Developing Machine Learning Models for Predicting Intensive Care Unit Resource Use During the COVID-19 Pandemic: Experiences from a Bi-Regional Health Care System Covering 2.5 Million Citizens in Denmark

Abstract

Introduction: Worldwide, the COVID-19 pandemic has put significant strains on hospital resources. Several studies have indicated that Machine Learning (ML) models can accurately predict adverse events (hospital and intensive care unit (ICU) admission, use of mechanical ventilation and death) on the individual patient level (micro prediction). We hypothesize that these micro predictions can be expanded to population wide (macro) forecasts for ICU resource planning.Methods: Electronic Health Record data from 34,012 SARS-CoV-2 positive patients from two Danish health care regions were extracted. Random Forest (RF) models were trained, using demographics, diagnoses code, laboratory tests and vital signs where available, to predict risk of ICU admission and use of mechanical ventilation after n days (n = 5, 10). These models were then used for a forecasting prognosis of the required hospital resources.Results: When using regular retraining on new data, RF models predicted 5-day risk of ICU admission with an area under the receiver operator characteristic curve (ROC-AUC) of 0∙986 and 5-day risk of use of ventilation with an ROC-AUC of 0∙995. The corresponding 5-day forecasting models predicted the needed ICU capacity with a coefficient of determination (R2) of 0∙930 and use of ventilation with an R2 of 0∙934. Performance was comparable but slightly reduced for 10-day forecasting models.Conclusions: Random Forest-based modelling of COVID-19 electronic health record data can be used for accurate 5- and 10-day forecasting predictions of ICU resource requirements.Funding Statement: The study was funded by grants from the Novo Nordisk Foundation to MS (#NNF20SA0062879 and #NNF19OC0055183) and MN (#NNF20SA0062879).Declaration of Interests: We declare no competing interests.Ethics Approval Statement: The study was approved by the relevant legal boards: the Danish Patient Safety Authority (Styrelsen for Patientsikkerhed, approval #31-1521-257) and the Danish Data Protection Agency (Datatilsynet, approval #P-2020-320). Under Danish law, these agencies provide the required legal approval for the handling of sensitive patient data, including EHR data, without patient consent.