR data.

RESULTS

On a hold-out test set, the algorithm attained an Area Under the Receiver Operating Characteristic (AUROC) of 0.85 and PPV of 0.25, relative to a cohort AKI prevalence of 5.21%, for long-horizon AKI prediction at a 72-hour window prior to onset. The ROC curve comparison of 72-hour prediction on the 10% hold-out test set is shown in Figure 1. The CNN model, which was provided text data through Doc2Vec input, outperformed the XGBoost model and the SOFA score.

CONCLUSION

A CNN machine learning- based AKI prediction model outperforms XGBoost and the SOFA scoring system, demonstrating superior performance in predicting acute kidney injury 72 hours prior to onset, without reliance on changes in serum creatinine.


ROC curve comparison of prediction performance using a CNN classifier, an XGB classifier, and the SOFA score, 72 hours prior to AKI onset on the MIMIC III ICU hold out data set.

c Copyright 2020 -2021 American Society of Nephrology. Reproduced with permission.
All ASN abstracts from the 2020 Annual Meeting are available at this link and also are archived in .pdf form at ASN-Online.org

Disclaimer: Abstracts often have errors, both typographical and otherwise. This posting is an electronic translation of submitted abstracts which has not been verified against the original submitted abstract nor with the authors for accuracy. As a result, there may be errors, especially with regard to drug doses, but not limited to these. Abstracts undergo only limited review, and data often are changed as a result of the peer review process, so their reliability is less than manuscripts published in peer-reviewed journals. In using these summaries, you are agreeing that you are aware of these limitations.

The materials are provided on an as-is basis without any warranty of any kind, either express or implied. In addition to errors, the information presented may be incomplete or outdated. The information contained is not intended nor recommended as a substitute for professional medical advice. You are advised to check the appropriate medical literature and the product information currently provided by the manufacturer of each device to be used or drug to be administered to verify the dosage, the method and duration of administration, or contraindications. It is the responsibility of the treating physician or other health care professional, relying on independent experience and knowledge of the patient, to determine drug, disease, and the best treatment for the patient.

To the fullest extent permitted by law, HDCN, ASN and their affiliates and suppliers disclaim all warranties, express or implied, including, but not limited to, any warranty of merchantability, non- infringement or fitness for a particular purpose.

In no event shall HDCN, ASN, or their affiliates or suppliers be liable for any damages whatsoever (including, but not limited to, direct, indirect, incidental, consequential, punitive or exemplary damages, or any damages for loss of profits, use, data, goodwill or other intangibles) arising from or in any way relating to these terms, the materials, or any information, goods or services obtained from or referred to in the materials, whether based on warranty, contract, tort (including, but not limited to, negligence), or any other legal theory, and whether or not any or all of the limited entities is advised of the possibility of such damages.