40th International Symposium on Intensive Care & Emergency Medicine 2020 – Part 2

Fig. 1 (abstract P513). Stationary graphical results. a) Air velocity in the airways, b) Temperature in the airways+lungs and c) Body temperature P513 Core warming as a treatment for coronavirus disease-2019 (COVID19) evaluation of the influence of lung density V Kostov, M Mercado-Montoya, N Bonfanti, E Gundert, AM Drewry, R Bedimo, K Kostov, S Shah, E Kulstad Walter Payton College Preparatory High School, Chicago, IL, United States; Universidad de Antioquia, Bioengineering Department, Medellín, Colombia; UT Southwestern Medical Center, Departments of Emergency Medicine and Anesthesia/Critical Care, Dallas, TX, United States; Washington University School of Medicine, Department of Anesthesiology, St. Louis, MO, United States; UT Southwestern Medical Center, Dallas, TX, United States; Life Science Angels, Chicago, IL, United States; Illinois Institute of Technology, Department of Bioengineering, Chicago, IL, United States; UT Southwestern Medical Center, Department of Emergency Medicine, Dallas, TX, United States Critical Care 2020, 24(Suppl 2):P513

Introduction: Lymphopenia appears to characterize COVID-19 [1]. Its depth on admission seems associated with prognosis [2], and exposes COVID-19 patients to secondary infections [3]. Current reports on immune response during severe COVID-19 are restricted to single cross sectional assessment on admission. Thus, we assessed the immune profile of patients admitted to the ICU for COVID-19-related ARDS and its evolution during the first week of stay. Methods: Prospective, observational study in consecutive patients admitted for COVID-19-related ARDS in our ICU. Peripheral blood was sampled on admission and on Day 3, 5 and 8 for immunophenotyping of lymphocytes (T, NK, B cells quantification, CD4+ and CD8+ T cells function), granulocytes (CD16-immatures granulocytes (IG), CD64+) and monocytes (mHLA-DR). Pro-inflammatory T cells cytokine production (IFN-ɣ, TNF-α, and IL-2) was assessed in 3 patients and 3 controls. Results: 13 patients were included (8 men (62%); 72 [64-76] y/o; 8 severe ARDS (62%)). All exhibited deep global lymphopenia that persisted to D7 (Fig 1). Increased proportion of regulatory T cells correlated with defective production of IFN-ɣ by CD4+ T cells. Accordingly, effector CD4+ frequencies were decreased; those of effector memories CD8+ and HLA-DR+, CD38+ activated CD8+ were increased. Increased number of granulocytes was observed with a rise of CD64 expression; IG were barely detected at admission. Monocyte counts were also increased with a major HLA-DR down regulation.

Conclusions:
Patients with COVID-19 showed sustained alterations of immune response with deep and persistent global lymphopenia which was correlated to increased T cell exhaustion and increased non-functional antigen presenting cells. These preliminary data warranting further confirmation suggest that therapies boosting host immunity could be a path worth following for COVID-19.

Introduction:
The primary aim of this study was to compare the effect of the COVID-19 outbreak on the mental health of ICU and non-ICU healthcare workers. The secondary goal was to look for gender differences in coping with this situation.

Methods:
A cross-sectional exploratory design was used. Participants were recruited in the first weeks of the Belgian outbreak. The inclusion criteria were healthcare workers working in high risk areas. Self-administered surveys concerning their education, work environment and personal experience used questions from standardized questionnaires. Correlations (95% level of significance) were searched using logistic regressions.

Results:
A total of 73 individuals completed the survey, giving a 75% participation rate. For our primary outcome, we didn't find a significant difference between ICU and non-ICU nurses. We did find a significant correlation for nurses originating from the OR (Odds = 2.07) and the orthopedic ward (Odds=2.07), and the change of behavior at home. For our secondary goal, we found a clear gender difference in coping with the outbreak. Females showed a greater presence of anxiety (M=0%; F=14%) and felt more worried about the situation. (M= 18.75%; F=45.5%). They were more afraid of getting infected (Odds= 1.02) and infecting their family (Odds=1.63). Additionally, the female gender had an influence on having behavioral changes at home (Odds=0.66). Furthermore, women tend to find more solace in their family and are more in need of a mental break in comparison to men. Other results are summarized in Table 1.

Conclusions:
Our study highlights that well-trained ICU nurses have the same risk for psychological distress than other nurses. Furthermore, female healthcare workers may need more psychological support. Protecting and educating healthcare workers seems to be an important component to address a pandemic outbreak. Finally, it could be interesting to compare this data with new surveys in the following months to evaluate the evolution of their mental health.  Introduction: Hub and spoke network is set in order to provide the best care for specialistic pathologies. New or unknown ills, therefore, may be preferentially centralized to hub hospital. Surge capacity, usually based on pre-disaster situation and different algorithms (despite some variations), follow the criterion "the more you have (Staff, Stuff, Space) the more you have to surge". Following this principles, COVID-19 pandemic have the potential to overwhelm hub hospitals, forcing a big reduction of specialistic activities for a long period with the risk of reducing the treatment capacity of many other life-threatening pathologies. We would like to present the activity of a network of hospitals in Northern Italy, and how they organized their ICU network to cope with the crisis.

Methods:
Eastern Piedmont is a region with 6 hospital, one hub university hospital and 5 spoke hospitals. The hub hospital is the only facility with advanced specialties. In order to keep specialistic services available, spoke hospitals proportionally increased their capacity more than the hub hospital by cutting their overall activity.

Results:
Spoke hospital ICU beds increment was 284% (min 225% max 417%) compared to a 155% increase of the hub hospital ( Figure 1). Total bed capacity went from 51 to 115 beds. The hub hospital managed to keep 7 specialistic elective surgical room open every day during the crisis, while 4/5 spoke hospitals held 1 each.

