Predictive Approaches for Acute Dialysis Requirement and Death in COVID-19.

Vaid, Akhil; Chan, Lili; Chaudhary, Kumardeep; Jaladanki, Suraj; Paranjpe, Ishan; Russak, Adam; Kia, Arash; Timsina, Prem; Levin, Matthew; He, John; Bottinger, Erwin; Charney, Alexander; Fayad, Zahi; Coca, Steven; Glicksberg, Benjamin; Nadkarni, Girish

BACKGROUND Acute Kidney Injury treated with dialysis initiation is a common complication of COVID-19 infection among hospitalized patients. However, dialysis supplies and personnel are often limited. METHODS Using data from adult hospitalized COVID-19 patients from five hospitals from the Mount Sinai Health System who were admitted from March 10th and December 26th, 2020, we developed and validated several models (logistic regression, LASSO, random forest, and XGBoost (with and without imputation)) for predicting treatment with dialysis or death at various time horizons (1, 3, 5 and 7 days) following hospital admission. Patients admitted to the Mount Sinai Hospital were used for internal validation while the other hospitals were part of the external validation cohort. Features included demographics, comorbidities, and laboratory and vitals signs within 12 hour of hospital admission. RESULTS 6093 patients (2,442 in training and 3,651 in external validation) were included in the final cohort. Of the different model approaches used, XGBoost without imputation had the highest area under the receiver curve on internal validation (range of 0.93-0.98) and area under the precision recall curve (range of 0.78-0.82 across) for all time points. XGBoost without imputation also had the highest test parameters on external validation (AUROC range 0.85 to 0.87, and AUPRC range of 0.27 to 0.54) across all time windows. XGBoost without imputation outperformed all models with higher precision and recall (mean difference in AUROC of 0.04 and mean difference in AUPRC of 0.15). Features of creatinine, Blood Urea Nitrogen, and Red cell distribution width were major drivers of model's prediction. CONCLUSIONS An XGBoost model without imputation for prediction of a composite outcome of either death or dialysis in COVID positive patients had the best performance compared to standard and other machine learning models.