Full description of the Boussignac valve and FF-CPAP is provided in the Additional file 1. Briefly, the principle of FF-CPAP is to add a filter, which acts as a “microbiological barrier” between the oro-nasal mask and the CPAP valve. Numerous filters with a viral filtration efficiency above 99.99% are available. We conducted the entire bench assessment with two different filters characterized by different humidification and mechanical properties: the DAR™ Adult–Pediatric Electrostatic Filter HME Small (Hygrobac S; Covidien, Medtronic, Parkway, MN, USA) and the Clear-Guard™ (Intersurgical®, Fontenay Sous Bois, France). The first one (subsequently named “DAR filter”) is a heat and moisture exchanger with high humidification performances ; the second one (subsequently named “Clear-Guard filter”) is an electrostatic filter with poor humidification performances but lower resistance. The resistance of each filter was measured at the following air flow rates: 30, 60, 90 and 120 L/min.
Static measurements of airway pressure
The Boussignac CPAP valve was connected to the airway opening of a test lung model with or without the filter with anti-viral properties. To evaluate the impact of the filter on the effective pressure transmitted to the patient, we used a Michigan test lung (Michigan Instruments, Grand Rapids, USA), with a simulated compliance of 50 mL/cm H2O and two simulated resistances (5 and 15 cm H2O/L/s). The oxygen flow meter (ball flowmeter, 0–30 L/min, Technologie Biomedicale S.A.S, Noisy-Le-Sec, France) was adjusted to set the CPAP level at 6 and 10 cm H2O without the filter. The airway pressure measured inside the test lung was compared to the set CPAP level (measured with the dedicated manometer) without and with the filter placed in between.
Dynamic assessment of FF-CPAP
First, we assessed the end expiratory pressure and tidal volume generated by the FF-CPAP at different oxygen flow rates while simulating various inspiratory efforts. An oro-nasal mask (AcuCare non-vented mask, ResMed) was strapped to the face of a RespiSim® Manikin (IngMar Medical, Pittsburg, PA, USA) and connected to a breathing simulator, Active Servo Lung 5000 (ASL5000®, IngMar Medical, Pittsburg, PA, USA; full methods in Additional file 1). The pressure into the oro-nasal mask was recorded at five constant oxygen flow rates: 10, 15, 20, 25, and 30 L/min while simulating four different inspiratory efforts (simulated inspiratory muscle pressures, Pmus): 5, 10, 15, and 20 cm H2O with the following respiratory mechanics: compliance = 50 mL/m H2O, resistance = 5 cm H2O/L/s.
Second, we assessed the impact of the additional resistive load related to the filter in dynamic conditions. The Boussignac valve was connected to the airway opening of the ASL 5000 lung simulator. The volume, airway pressure and Pmus were recorded without and with the filter in the following eight conditions: at two simulated effort (5 and 10 cm H2O of Pmus), two simulated resistances (5 and 15 cm H2O/L/s) with a constant compliance of 50 mL/cm H2O and two levels of CPAP (6 and 10 cm H2O). The decrease in volume (Delta Vt) induced by the filter and the maximum change in airway pressure between inspiration and expiration [expressed as peak-to-peak airway pressure (P-P)] were measured for each condition.
Dynamic pressure–volume loops were reconstructed based on volume and airway pressure recordings to calculate the work of breathing imposed by the device (WOBimposed, see Additional file 1 for more details). In each condition, the relative change in WOBimposed induced by the filter (ΔWOBimposed) was calculated and expressed as a percentage of the WOBimposed without the filter. Dynamic pressure–volume loops were also reconstructed based on volume and muscle pressure recordings, to calculate the theoretical increase in patient’s work of breathing required to maintain the tidal volume constant (see Additional file 1 for more details). Relative changes in patient’s work of breathing needed to maintain the tidal volume constant was calculated and expressed as a percentage of its value without the filter.
In four patients receiving ventilatory support with the FF-CPAP with the DAR filter at four different oxygen flow rates (15, 20, 25, and 30 L/min), the pressure into the oro-nasal mask was recorded (see Additional file 1). A written informed consent was obtained from each patient and this physiological evaluation was approved by Mondor Institutional Review Board.
Setup of intermediate care units and related training
The hospital admitted the first COVID-19 patient on February 15th, 2020. By March 14th, 2020, 52 patients were hospitalized, of whom 12 in ICU. Two intermediate care units (20 beds) with a 1/6 patient–nurse ratio were then created to treat patients with COVID-19 related acute hypoxemic respiratory failure (COVID-AHRF) who did not require immediate intubation.
Training program was rapidly programmed to enable non-ICU nurses and doctors to use FF-CPAP on COVID-AHRF patients. This training had to be continuous and at distance, hence the choice of a short (5 min) video tutorial (e-Video in the Additional file 2 or: http://www.reamondor.aphp.fr/covid19/). This tutorial was available on every computer in intermediate care units and integrated into a massive open online course (MOOC) dedicated to COVID-19 patients’ care (https://www.fun-mooc.fr/courses/course-v1:UPEC+169003+archiveouvert/about and https://covid19.coorpacademy.com/dashboard). The usefulness of this video tutorial was assessed retrospectively through a survey covering the medical and paramedical staff of the intermediate care units. We asked them to assess several statements (see Additional file 1) using a Likert scale model (strongly disagree/disagree/neutral/agree/strongly agree).