Conclusions:
An increment of 200% in capacity is usually considered the goal of crisis surge response but this could be achieved for brief period of times without affecting the quality of care. Considering the duration of the epidemic and the need to keep specialistic services available for all the patients, it is fundamental to adapt the current surge capacity model to take into account other hospitals in the network to deliver the best possible quality of care. We believe it is possible to use this organizational model to better estimate surge capacity during pandemics in similar hospital networks. There is increasing evidence that coronavirus disease 2019 (COVID-19) causes micro-and macrovascular thrombi. However, the best prophylactic anticoagulation strategy for critically ill patients with COVID-19 remains unclear with no concordant official guidance [1]. The purpose of this survey is to establish which anticoagulation strategies are being used for critically ill patients with COVID-19 within the Northwest of England.

Methods:
We conducted a 13-item online survey assessing local anticoagulation strategies and thromboembolic events in adult, critically ill patients with COVID-19. Clinical directors of critical care units within Northwest England were invited to participate and share the questionnaire with their consultant and junior doctor colleagues. The survey was conducted between April 17th and May 1st 2020.

Results:
There were 38 respondents, of which 66% were consultants and 29% were junior doctors. The majority of participants (91%) reported regularly using prophylactic anticoagulation. Half of respondents (49%) reported that they had changed their anticoagulation strategy to split dose low molecular weight heparin (LMWH), whilst 31% had changed to therapeutic dose LMWH. D-dimer was used to guide anticoagulation by 15% of respondents, with cut offs of 1000 and 3000 ng/ml as an indicator to give higher dose LMWH. Anti-Xa levels were used by 17% of participants to guide the dose of prophylactic anticoagulation.

Conclusions:
Prophylactic anticoagulation is still routinely used to treat critically ill patients with COVID-19 in the Northwest of England, but there is variation in practice, and deviation from usual practice due to clinical evidence of thromboses. There is an urgent need for data to guide the anticoagulation strategy in patients with COVID-19.

P521
Non-invasive brain compliance findings in patients with COVID-19 in a Brazilian intensive care unit SS Rojas 1   . ICU beds and OR capabilities available in Eastern Piedmont before and during the COVID-19 crisis to 1. ICP pulses were used to calculate the P2/P1 ratio with a mean of 1.08, standard deviation of 0.24 and median of 1.02 in a total 258 observations, with approximately 20 observations per patient ( Figure  1). At the 0.05 significance level, the true mean value of P2/P1 is between 1.009 and 1.159. One P2/P1 ratio differs from the others and represents a patient with rheumatoid arthritis, on treatment with corticoids and hydroxychloroquine.

Conclusions:
Respiratory distress may produce BC impairment even in early development of this disease. This equipment may represent a tool for the early detection of neurological changes in these patients and help in decision-making.

Introduction:
The novel coronavirus SARS-CoV-2 and the resulting disease COVID-19 is known to have mild as well as critical courses requiring vv-ECMO therapy. Also, COVID-19 seems to be associated with hypercoagulability and thrombosis [1]. Indeed, our center as well did witness an increase of thrombotic complications especially in the extracorporeal circuit in these patients, particularly of the centrifugal pump. Thus, we investigated the rate of vv-ECMO complications in these patients.

Methods:
All COVID-19 cases admitted on our ICU who received vv-ECMO therapy were compared with vv-ECMO patients treated on our ICU during the years 2018 and 2019. Aside from baseline characteristics, duration of ICU stay and length of vv-ECMO therapy, all circuit related complications resulting in partial or complete exchange of the extracorporeal system were registered. Events clearly documented as non-thrombotic were excluded. Also D-Dimer measurements prior to these events were analyzed.

Results:
In total, 66 patients were analyzed, 55 non-COVID-19 vs. eleven COVID-19-related. To this date, six COVID-19 patients were still treated on our ICU. The two groups did not differ in age, BMI and severity of illness (RESP Score mean 0.89 vs 1.09, p 0.85). A significantly higher rate and probability of centrifugal pump thrombosis needing exchange could be observed in the COVID-19-group (see figure 1). In total, 16 centrifugal pump thromboses did occur in the non-COVID-19-group compared to nine in the COVID-19-group. In addition, the most recent D-Dimer measurements prior to the events were significantly lower in the non-COVID-19 group (mean 15.48 vs 26.59, p < 0.05).

Conclusions:
The coronavirus SARS-CoV-2 induced infection is associated with higher rates of thrombotic events of the extracorporeal system during vv-ECMO therapy and the medical team should be watchful to counteract these complications. Introduction:

Introduction
In a previous study the increase in levels of some microbial phenolic metabolites (phenyllactic (PhLA), p-hydroxyphenylacetic (p-HPhAA), p-hydroxyphenyllactic (p-HPhAA) acids) simultaneously with a rise in levels of mitochondrial dicarboxylic (succinic and fumaric) acids were detected in serum samples of septic patients. A close direct correlation of serum concentrations of these metabolites confirms the participation of the phenolic metabolites in the development of mitochondrial dysfunction in sepsis [1]. The purpose of this study was to find out whether there are similar correlations in the cerebrospinal fluid (CSF).

Results:
Phenolic acids were found in CSF samples at low concentrations 0.1-1.1 μM. Dicarboxylic acids were detected in different concentration ranges: succinic acid 0.6-111 μM, fumaric acid 0.1-5.3 μM. Close direct correlations (r) of microbial and mitochondrial metabolites were revealed (Table 1). Correlation between fumaric and succinic acid was 0.78. These correlations are similar to those obtained previously in the sera of septic patients [1].

Conclusions:
Correlations of phenolic and mitochondrial metabolites in CSF are similar to those previously found in the blood of critically ill septic patients. The new data relate to the revealed correlation between the metabolite of microbiota p-HBA and mitochondrial metabolites which needs to be explained in further research.

