Variations in Presentation and Management of Critically Ill Coronavirus Disease 2019 Patients: A Multi-Center Descriptive Analysis

Rationale: Early in the coronavirus disease 2019 (COVID-19) pandemic there was significant practice variation among hospitals regarding the choice and timing of treatments for acute respiratory failure. It is unknown whether this practice variation contributed to outcome differences. Methods: We performed a retrospective study of all adult patients with respiratory failure due to COVID-19 admitted between March 11 and May 31, 2020 to a medical or surgical ICU at three Massachusetts hospitals. Medical charts were manually reviewed by physicians and abstracted into a standardized REDCap database. Chi-square test for categorical variables and Kruskal-Wallis test for continuous variables were performed using R version 4.0.2. Results: Data from 429 patients were analyzed. Among the three institutions, there were significant differences in race, prevalence of hypertension and diabetes mellitus, duration of COVID-19 symptoms on presentation, and days between admission and intubation. Significant differences were observed in presentation acuity by sequential organ failure assessment (SOFA) score but not simplified acute physiology score (SAPS) or PaO2:FiO2 ratios. Hospital A intubated more patients on the day of admission and utilized more inhaled nitric oxide and less immunosuppression (steroids, anti-IL6 agents). Hospital B treated more patients with remdesivir, other experimental antivirals, and early paralysis (within 48 hours of intubation) but less awake prone positioning. Hospital C utilized more non-invasive positive pressure ventilation (NIPPV) and high flow oxygen in lieu of intubation;it also administered more statins and steroids for acute respiratory distress syndrome (ARDS) and used less early proning within 48 hours of intubation. No difference in hydroxychloroquine use was seen across institutions. There were no statistical differences across hospitals in reintubation, ventilator-free days at 28 days, or in-hospital mortality. Transition to comfort measures was more common at hospital C. There was a trend at hospital A toward lower 30-day (A=25.3%, B=32.1%, C=39.4%;p=0.054) and 90-day (A=28.5%, B=36.1%, C=41.4%;p=0.085) mortality. At hospital A there was significantly longer hospital length-of-stay (A=25.0, B=19.0, C=15.0;p=0.004) and ICU length-of-stay (A=18.0, B=15.0, C=12.0;p=0.001). Conclusions: Early in the COVID-19 pandemic in Massachusetts, there were significant differences in patient characteristics and treatments administered across three institutions. One institution demonstrated a trend toward lower 30-day and 90-day mortality despite later presentation from symptom onset, higher admission acuity, and less utilization of remdesivir or steroids. Practice variation across institutions may explain differences in outcomes, independent of baseline characteristics, and should be studied further as it may inform future management of COVID-19.