Associations Among Patient Race, Sedation Practices, and Mortality in a Large Multi-Center Registry of COVID-19 Patients

Abstract

Introduction: Racial and ethnic minorities have accounted for the majority of intensive care unit (ICU) hospitalizations for COVID-19. At the same time, ICUs were forced to deviate from long-established care processes in response to a steep increase in admissions and to prevent healthcare worker infections. These shifts may have resulted in changes to sedation practices, such as level of sedation or sedation holidays, that differed by patient race or ethnicity. We aimed to examine associations among patient race and ethnicity, sedation practices, and mortality in a large, national sample of patients receiving mechanical ventilation for COVID-19. Methods: We analyzed granular daily data from the Viral Infection and Respiratory Illness Universal Study (VIRUS) Registry for COVID-19 patients admitted to ICUs between February and November 2020. We included patients over 18 years of age, who were mechanically ventilated following clinical or PCR-confirmed COVID-19 diagnosis. We will calculate descriptive statistics for mortality at discharge and 28 days by patient race/ethnicity, sex, and two care processes associated with mechanical ventilation: sedation level and sedation holidays. We will estimate risk-adjusted, hospital-level mortality differentials by race. We will use mixed effects logistic regression and causal mediation analysis to test associations among patient race/ethnicity, sedation practices for mechanical ventilation, and mortality at 28 days, controlling for comorbidities, markers of severity, and time to admission, and adjusting for clustering by ICU. Results: Among 19,626 patients hospitalized for COVID-19, 8,668 (14.6%) received mechanical ventilation at 238 hospitals. The median age was 62 (IQR 40-72) and 45.1% were female. Among hospitalized patients, 23.3% self-identified as Hispanic, 26.6% as non-Hispanic Black, 35.6% as non-Hispanic White, and 14.5% as non-Hispanic and another racial group. Approximately 1% (n=236) of patients were missing race/ethnicity. At 28 days, 20.7% (n=4,076) of hospitalized patients were deceased. Use of benzodiazepines was highly clustered by hospital (intraclass correlation coefficient of 0.63). In cluster-adjusted analyses, Hispanic patients were more likely to receive benzodiazepines at least once during hospitalization than either non-Hispanic White (Odds Ratio (OR) 0.76, p=0.013) or non-Hispanic Black (OR 0.70, p=0.003) patients. Multivariable mixed effects and causal mediation analyses are ongoing. Conclusions: Sedation practices, such as level of sedation and sedation holidays, are associated with mortality;yet these practices may differ based on a patient's race or ethnicity. We will leverage a unique, multi-center database with granular clinical information to understand how these differences may influence racial and ethnic disparities in respiratory failure.