This was a single-center retrospective study conducted in Henri Mondor University Hospital, Créteil, France, and approved by the institutional ethical committee of the French Intensive Care Society as a component of standard care. In accordance with French law, the patient's consent was waived, but each patient or his or her next of kin has been informed and given the opportunity to refuse the use of his or her personal data.
All consecutive patients with a “full code” order who received FF-CPAP as the first line ventilatory support for COVID-AHRF between March 14th and April 14th, 2020, were included. In case of “do not intubate” order upon FF-CPAP initiation, patients were not included. COVID-AHRF was defined as acute dyspnea (with a respiratory rate > 25 breaths/min and/or active contraction of accessory respiratory muscles), with escalating oxygen therapy ≥ 6 L/min with a non-rebreather facemask to maintain SpO2 > 90%, and new pulmonary infiltrates on chest X-rays  in a patient diagnosed with COVID-19. The latter was defined by a positive SARS-CoV-2 PCR on a naso-pharyngeal swab and/or a compatible computed tomography scan (CT-scan).
FF-CPAP was assembled with the DAR filter and the same filter was left in place during the whole duration of FF-CPAP therapy. FF-CPAP support was initiated in patients with COVID-AHRF as defined above. FF-CPAP was not advised in case of hemodynamic instability or impaired neurologic status. The minimal oxygen flow rate with FF-CPAP was 15 L/min. The FF-CPAP was used in all cases in a continuous pattern interrupted to allow the patient to eat or whenever exceeded patient’s tolerance due to discomfort. During such interruptions, oxygen was supplied via the non-rebreather facemask. The presence of previously predefined respiratory criteria for intubation at the time of CPAP initiation was sought (intubation is recommended when at least two of such criteria are present) : respiratory rate of > 40 breaths/min, signs of high respiratory muscle workload (meaning active contraction of accessory respiratory muscles), copious tracheal secretions, acidosis with pH < 7.35, and SpO2 < 90% for more than 5 min. Patients were intubated in case of persistence or emergence of signs necessitating intubation despite FF-CPAP therapy.
The main aim of the study was to assess the feasibility, efficiency and safety of using FF-CPAP to maintain adequate levels of oxygenation and to manage a massive influx of COVID-AHRF patients out of the ICU. Thus, we assessed the following main end points: (1) the effect of FF-CPAP on respiratory symptoms (decrease in respiratory rate) and oxygenation (increase in SpO2); (2) the duration of FF-CPAP therapy; (3) the proportion of patients who were ultimately not intubated, especially among patients exhibiting predefined criteria for intubation upon FF-CPAP initiation; (4) the proportion of patients remaining in intermediate care units without ICU admission; (5) the incidence of severe adverse event defined as hypoxemic cardiac arrest prior to intubation under FF-CPAP therapy; (6) potential factors associated with intubation in this population.
We reviewed electronic medical records, laboratory and initial CT-scan findings for all patients. We collected data on age, sex, body mass index, medical history (smoking, chronic respiratory, cardiac, or kidney diseases, cancer), symptoms potentially related to COVID-19 (fever, cough, dyspnea, malaise, rhinorrhea, headache, vomiting, diarrhea, myalgia, and chest pain), and pre-hospitalization treatment (angiotensin-converting enzyme inhibitors, angiotensin II receptor blockers, corticosteroids, and non-steroidal anti-inflammatory drugs within the 7 days before hospital admission). Laboratory values at baseline were retrieved. Vital signs (respiratory rate, heart rate, mean blood pressure, oxygen flow rate) within the 24 h prior to FF-CPAP initiation as well as during the first hour of FF-CPAP therapy were collected. Duration of FF-CPAP delivery, the need for intubation, cardiac arrest prior to intubation and death within 28 days were also collected.
Data were analyzed using SPSS Base 20.0 statistical software package (SPSS, Chicago, IL, USA).
In the bench part of the study, normality of data’s distribution was verified using the Kolmogorov–Smirnov test of normality. Results were thus presented as means ± standard deviation. Comparisons between the conditions were performed using paired t test.
In the clinical assessment, no a priori sample size calculation was performed. The sample size was planned to correspond to the number of patients satisfying the inclusion criteria during the study period. Continuous data were expressed as medians (25th–75th percentiles) and compared using Mann–Whitney test for independent variables and Wilcoxon signed rank test for related variables. Categorical variables, expressed as percentages, were evaluated using Chi-square or Fisher exact tests as appropriate. The accuracy of respiratory rate measured before FF-CPAP initiation in detecting the need for intubation was assessed by receiver operating characteristic (ROC) curves. The threshold value of respiratory rate to predict intubation was then determined from analysis of ROC curves as the value that displayed the best compromise between sensitivity and specificity. Cumulative probability of intubation was evaluated using standard Kaplan–Meier actuarial techniques to estimate survival probability. Two-sided p values of < 0.05 were considered significant.