Abstract: SA-OR02
Machine Learning for Predication of Severe AKI in Hospitalized Patients with COVID-19
Session Information
- Coronavirus: Research Abstracts
October 24, 2020 | Location: Simulive
Abstract Time: 05:00 PM - 07:00 PM
Category: Coronavirus (COVID-19)
- 000 Coronavirus (COVID-19)
Authors
- Chaudhary, Kumardeep, Icahn School of Medicine at Mount Sinai, New York, New York, United States
- Chan, Lili, Icahn School of Medicine at Mount Sinai, New York, New York, United States
- Saha, Aparna, Icahn School of Medicine at Mount Sinai, New York, New York, United States
- Chauhan, Kinsuk, Icahn School of Medicine at Mount Sinai, New York, New York, United States
- Vaid, Akhil, Icahn School of Medicine at Mount Sinai, New York, New York, United States
- Murphy, Barbara T., Icahn School of Medicine at Mount Sinai, New York, New York, United States
- He, John Cijiang, Icahn School of Medicine at Mount Sinai, New York, New York, United States
- Coca, Steven G., Icahn School of Medicine at Mount Sinai, New York, New York, United States
- Nadkarni, Girish N., Icahn School of Medicine at Mount Sinai, New York, New York, United States
Background
Preliminary reports indicate that acute kidney injury (AKI) is common in coronavirus disease (COVID-19) patients and is associated with worse outcomes. Identification of patients at high risk for developing severe AKI in hospitalized COVID-19 patients in the United States is not well-described.
Methods
This is a retrospective observational study of patients aged ≥18 years with laboratory confirmed COVID-19 admitted to the Mount Sinai Health System between February 27 and April 15, 2020. We trained and tested a machine learning algorithm, extreme gradient boosting (XGBoost), a boosted decision-tree based machine learning (ML) model, with 5-fold cross validation to predict AKI requiring dialysis. Patients from the Mount Sinai (MSH) were randomly split into a training and validation set for the model. To increase model generalizability and help minimize bias, the model’s performance was assessed on a test set composed entirely of patients from the other hospitals in the Mount Sinai Health System (MSHS). Input features for the model included demographics, laboratory values, and vital signs that occurred in the first 48 hours of admission.
Results
Of 3,235 hospitalized patients with COVID-19, AKI occurred in 1406 (46%) patients and 280 (20%) with AKI required dialysis. In the training set (n=1,317 patients), the classifier achieved good performance with an area under the receiver operating characteristic curve (AUROC) of 0.79 and area under the precision recall curve (AUPRC) of 0.38 for predicting AKI requiring dialysis. Performance was similar in the testing set (n=1,918) with 0.79 AUROC and 0.36 AUPRC. The features that had a larger impact on the model included serum creatinine, age, potassium, and heart rate.
Conclusion
A machine-learned model using admission features had good performance for dialysis prediction and could be used for resource allocation.