This prospective observational study was carried out at the Clinical Hospital of the UC-CHRISTUS Health Network, in Santiago, Chile. Patients with laboratory confirmed SARS-CoV-2 infection and moderate to severe ARDS (19), were consecutively included between March 17 and July 31, 2020. Admission pathways comprised the emergency department, and basic ward.
The Institutional Ethics Committee approved this project (Research Ethics Committee Nº 200504004, Faculty of Medicine, Pontificia Universidad Católica de Chile), and waived the need for informed consent.
Laboratory confirmation of SARS-CoV-2 was defined as a positive real-time reverse transcriptase polymerase chain reaction (RT-PCR) result of nasal and pharyngeal swabs.
Our university hospital has a 32-bed both medical-and-surgical ICU with extracorporeal membrane oxygenation (ECMO) capability. Due to the COVID-19 pandemic, the hospital ICU capacity was surged incorporating up to 56 beds from other reconverted units, as needed.
Intensivists and ICU-trained nurses were deployed to these expanded ICUs to ensure a similar standard of care.
All patients undergo an initial respiratory failure management protocol that includes HFNC and the awake prone positioning if tolerated. Orotracheal intubation and connection to MV was performed if the patient had an increased WOB (tachypnea > 30 / min and the use of accessory muscles, paradoxical breathing, altered consciousness, or an hyperadrenergic state), refractory hypoxemia (O2 saturation < 90% despite prone position and HFNC at maximum fraction of inspired oxygen (FiO2) or the presence of concomitant shock (Additional file 1). The decision to intubate was by attending physicians, and MV started in the volume-control ventilation mode according to local management protocol (Additional file 1).
Data Collection
Data were recorded prospectively by the research team in an electronic worksheet. The study data were recorded prospectively by the research team during the patient’s stay in the ICU and collected and managed using the REDCap electronic data capture tools hosted at Pontificia Universidad Católica de Chile.
Clinical data included sex, age, weight, height, medical comorbidities, days since the start of symptoms, laboratory parameters, and PaO2/FiO2 ratio by the time of hospital admission. The Acute Physiology and Chronic Health Evaluation (APACHE) (20) and Sequential Organ Failure Assessment (SOFA) scores(21) were calculated within 24 hours of ICU admission. Subsequently, clinical, laboratory and ventilatory parameters were recorded both on day 1 from the start of invasive MV and included: respiratory support mode, positive end-expiratory pressure (PEEP) level, arterial blood gases and FIO2, and respiratory system compliance (Crs).
Outcomes
Time of intubation was defined as the time from hospital admission to endotracheal intubation, and classified as lower than 72 hours (early) and equal or higher than 72 hours (late), according to published data (17) and authors consensus. The primary outcome was in-hospital death after receiving MV. Secondary outcomes included duration of MV, ICU and hospital length of stay, and mortality on day 28 and on discharge from the ICU.
Statistical Analysis
For variables with non-normal distribution, non-parametric tests were used. Accordingly, descriptive statistics are shown as medians [interquartile range 25–75] or percentages (%). Mann-Whitney U, Kruskal Wallis, chi-square, and Fisher’s exact, were used when appropriate. We performed univariate analyses between main outcome and clinical and laboratory variables. Those with a univariate p-value of 0.1 were included in the multivariate analysis plus other clinically relevant ones. Logistic models were fitted testing individual and interaction variables. Multivariable fractional polynomial methods were also used in order to preserve the continuous nature of important covariates as PaO2/FiO2 ratio, time to intubation and age, since we did suspect that their relationships with other variables were non-linear.
As mentioned, for hypoxemia and intubation times we used published cutoffs (17).
We did not use APACHE II for fitting multivariate models to avoid overadjustment, since age was incorporated as an individual variable or in the CALL score, PaO2/FiO2 ratio was a relevant individual variation, and comorbidities were represented in the regression models, as well.
Data was analyzed using Graphpad Prism 7 (Graphpad Softwares, La Joya, CA), and Stata 16 (StataCorp, College Station, TX. USA) statistical packages. Two-tailed p value of < 0.05 was considered as statistically significant.