ABSTRACT
INTRODUCTION: To describe the epidemiology of Coronavirus Disease 2019 (COVID-19)-related critical illness at a diverse academic health system. METHODS: We performed a single-health system, multihospital retrospective cohort study of patients with COVID- 19-related critical illness who were admitted to an intensive care unit (ICU) at any of five hospitals within the University of Pennsylvania Health System. We report descriptive statistics for patient demographics, comorbidities, acute physiology parameters, receipt of ICU therapies, hospital outcomes, and survivorship. Using multivariable linear and logistic regression, we evaluated trends over time in all-cause 28-day in-hospital mortality, the primary outcome, and in patient acuity, and we evaluated candidate prognostic risk factors for association with mortality. RESULTS: 468 patients with COVID-19-related critical illness had a median age of 65 years (interquartile range [IQR] 54-74), were more likely male (57.7%), were more likely Black race (52.8%), and had a high co-morbidity burden (71.8% with ≥ 2 points on the Charlson Comorbidity Index). At least once during their hospitalization, 319 (68.2%) patients were treated with mechanical ventilation and 121 (25.9%) with vasopressors. Outcomes were notable for 29.9% all-cause 28-day in-hospital mortality (37.0% among those who received mechanical ventilation and 14.8% among those who did not receive mechanical ventilation), 8-day (IQR 3-17) median ICU length of stay, 13-day (IQR 7-25) median hospital length of stay, and 10.8% all-cause 30-day readmission rate. Mortality decreased over time from 43.5% (95% confidence interval 31.3%-53.8%) to 19.2% (11.6%- 26.7%) between the first and last 15-day periods in the fully adjusted model. Risk factors at ICU admission prognostic for mortality included increasing age, peripheral vascular disease, low or high body mass index, abnormal mental status, hypoxemia, tachypnea, and thrombocytopenia. CONCLUSIONS: Among patients with COVID-19-related critical illness admitted to the ICU at an academic health system in the U.S., a finite set of patient-level factors were prognostic for mortality and mortality decreased over time.