Conclusions:
This retrospective study showed that doctors and nurses are not compliant with the NICE guidelines as 50% of CT head were not warranted. CT head was not being performed within the 1st hour of head injury presentation: delay of approximately 2-hour 20minutes between triage and assessment by a medical practitioner. 75% of data suggest CT head was being reported within 1 hour, which is compliant with NICE guidelines. Moreover, it was impossible to say whether some of the risk factors such as vomiting, seizures or LOC was not assessed or merely not documented. This data suggested that an intervention is required to improve the documentation process and compliance of doctors and nurses in order to improve the delays between triage, assessment and performing CT head. Furthermore, the cost of unwarranted CT scans and patients being exposed to unnecessary radiation could be reduced. Hence, we are implementing two quality improvements projects in order to address the above issues.

Introduction:
Patients with moderate traumatic brain injury (mTBI) are 1.5 times more frequent than those with severe TBI and some of them will develop secondary neurologic deterioration (SND) within the first 7 days [1,2]. However, identifying at risk patients of SND is still challenging. This study aimed to determine risk factors associated with SND after mTBI.

Methods:
We conducted a single center retrospective study. Adults admitted in Brest hospital between 2015 and 2018 for mTBI, defined by Glasgow coma scale (GCS) score 9-13, were eligible. We assessed clinical patients' characteristics in the prehospital setting, at admission and during the first 72 hours of hospitalization. Biology, transcranial Doppler (TCD) and CT examinations were also reported. SND was defined either by a decrease in GCS or by a deterioration in neurologic status sufficient to warrant intervention like mechanical ventilation, transfer to the ICU or neurosurgery [3]. Factors statistically associated with SND were identified. Results: 147 patients were included, mean patient age was 51.5 years (± 18.94 years) and 81% of patients were men. Mean GCS score was 11.4 (± 1.63). 46 (31.3%) showed SND and 14-day mortality rate was of 18.4%. Patients with SND were older (p<0.001), had higher hypoxemia and intubation rates (p=0.011, p=0.002 respectively) before admission. They had significantly higher IGS2 scores on admission (p <0.001), and more frequent hyperglycemia (>8 mmol/l) and hypotension (SBP < 90 mmHg) (p=0.001 and p<0.001 respectively) during the first 72 hours. Neuro-worsening was also associated with abnormal TCD (pulsatility index > 1.4) and abnormal head CT scan on admission (higher categories at Marshall CT score) (p=0.004 and p<0.001 respectively). 32% of patients with SND required neurosurgery.

Conclusions:
We report for the first time the largest study about early outcome after mTBI. About a third of mTBI patients showed SND and early factors could be used for determining at risk patients. For these patients, treatment includes surgical resection and postoperative management in the ICU where multimodal monitoring, goaldirected therapy and analysis of serum markers are essential. Serum lactate is a commonly used marker of global tissue perfusion, so it correlates to prognosis and clinical outcomes of patients. Thus, blood lactate has been used widely in critically ill patients. This study aims to assess the behavior of serum lactate as an outcome predictor in patients undergoing surgical resection of brain tumors.

Methods:
After Institutional Review Board approval was obtained, a crosssectional study was conducted. All patients older than 18 years undergoing brain tumor resection between January 2015 and December 2019 were included. We excluded septic patients and those with hyperlactatemia. For analysis patients were divided into two groups according to their serum lactate levels: <2 mmol/l (normal) and ≥2 mmol/l (hyperlactatemia). Measured clinical outcomes included duration of mechanical ventilation, reintubation, length of stay in ICU, length of hospital stay, readmission to the ICU and mortality. P<0.05 was considered statistically significant.

Results:
A total of 225 patients were analyzed, of which 154 patients (64.4%) had a normal level of serum lactate and 80 patients (35.6%) presented hyperlactatemia. There were no statistically significant differences in duration of mechanical ventilation (p=0.07), reintubation (p=0.06), length of stay in ICU (p >0.05), length of hospital stay (p>0.05), and readmission to the ICU (p=0.09). Mortality was higher in patients with hyperlactatemia (p=0.04).

Conclusions:
Serum lactate is not a good predictor of postoperative clinical outcomes in patients undergoing brain tumor resection. However, it may predict mortality in these patients, so it is necessary to conduct another study with larger sample size.

P531
Targeted temperature management and neurological outcome after out of hospital cardiac arrest J Ellis, C Pritchett, J Thorns, BA Alberts, R Tedstone, M Spivey Royal Cornwall Hospital, Intensive Care, Truro, United Kingdom Critical Care 2020, 24(Suppl 2):P531

Introduction:
Our 2017 audit of targeted temperature management (TTM) in patients admitted to ICU after an out of hospital cardiac arrest (OOHCA) found that patient temperatures were above those recommended by the Resuscitation Council (RC). Previous data showed that cooling was best achieved in 2013; the year with the highest use of invasive cooling. Following the audit, additional education was given to ICU staff on TTM. We repeated this audit to assess the effect of that intervention.

Methods:
Patients admitted following OOHCA between 01/04/2017 and 31/06/ 2019 were identified, excluding patients aged under 18. Baseline data included gender, age, total 'downtime', presenting rhythm, and survival to discharge. Data was also collected on cooling method, whether TTM was documented in the medical plan, and temperature at 0, 12, 24, 48 and 72 hours after ICU admission. Cerebral performance category (CPC) was gaged using clinic letters. Results: 96 patients were admitted to ICU following OOHCA between 01/04/ 2017 and 31/06/2019. Of these, 71.9% were male, the mean age was 63, and 38.5% of patients admitted survived to hospital discharge. TTM was documented in the medical plan for 58.3% of patients. The percentage of patients receiving invasive cooling in 2017-19 was 22.6%, 25.0% and 29.4% respectively. Across all years, the mean temperatures at 0, 12, 24, 48 and 72 hours following ICU admission were 34.9°C, 36.4°C, 36.5°C, 36.6°C and 36.9°C. Comparison between patients with a CPC of 1 and CPC of 5 showed no significant difference in temperature during ICU stay.

