The Repurposed Use of Anesthesia Machines to Ventilate Critically Ill Patients With Coronavirus Disease 2019 (COVID-19)

Maurizio Bottiroli; Angelo Calini; Riccardo Pinciroli; Ariel Mueller; Antonio Siragusa; Carlo Anelli; Richard D. Urman; Ala Nozari; Lorenzo Berra; Michele Mondino; Roberto Fumagalli


BMC Anesthesiol. 2021;21(155) 

In This Article


Data Source

The COVID-ICU multicenter registry is an international data repository that started on March 1st, 2020 and is currently ongoing. It includes de-identified daily data relative to critically ill patients with confirmed COVID-19 admitted to the ICU. The anonymized data collection strategy utilizes a secure cloud-based online platform (Studytrax, Macon, GA).[9] The present study received approval by the Institutional Review Board of the coordinating institution (Massachusetts General Hospital, Boston, MA, USA, Protocol #2020P000760) with a waiver of informed consent. Patients included in this study were all admitted to Niguarda Hospital, Milan, Italy (Approval No.183–15,042,020).

Identification of the Study Cohort

In the first week of March, due to an overwhelming need for ICU beds, Niguarda Hospital (Milan, Italy) converted a large postoperative 27-bed ICU and an OR in two COVID-19 specific ICUs. The former unit was a 27-bed standard ICU, fully equipped with state-of-the-art ICU ventilators and was used entirely for COVID-19 from March 12th to April 15th, 2020. The second unit was emergently opened (from March 5th to May 15th, 2020) in an adjacent OR area. This OR ICU consisted of a 15-bed temporary ICU, including five beds from a PACU space and 10 beds placed in five separate ORs (two beds per single OR). This provisional unit was equipped with seven Anesthesia Machines, alongside eight standard ICU ventilators. These two units were equipped with the same medical, nursing, and support staff who usually worked in the standard postoperative ICU. Physicians and nurses within both the standard ICU and the OR ICU were highly trained individuals in the field of anesthesia and critical care. Particularly, the OR ICU was equipped with nurses skilled in the use of standard ICU ventilators and Anesthesia Machines. To further improve the Anesthesia Machines management in the OR-ICU, we planned weekly training sessions on their use by an expert anesthesiologist and two expert nurses of anesthesia. Moreover, every morning an expert nurse of anesthesia and an Anesthesiologist revised the Anesthesia Machines in function (filters, circuits and ventilator) in order to improve patient's safety.

During the study period, patients were admitted to either unit and assigned to an ICU ventilator, or an Anesthesia Machine, based on bed availability. Patients assigned to Anesthesia Machines were never switched from ICU ventilators and likewise switching did not occur for patients assigned to ICU ventilators.

Study Design

We designed a retrospective study comparing intubated COVID-19 patients whose care involved the use of Anesthesia Machine (AM group) to a cohort receiving care involving the use of standard ICU ventilators (ICU-VENT group) admitted during the same period. The "Strengthening the Reporting of Observational studies in Epidemiology" (STROBE) guidelines were followed.[10]

Patients 18 years-or-older admitted to either the standard ICU or the OR ICU were included. A confirmed diagnosis of SARS-CoV-2 infection was required. Patients receiving less than 48-h of mechanical ventilation were excluded. The primary endpoint was to assess a difference in 60-day survival between the two groups. Secondary endpoints included evaluation of differences between AM and ICU-VENT groups in terms of ICU or hospital length of stay, ventilator-free days, ICU and hospital free days, need for ECMO, need for tracheostomy, incidence of barotrauma (defined as spontaneous pneumothorax or pneumomediastinum) and need for emergency endotracheal tube exchange secondary to airway occlusion. We defined ventilator-free days as the total number of calendar days or part calendar days of unassisted breathing during the first 60 days after ICU admission; all patients who died by day 60 were assigned zero ventilator-free days. ICU and Hospital free days were defined respectively as the total number of calendar days the patient spent out of the ICU or Hospital during the first 60 days after ICU admission. All patients who died by day 60 were assigned zero ICU or Hospital-free days. All outcomes and variables included in this analysis were abstracted from the medical record using clinically documented values.

