ABSTRACT
RATIONALE Acute hypoxemic respiratory failure (AHRF) is the major complication of coronavirus disease 2019 (COVID-19), yet optimal respiratory support strategies are uncertain. We aimed to describe outcomes with highflow oxygen delivered through nasal cannula (HFNC) and non-invasive positive pressure ventilation (NIPPV) in COVID-19 AHRF and identify individual factors associated with non-invasive respiratory support failure. METHODS We conducted a retrospective cohort study of hospitalized adults with COVID-19 within a large academic health system in New York City early in the pandemic to describe outcomes with HFNC and NIPPV. Patients were categorized into the HFNC cohort if they received HFNC but not NIPPV, whereas the NIPPV cohort included patients who received NIPPV with or without HFNC. We described rates of HFNC and NIPPV success, defined as live discharge without endotracheal intubation (ETI). Further, using Fine-Gray sub-distribution hazard models, we identified demographic and patient characteristics associated with HFNC and NIPPV failure, defined as the need for ETI and/or in-hospital mortality. RESULTS Of the 331 patients in the HFNC cohort, 154 (46.5%) patients were successfully discharged without requiring ETI. Of the 177 (53.5%) who experienced HFNC failure, 100 (56.5%) required ETI and 135 (76.3%) patients ultimately died. Among the 747 patients in the NIPPV cohort, 167 (22.4%) patients were successfully discharged without requiring ETI, and 8 (1.1%) were censored. Of the 572 (76.6%) patients who failed NIPPV, 338 (59.1%) required ETI and 497 (86.9%) ultimately died. In adjusted models, significantly increased risk of HFNC and NIPPV failure was observed among patients with co-morbid cardiovascular disease (sub-distribution hazard ratio (sHR) 1.82;95% confidence interval (CI), 1.17-2.83 and sHR 1.40;95% CI 1.06-1.84, respectively). Conversely, a higher oxygen saturation to fraction of inspired oxygen ratio (SpO2/FiO2) at HFNC and NIPPV initiation was associated with reduced risk of failure (sHR, 0.32;95% CI 0.19-0.54, and sHR 0.34;95% CI 0.21-0.55, respectively). CONCLUSIONS A subset of patients with COVID-19 AHRF was effectively managed with non-invasive respiratory modalities and achieved successful hospital discharge without requiring ETI. Notably, patients with co-morbid cardiovascular disease and more severe hypoxemia experienced lower success rates with both HFNC and NIPPV. Identification of specific patient factors may help inform more selective use of non-invasive respiratory strategies, and allow for a more personalized approach to the management of COVID-19 AHRF in pandemic settings.
ABSTRACT
Rationale: High patient volume and limited ICU resources associated with the COVID-19 pandemic have exacerbated ICU capacity strain, leading to longer pre-ICU lengths-of-stay (LOS). We examined the patient- and hospital-level predictors of pre-ICU LOS, and the association of pre-ICU LOS on in-hospital mortality for patients with COVID-19. Methods: Data were derived from the Study of the Treatment and Outcomes in Critically Ill Patients with COVID-19 (STOP-COVID), a multicenter cohort study of critically ill adults with COVID-19 admitted to 68 US hospitals. All patients had a minimum of 28-day follow-up;those discharged from hospital were presumed alive. The primary outcome was pre-ICU LOS, dichotomized into brief (≤1 day) vs. prolonged (>1 day). We constructed a multivariate mixed effects model, adjusting for patient factors (e.g., demographics, comorbidities, and pre-hospital symptom duration) and hospital factors (pre-COVID ICU beds number, countylevel case rates of COVID-19 (number of cases per 100,000 residents), and the hospital site itself) to determine predictors of pre-ICU LOS. Using 1:3 propensity score matching for pre-ICU period, we used multivariate mixed effect modelling to examine the association between pre-ICU LOS and in-hospital mortality. Results: A total of 4738 patients with complete data were admitted to the ICU, 36.6% were female, with median age 62 years (IQR 52-71). The majority (85.5%) were admitted from the ED or wards, with 62.5% classified as having a brief pre- ICU LOS. While demographics and co-morbidities (cancer, diabetes, and end-stage renal disease) were not associated with pre-ICU LOS, pre-existing lung disease (OR 1.33, 95% CI 1.02-1.74) was a patient-level predictor of a brief pre-ICU LOS as compared to a prolonged LOS. Having more available ICU beds (>100 vs. 0-48 ICU beds, OR 1.41, 95% CI 1.03-1.92) was a hospital-level predictor of a brief pre-ICU LOS. More patients were intubated at the time of ICU arrival in the prolonged pre-ICU LOS group, compared to the brief LOS group (64.6% vs. 59.2%, p≤0.001). In the mixed model, propensity matched for pre-ICU LOS, and adjusted for patient/hospital characteristics, differential pre-ICU LOS was not predictive of in-hospital mortality (OR 1.22, 95%CI 0.81-1.87), though oxygen support modality was associated with mortality. Conclusion: Patient- and hospital-level factors, such as ICU capacity, had an impact on pre-ICU duration, with more patients requiring a higher level of oxygen support at ICU arrival if admitted later in their course. However, once adjusting for clinical and hospital factors, pre-ICU LOS was not associated with in-hospital mortality.