Conclusions:
The results of this audit are similar to those obtained in 2017, implying additional education has not resulted in better adherence to RC guidelines of TTM below 36°C for 24 hours. However, average temperatures in the first 72 hours are below the recommended 37.6°C. In the future, it would be interesting to compare incidences of pyrexia in those cooled passively compared to those given invasive cooling and to assess for differences in outcomes.

P532
Outcomes from out-of-hospital cardiac arrests in Kaunas Introduction: EuReCa TWO study [1] has recently reported large variation in out-ofhospital cardiac arrest (OHCA) outcomes across the Europe. Unfortunately, Lithuania did not participate in that study. We sought to fill the gap and describe the epidemiology and outcomes from Kaunas, the second largest Lithuanian city with a population of 0.29 million (9.7% of all population in Lithuania).

Methods:
The incidence, demographics and outcomes of patients who were treated for an OHCA between 1st January 2017 and 31st December 2018 in Kaunas Emergency medical Service (EMS) were collected and are reported in accordance with 2014 Utstein recommendations.

Results:
In total, 524 OHCA cases of EMS treated cardiac arrests were analyzed. The mean age was 68.6 (SD = 16) years and 66.3% were male. 74% OHCA cases occurred at home and 63.2% were witnessed. The initial rhythm was shockable in 28.6% and non-shockable in 66.6% of all cases. Return of spontaneous circulation (ROSC) at hospital transfer was 28.2% in 2017, and 31.8% in 2018. Survival to hospital discharge was 12.1% in 2017, and 12.7% in 2018. Survival to hospital discharge in Utstein comparator group (OHCA witnessed by a bystander, and shockable rhythm) was 31.8% in 2017, and 44.2% in 2018.

Conclusions:
ROSC and survival to hospital discharge in Kaunas were slightly better to those reported as median in EuReCa TWO study.  Over a third of critical illness survivors suffer from mental health problems following hospital discharge. Memories of delusional experiences are a major risk factor. Research on the formation of fear demonstrates that if mitigating information about a traumatic event is introduced during the time of memory formation/upon its subsequent recall, the emotional experience of the memory may be modified. Given that semantic processing continues during altered states of consciousness, we trained doulas to provide early psychological support for the critically ill in parallel with medical treatment. Stakeholder acceptance is a vital part of the intervention success. Methods: ICU nurses who witnessed the interventional session were given a paper questionnaire with multiple choice and open-ended items regarding their impressions. Results: 43 patients received the intervention; 32 nurses provided feedback.
When asked about what they liked, all provided positive comments about the intervention (see Figure 1). When questioned about what they didn't like, only 11 (34%) commented including that positivity may not always reflect reality, that it may overstep the nurse's job, that the intervention should be longer, or recommending better coordination with nurses/shift changes. When asked whether communicating with the ICU doulas was also helpful for the nurse, 26 (81%) answered yes. Specifically, nurses liked: aspects of the intervention that provided patients with reorientation and reassurance; ICU doulas being a liaison with the family and team and answering questions; and, being able to step away to complete other tasks while someone was there with the patient. Of the 6 respondents who did not find the intervention helpful to nurses, only two elaborated that the intervention appeared redundant to the care nurses already provide.

Conclusions:
The majority of bedside ICU nurses welcome the presence of ICU doulas and the psychological support intervention they provide.

Introduction:
High thrombus burden is an independent risk factor for death and complications, including no reflow, during primary percutaneous coronary intervention (PCI) for STEMI. The aim was to investigate whether a strategy of mechanical thrombectomy in combination with deferred stenting is associated with a reduced incidence of slow-or no-reflow, and other thrombotic complications compared with stenting in patients with high thrombus burden.

Methods:
A total of 210 patients with STEMI and high thrombus burden treated with thrombus aspiration in combination with glycoprotein IIb/IIIa inhibitors with or without stent implantation. Patients were divided into 2 groups: non-stent PCI group (n = 105) and stent PCI group (n = 105). The end points were a myocardial blush grade of 0 or 1 (defined as absent or minimal myocardial reperfusion, respectively) and the postprocedural frequencies of a TIMI flow grade of 3, 48 hours after primary PCI, complete resolution of ST-segment elevation immediately after primary PCI, target vessel revascularization, reinfarction, death, and the combination of major adverse cardiac events by 30 days after randomization.

Results:
A myocardial blush grade of 0 or 1 occurred in 26.3% of the patients in the stent PCI group and in 17.1% of those in the non-stent PCI group (p < 0.05). Complete resolution of ST-segment elevation occurred in 86.6% and 78.2% of patients, respectively (p = 0.35). At 30 days, the rate of death in the stent PCI group and non-stent PCI group was 1.7%, and 1.0%, respectively (p = 0.33), and the rate of adverse events was 12.1% and 2.2%, respectively (p < 0.01).

