Singh K, et al. medRxiv 2020.
ABSTRACT
INTRODUCTION: The coronavirus disease 2019 (COVID-19) pandemic is straining the capacity of U.S. healthcare systems. Accurately identifying subgroups of hospitalized COVID-19 patients at high- and low-risk for complications would assist in directing resources.
OBJECTIVE: To validate the Epic Deterioration Index (EDI), a predictive model implemented in over 100 U.S. hospitals that has been recently promoted for use in COVID-19 patients.
METHODS: We studied adult patients admitted with COVID-19 to non-ICU level care at a large academic medical center from March 9 through April 7, 2020. We used the EDI, calculated at 15-minute intervals, to predict a composite adverse outcome of ICU-level care, mechanical ventilation, or death during the hospitalization. In a subset of patients hospitalized for at least 48 hours, we also evaluated the ability of the EDI (range 0-100) to identify patients at low risk of experiencing this composite outcome during their remaining hospitalization. We evaluated model discrimination and calibration using both raw EDI scores and their slopes.
RESULTS: Among 174 COVID-19 patients meeting inclusion criteria, 61 (35%) experienced the composite outcome. Area under the receiver-operating-characteristic curve (AUC) of the EDI was 0.76 (95% CI 0.68-0.84). Patients who met or exceeded an EDI of 64.8 made up 17% of the study cohort and had an 80% probability of experiencing the outcome during their hospitalization with a median lead time of 28 hours from when the threshold was first exceeded to the outcome. Employing the EDI slope lowered the AUCs to 0.68 (95% CI 0.60-0.77) and 0.67 (95% CI 0.59-0.75) for slopes calculated over 4 and 8 hours, respectively. In a subset of 109 patients hospitalized for at least 48 hours and who had not experienced the composite outcome, 14 (13%) patients who never exceeded an EDI of 37.9 had a 93% probability of not experiencing the outcome throughout the rest of their hospitalization, suggesting low risk.
CONCLUSION: In this validation study, we found the EDI identifies small subsets of high- and low-risk patients with fair discrimination. These findings highlight the need for hospitals to carefully evaluate prediction models before widespread operational use among COVID-19 patients.
PMID:32511650 | PMC:PMC7277006 | DOI:10.1101/2020.04.24.20079012