ICU Care

Patients were treated according to internationally recognized standards of care.[11] Details regarding mechanical ventilation settings, the use of respiratory failure rescue strategies, and COVID-19 specific therapies are reported in the online supplement (See "ICU care and COVID-19 specific therapies" in Supplemental Digital Content 1).

Anesthesia Machine Setup

The Primus™ workstation (Dräger, Lubeck, Germany) was the only Anesthesia Machine used. A heat and moisture exchanger with a filter (HMEF) and filter exclusive to the airway were used for every patient. The HMEF was placed at the endotracheal tube mouthpiece (DAR™ Adult-Pediatric Electrostatic Filter HME, Small, Medtronic, Minneapolis). The airway filter without HME (DAR™ Electrostatic Filter, Large, Medtronic, Minneapolis) was placed at the end of the expiratory limb of the circuit.[8] Both devices were routinely changed every 24 h or if there were signs of obstruction. For every Anesthesia Machine, a successful startup test was performed at baseline and at least every 72-h. A rebreathing circuit was in place, with a soda-lime scavenger to adsorb carbon dioxide. During approximately the first month of the study, the total fresh gas flow rate was maintained at 50–60% of the patient's minute ventilation, with the intent to spare sevoflurane and oxygen. Following the publication of consensus recommendations, the fresh gas flow was increased to around 80% of minute ventilation in patients receiving halogenates, and to over 100% in patients without inhaled anesthetic.[8,12]

ICU Ventilators

The SERVO-i Mechanical Ventilator (Getinge, Gothenburg, Sweden) is the primary ICU ventilator in use at the study institution. A similar filtering strategy to that used in the AM group was initially implemented, with an HMEF and an exclusive airway filter placed at the endotracheal tube and expiratory inlet on the ventilator, respectively. However, as soon as adequate supplies of personal protective equipment and proper isolation logistics could be guaranteed, active humidification was preferred to HMEF for most patients in the ICU-VENT group.

Statistical Analysis

The statistical plan was written after the data were accessed and no statistical power calculation was performed prior to the start of the study. Instead, the sample size was based on all available data from the time period in which both units were functioning as COVID ICUs.

Baseline characteristics are presented as median and interquartile range for continuous covariates, and proportions for categorical variables. The Mann-Whitney U test or Fisher's exact test were used for differences between the two groups (AM group vs. ICU-VENT group), and survivors vs. non-survivors, as appropriate. Kaplan-Meier survival analysis was used to compare 60-day survival between the two cohorts. Significance was assessed using a Log-Rank test. There were no censored survival data in this study.

In order to evaluate the impact of receiving care with an Anesthesia Machine on patient survival, we performed an adjusted and multivariable analysis. After confirming the proportional hazards assumption was met (p = 0.86), Cox Regression models were performed, in which we assessed the relationship with 60-day mortality. In the adjusted model only variables with p < 0.10 on the univariate screen were considered candidate variables for inclusion. Using this list of variables, backwards selection was then performed (considering p < 0.1 for exclusion at each step) in order to elucidate a final model. Hazards ratios (HR) and their associated 95% confidence intervals (CI) from the final model are presented. SPSS software v26 (Microsoft Corporation – Redmond, USA) and SAS 9.4 (SAS Institute Inc., Cary, NC) was used for data analysis. Two-sided p-values < 0.05 were considered statistically significant.

A sensitivity analysis was performed in order to address the possibility of biased estimates of the predictors by using a data-driven variable selection process. In this analysis 100 bootstrapped samples were obtained from the original dataset with replacement, and backward selection was similarly performed as above for each of the bootstrapped samples. Variables in the final sensitivity model were based on frequency and included the four most common variables in order to maintain model parsimony. Full results of the model selection process are detailed in Supplementary Digital Content 2Table 4.

Rates of missing data are reported in Supplementary Digital Content 2Table 3. No imputation was performed for missing data.