Conclusions:
Mechanical thrombectomy in combination with glycoprotein IIb/IIIa inhibitors without stenting is applicable and effective method in a large majority of patients with myocardial infarction with ST-segment elevation and high thrombus burden. It results in better reperfusion outcomes than conventional PCI with stent, irrespective of clinical and angiographic characteristics at baseline. Cardiac function is known to be negatively impacted by sepsis. Monitoring cardiac output (CO) and stroke volume (SV) trends over the course of treatment may provide insight into cardiac function and predict patient outcome. The goal of this study was to explore the relationship between the change in cardiac output over time in septic shock. Methods: FRESH is a randomized controlled study evaluating the hemodynamics in critically ill patients with sepsis or septic shock (NCT02837731). Patients randomized to PLR guided resuscitation received hemodynamic monitoring for 72 or until ICU discharge, whichever occurred first (Starling SV, Cheetah Medical). Patients that exhibited an improvement in cardiac output at 12, 24, 36 and 48 hours were compared to those who did not exhibit improvement. Overall improvement in cardiac output (first CO measurement compared to last CO measurement) was also compared between groups. Results: 90 patients with septic shock received hemodynamic monitoring over a 72 hour monitoring period. 60 % were female, and the average age was 61 years. Overall, 44% of assessments demonstrated a fluid responsive positive response after receiving initial resuscitation fluid of 2.3 L. Patients who exhibited improved CO at 48 hours received less fluid over the course of their ICU stay. This difference was consistent both when pre-enrollment fluids were included (5985.6 ± 2293.0 vs 8667.2 ± 3750.6, p=0.01) and only when post enrollment fluids were used (4056.2 ± 2149.2 vs 6296.5 ± 3646.8, p=0.024). Notably, patients who exhibited an overall improvement in CO also exhibited a decreased in serum creatinine over the study period (-0.28 ± 0.27 vs 0.52 ± 0.23, p=0.029) ( Figure 1).

Conclusions:
We have previously shown that patients who improve CO in response to the resuscitation exhibited improved outcome. Trending cardiac output over the ICU stay revealed additional usefulness in predicting patients with improved outcome. The results highlight the importance of trending hemodynamics in therapy. Introduction: SARS-CoV-2 infection may progress to acute respiratory failure (ARF), cardiac injury, renal failure and liver dysfunction. Myocardial depression in sepsis represents a major predictor of unfavorable outcome, leading to a mortality rate close to 70%. Sepsis-induced cardiomyopathy (SIC) is characterized by a global but reversible myocardial dysfunction with left ventricular (LV) dilatation and depressed LVEF [1]. While LVEF calculation is easy to acquire, it remains variable with beat to beat and highly dependent on loading conditions. Speckle tracking echocardiography (STE) is a recent imaging technology allowing early detection of LV dysfunction, prior to decrease in LVEF [2].

Methods:
We investigated a COVID-19 patient with severe ARDS (P/F ratio <100) and septic shock. A TTE was performed to assess LV function and standard echocardiographic variables. QLAB cardiac analysis was performed to assess strain imaging.

Conclusions:
We illustrated STE findings in a COVID-19 patient admitted with hypoxemic ARF and septic shock. While LVEF was normal with the Simpson's biplane method, average strain in 4-Ch view was -13.4% and GLS was -13.6%, indicating reduced LV longitudinal function. In this regard, STE may improve prognostication over LVEF to assess LV systolic function in critically-ill patients with SARS-Cov-2 infection.

Introduction:
The incidence of pleural effusions (PEs) radiologically in acute decompensated heart failure was estimated by 50 % [1]. The prevalence of PEs using chest ultrasound in patients with heart failure with preserved ejection fraction (HFpEF) was not studied before. The aim was to determine the prevalence and severity of PE using bedside chest ultrasound in patients with HFpEF.

Methods:
We prospectively evaluated 85 patients admitted to the coronary care unit with acute pulmonary edema. 27 patients had LVEF below 50% and they were excluded. We used bedside ultrasonography to document the presence of pleural effusions ( Figure 1) and estimate the amount using Goecke 2 formula. Results: 58 patients were recruited with estimated LVEF above 50% and median age of 73 years. Pleural effusion was detected by chest ultrasound in 54 patients (94%) with 91% being bilateral and of moderate amount. The mean value of E/A and E/E' ratio was 1.92 and 19.9 respectively where mean deceleration time (DT) was 171 ms, however, these parameters were not related to severity of pleural effusion (p= 0.52, 0.98 and 0.7 respectively).The N-terminal pro-brain-type natriuretic peptide (NT-proBNP) was significantly higher in patients with pleural effusion (p=0.046).

Conclusions:
PEs are present on chest ultrasound in 94% of patients with HFpEF and mainly bilateral. The degree of PEs was moderate and was not related to the LV indices of HFpEF.

Introduction:
In two independent observational cohorts MR-proADM values identified low disease severity patients without risk of disease progression in the ED with no 28 days mortality that wouldn´t require hospitalization [1]. This interventional study aimed to provide guidance to safely reduce the number of hospital admissions by implementing a MR-proADM algorithm that identifies low risk patients not requiring hospitalization.

Methods:
A randomized controlled interventional multicenter study in 4 EDs in Spain. The study protocol was approved by the Ethics Committees of the hospitals. Control arm patients received standard care. MR-proADM guided arm patients with low MR-proADM value (≤0.87 nmol/l) were treated as out-patients, with high MR-proADM value (>0.87 nmol/l) were hospitalized ( Figure 1). The hospitalization rate was compared between the study arms.

Results:
Two hundred patients with suspicion of infection were enrolled. In the MR-proADM guided arm the hospital admission rate in the intention-to-treat population (ITT-P) was 17% lower than in the control arm (40.6% vs. 57.6%, p=0.024) and 21% lower in the perprotocol population (PP-P) (37.2% vs. 57.6%, p=0.009). The mortality rate in the out-patients group was 0% (5.1% in hospitalized patients in ITT-P). No significant differences for the safety endpoints re- admission and re-presentation rates were observed. The re-admission rate was slightly but not significantly higher in the MR-proADM guided arm compared to the control arm (PP-P: at 14 days 9.3 % vs. 7.1%, p=1; at 28 days 11.1% vs. 9.5%, p=1). The rate of 28 days representation was slightly lower in the MR-proADM guided arm compared to the control arm (20.4% vs. 26.2%, p=0.668; PP-P). Conclusion: Implementing a MR-proADM algorithm efficiently and sustaining optimizes ED workflows. Hospitals can highly benefit from a reduced rate of hospitalizations by 21% using MR-proADM. Introduction: ICU-acquired weakness has shown to persist beyond the ICU stay and to associate with long-term functional impairment of ICU survivors. Underlying mechanisms remain unclear, but illness-induced aberrant DNA methylation could be involved due to its potential longlasting impact on gene expression. Recently, DNA methylation alterations have been identified in peripheral blood of pediatric ICU patients, which were found to explain part of their long-term developmental impairment [1]. Whether DNA methylation in muscle is altered by critical illness is unknown.

Methods:
We extracted DNA from skeletal muscle biopsies collected on day 8±1 in ICU from patients included in the EPaNIC trial (n=188) and from 20 matched healthy controls. Genome-wide DNA methylation was determined with Infinium® HumanMethylation EPIC-BeadChips, which interrogate more than 850000 CpG sites. Methylation status of individual CpG sites and of DNA regions of patients and controls were compared, using stringent corrections for multiple comparisons.

Results:
In DNA extracted from ICU patients, 565 CpG sites, associated with 400 unique genes, were differentially methylated as compared with controls, with an average difference in methylation of 3.2% (SEM 0.07%, p<0.00005). Many of the associated genes were identified as highly relevant for muscle structure, function and/or weakness. In addition, in patients as compared with controls, we identified two hypomethylated regions (family-wise error rate <0.05), spanning 18 and 3 CpG sites in the promotor regions of the HIC1 and NADK2 genes, respectively. HIC1 and NADK2 play important roles in muscle regeneration and postsynaptic acetylcholine receptor regulation, and in mitochondrial processes, respectively.

Conclusions:
The DNA methylation signature in skeletal muscle is altered by critical illness, which may provide a biological basis for the long-term persistence of weakness in ICU survivors.

Introduction:
Postoperative morbidity still remains a major concern for the survival and cost of treatment, negatively affecting either of them. The aim of the study to describe the most common complications after abdominal surgery (POC) and their impact on length of hospital stay (LOS) and in-hospital mortality. We also explored which patients were more at risk for these complications.

Conclusions:
Complications are frequent after abdominal surgery (incidence: 40.74%) and the most common complications were pulmonary complications. Postoperative complications increase length of hospital stay and in-hospital mortality. Efforts should be done to prevent complications, modifying their risk factors and optimizing the treatment of underlying diseases.

Introduction:
Intensive care unit (ICU) survivors are at risk of emergency hospital readmissions. A risk prediction tool may help identify high-risk patients in need of preventative interventions after discharge. Scottish Patients at Risk of Readmission and Admission (SPARRA) predicts 1year risk of emergency hospital admission in the general Scottish population. The aim of this study was to externally validate SPARRA in ICU survivors. We also aimed to compare the performance of the two SPARRA scoring systems: automated scores (imported from national database) and manual scores (generated by entry of predictors into online risk calculator).

Methods:
Retrospective study of 264 patients discharged from a single ICU from 2015-2018. Scores were matched to time of hospital discharge: automated scores were obtained for all 264 participants; 60 patients were also scored manually. Observed rates of emergency admission in the year from hospital discharge were recorded. Discrimination was assessed by area under the ROC curve (AUC); calibration plots were drawn and the mean absolute error (MAE) between predicted and observed readmission rates was calculated. Results: 160 patients (60.6%) had ≥1 emergency admission in the follow-up year. Manual scores had poor discrimination: AUC 0.59 (95% CI 0.44-0.74); automated SPARRA scores had moderate discrimination: AUC 0.73 (95% CI 0.65-0.81). The ROC curves are shown in Figure 1. Both scoring systems were poorly calibrated with underestimation of admission rates across all risk strata. The MAE was 16.5% (95% CI 10.2-22.8) in automated scores and 20.9% (95% CI 13.8-28.0) in manual scores.

Conclusions:
Emergency hospital readmissions are common in the year after hospital discharge following an intensive care admission. Automated SPARRA scores performed better than manual SPARRA scores in parameters of discrimination and calibration. However, neither method reliably predicted emergency readmission in this cohort. A new risk prediction model should be developed for use in ICU survivors. Early goals of care documentation (GOC) is vital to avoid futile/unwarranted interventions in the event of deterioration. It is however unclear whether being frail impacted GOC timing. We aimed to investigate the impact of frailty, as measured by Hospital Frailty Risk Score (HFRS), on timely GOC and outcomes in all newly admitted older patients.

Conclusions:
Older frail patients were more likely to have timely GOC than older non-frail patients. Frail patients had more RRCs, longer hospitalization and increased hospital mortality. Early GOC may avoid burdensome treatments. Introduction: Pancreatic robotic surgery (PRS) allows less invasivity and a better anatomical view than open surgery. From the anesthesiological point of view, this approach requires greater attention to hemodynamics and ventilation as complications can occur during or after surgery. This observational study aimed at evaluating the first 2 years of PRS procedures so as to foster a common anesthesiological management.

Methods:
We analyzed the data of patients submitted to PRS between December 2017 to January 2020. Standard general anesthesia was performed (induction with propofol, opioid and rocuronium for neuromuscular blockade, maintenance with alogenate or TIVA/TCI), using VCV or PCV-VG ventilation; vital signs and TOF were constantly monitored intraoperatively. We recorded baseline clinical characteristics, pre-operative SOFA score, positive pre-operative surveillance rectal swabs, clinical(AKI, pneumonia, fever and/or sepsis) and surgical complications (pancreatic fistulae, abdominal collections, bleeding); 1st post-operative day(POD) SOFA score; ICU and hospital length of stay(LOS). Results: 66 pts(14-77 y, 21 males) were analyzed, ASA 1-3,mean BMI 20.3 kind of interventions were performed: 5 enucleoresections, 27 spleen preserving distal pancreatectomy, 34 distal splenopancreatectomy. Mean duration of surgery was 426' with a mean laparoscopic IAP of 12 mmHg. TAP block was performed in 21 pts. Pre-operative SOFA score was 0 for all patients. Pre-operative rectal swabs showed 1 case of E. coli ESBL. Pre-and PO SOFA score were not statistically different.9 patients were admitted to ICU with maximum LOS of 5 days. Mean LOS in the hospital was 11 days. 65% of patients had complications (21 sepsis, 23 fistulae, 16 abdominal collections, 5 PO bleeding, 4 pneumonias, 2 pneumothorax, 1 AKI). We recorded 1 case of PO positive rectal swab (C. freudii ESBL) and 3 cases of MRSA bloodstream infection.

Conclusions:
IO management did not alter pts' SOFA score but postoperative complications prolonged both ICU and hospital LOS.

P544
Arterial lactate concentration on intensive care unit admission predicts long term mortality after liver transplantation LD Drouard CHU Erasme, Anesthesiology, Brussels, Belgium Critical Care 2020, 24(Suppl 2):P544 Introduction: Arterial lactate is a routinely measured biomarker that reliably detects tissue hypoxia. High risk patients undergoing major surgery are at risk of developing hyperlactatemia. The aim of this study was to investigate the association between arrival ICU lactate concentration and mortality at 1 year, postoperative complications and 30-day mortality after liver transplantation.

Methods:
We identified all liver transplant patients in Erasme hospital from September 2013 to December 2019.The primary outcome was to determine if early hyperlactatemia was associated with mortality at one year. Other outcomes included the association between increased lactate, postoperative complications and 30-day mortality. A multivariate analysis determined the independent association of lactate on one year mortality. Receiver operating characteristic (ROC) curves were established for one-year mortality, 30-day mortality, and postoperative complications. 95% IC were calculated with the Delong method.

Conclusions:
In this retrospective study we report an association between early postoperative lactate level and major postoperative complications, 30-day mortality, and 1 year mortality. Lactate concentration on ICU arrival is an easily accessible biomarker that may predict outcome and identify high risk patients. Introduction: Stress-induced hyperglycemia is a common complication associated with higher morbidity and mortality in ICU patients. The Stochastic TARgeted (STAR) glycemic control (GC) framework provides consistent, safe, effective control in different ICUs and countries. It is a patient-specific, risk-based protocol controlling both insulin and nutrition dosing. This study analyses safety and efficacy of STAR at the University Hospital of Liège, Belgium, and assesses the impact of also modulating nutrition on GC outcomes.

Methods:
Patients are included after 2 blood glucose (BG) > 145 mg/dl. The study compares STAR (N=14; modulating insulin and nutrition) and STAR-IO (N=15; controlling insulin only, leaving nutrition at clinical discretion). STAR controls nutrition between 30-100% of goal feed if insulin alone cannot safely reduce BG. The target band is 80-145 mg/ dl for both arms. GC was stopped after 72h or if BG in target at insulin rate ≤2U/h for 6 hours. Performance is assessed by %BG in target and the median [IQR] BG, and safety by %BG below (<80 mg/dl) and above (>145 mg/dl) target. Ethics approval was granted by the University Hospital of Liège Ethics Committee. Results: Table 1 shows better performance for STAR (83% vs 78% BG in target), and better safety from hypoglycemia (0.7% vs 1.5%) and hyperglycemia (17% vs 21% Introduction: Sex differences in the metabolic response to critical illness are unknown. This retrospective analysis examines potential differences in the evolution of insulin sensitivity (SI) and its variability (%ΔSI) between sexes. Significant differences would suggest differences in the metabolic stress response and glycemic response to insulin therapy, and, thus, the need for more personalized glycemic control (GC).

Methods:
Retrospective data from 145 ICU patients (N=8710 hours) are used to hourly identify hourly model-based SI and its rate of change %ΔSI in 6-hour blocks from ICU admission to 72 hours. The evolution of SI and %ΔSI are compared for males and females. Hypothesis testing (95% confidence interval (CI) bootstrapped difference in medians) assesses if differences are significant, and equivalence testing assesses if differences are clinically equivalent.

Results:
Females have significantly lower SI levels than males (p<0.05), and this difference is not clinically equivalent (Figure 1; top). Differences in %ΔSI are not significant (p>0.05), and these differences are clinically equivalent (Figure 1; bottom).

Conclusions:
Given significantly lower SI levels, but equivalent variability, for women, equally safe and effective GC should be achievable for both sexes. However, women may require more insulin to achieve these goals. GC protocol designs should thus account for these differences in the future.

P547
Reduced workload in the STAR glycemic control framework: quantifying the safety trade-off Introduction: The Stochastic Targeted (STAR) glycemic control (GC) framework is a model-based, risk-based insulin and nutrition protocol providing safe, effective GC to nearly all patients. STAR currently uses 1-3 hourly measurement intervals, averaging 12 measurements per day. This study assesses the impact of increasing its measurement interval up to 6-hourly on GC safety and efficacy.

Methods:
STAR identifies patient-specific model-based insulin sensitivity (SI) using clinical data and forecasts its future variability to obtain an insulin and nutrition intervention minimizing hypoglycemic risk. STAR also modulates nutrition intervention to reduce persistent hyperglycemia for highly insulin resistant patients. STAR is adapted to allow 1-6 hourly measurement interventions. Validated virtual trials assess GC outcomes of the original STAR (STAR-3H) and the STAR 1-6 hourly (STAR-6H) protocols on the same 681 underlying virtual patients based on retrospective clinical data.

Results:
Results are presented in Table 1

Conclusions:
The risks associated with the reward of reducing workload are slightly reduced safety, performance, and nutrition rates. Overall, despite using these longer measurements intervals, STAR still managed to provide highly safe, effective GC for nearly all patients.  slightly higher than normal. Moreover, in patients Antithrombin III was decreased to 77.9 (66.4-86.9)% (normal over 83%), the level of D-dimers was increased 0.65 (0.43-2.14) μg / ml (normal to 0.26 μg/ ml), and was increased activity of von Willebrand factor 364.5 (331.5-456.8)% (normal to 125.2% -169.7%).

Conclusions:
In patients with COVID-19 infection antithrombin III was increased, the level of D-dimers was increased, and the activity of von Willebrand factor was increased. Introduction: Anemia is a common finding in preoperative period, and often increases the need for blood transfusion. The aim of the study: to explore the relation between moderate or severe preoperative anemia (Hb levels <10 g/dl) and postoperative outcome and also the impact of preoperative blood transfusion on the outcome.

Introduction:
The blood gas sample is the most used paraclinical test at the ICU and recommended to be analyzed instantly without any storage. At a clinical setting with acute and unexpected work tasks, blood gas samples might be left unanalyzed and, consequently, replaced by a new patient sample. We hypothesize that minor storage with delayed analysis is unproblematic. We aim to define the timespan venous blood gas samples maintain stable values. The study is ongoing.

Methods:
From each of ten healthy participants we have obtained 16 blood samples. This is followed by an ongoing inclusion of ICU patients with an abnormal pH, and we increase the number of samples to 20. Sampling is done through an intravenous access (central or elbow flexion), a three-way stopcock and 1.8 ml heparinized syringes. Samples are either packed in separate ice bags or stored at room temperature. The ABL800 FLEX analyses samples for pH, pO2, pCO2, HCO3-, glucose and lactate. The first and last obtained samples are analyzed immediately and used as a reference, followed by a cooled and un-cooled sample every 15th-18th minute. In the case of the ICU samples, the test period is prolonged from two to three hours.

Introduction:
Hypogammaglobulinemia [1] is one of the important predisposing factor for infection in chronic lymphoid leukemia (CLL) and to prevent it, needs usually the prophylactic administration of immunoglobulins. Nevertheless, intravenous immunoglobulins (IVIg) as curative treatment [2] was never well evaluated in these patients. So the objective of our study was to analyze the mortality and the morbidity of 30 patients with CLL admitted in an university hospital ICU for severe pulmonary infections and receiving or not administration of IVIg as curative treatment.

Methods:
We conducted a prospective observational study from April 2018 to April 2020 and selected 30 patients with early diagnosed and untreated CLL. Patients were 75+-10 years old and were mechanically ventilated for severe pneumoccoci bilateral pulmonary infections. We divided the patients in two groups: A first group (experimental group) of 15 patients received IVIg 1g/kg associated with antibiotics on the first day of the diagnosis of bacterial pneumonia and a second group (control group) of 15 patients received placebo and antibiotics. In each group, we performed a dosage of IgG before and one week after the administration of IVIg.

Results:
Before IVIg, the level of IgG was in the normal range (8 g/l) in the two groups. One week after IVIg, in the experimental group, the level was significantly higher than in the control (ctrl) group (14.73+-1.16 VS 8+-1.25 g/l) p<0.0001 (

Introduction:
Hemophagocytic lymphohistiocytosis (HLH) is a life threatening condition caused by excessive stimulation of lymphocytes and histiocytes that secrete inflammatory cytokines. In this study, the effect of therapeutic plasma exchange in HLH and MAS has been observed.

Methods:
The patients who were diagnosed with HLH and MAS in PİCU, were included in this study. The age, sex, PRISM scores, diagnosis, the length of hospital stay, C-reactive protein and procalcitonin levels, thrombocyte counts, fibrinogen and ferritin levels, neutrophil and lymphocytes counts, hemoglobin, triglycerides, d-dimer and albumin levels and the need for therapeutic plasma exchange of the patients were noted.

Results:
Twenty male (74.1%), seven female (25.9%) patients were included in the study. The median age of the patients was 3.5(0.6-5.8), the length of hospital stay was 11(5.0-22.0). Genetic causes existed in 13 patients (48%), 6 (22.2%) of the patients had causes secondary to infections, 5 of them (18.5%) had rheumatologic causes, and 3 (9.3%) of them had other causes. Therapeutic plasma exchange was performed on 21 (81.5%) patients. Intravenous immunoglobulin was given to 19 (70.4%) of them. Eight (36%) patients in the plasmapheresis group and 1 (20%) patient who was not in the group died. The standard mortality rate (observed/expected mortality) of the patients who were not in the plasmapheresis group was 1.74, higher than of the ones in the plasmapheresis group (1.52). The incidence of death of the ones whose albumin levels were under 2.49 was 32.9 times higher than the ones whose levels were above 2.49 (p:0.001). There was no related mortality except for the albumin levels.

Conclusions:
Although there was no statistically meaningful difference in terms of mortality in between the two groups, the calculated standard mortality rate was higher in the patients who did not have plasmapheresis.
These results need to be supported with wider studies since the number of cases was not enough in our study. Preparing staff for redeployment to intensive care medicine (ICM) during the COVID-19 pandemic took a multi-faceted approach in our Trust, which includes 5 separate hospitals. In addition to face-to-face training, we found that online resources were positively received by staff preparing for redeployment.

Methods:
An entirely new set of educational and reference resources were rapidly collated onto an existing online platform previously used for Statutory and Mandatory Training. These resources were organized into general information for all healthcare professionals (e.g. personal protective equipment and resuscitation) and profession-specific sections aimed at preparing doctors and nurses for their new roles. Within the profession-specific sections, the material was arranged according to where the staff member would be redeployed: ICM or medical wards. Efforts were made to ensure material was as accessible as possible on any device in order to optimize the user experience. As a sample, all 131 staff who accessed the resources over 3 days (18-20 April) were contacted by email to complete feedback via SurveyMonkey.
Results: 78% of respondents found the resources to be very or extremely useful ( Figure 1) and 73% found the usability and design of the site to be very or extremely clear. 46% of respondents indicated that they were most interested in the ICM resources.

Conclusions:
The COVID-19 pandemic focused attention on optimizing the provision of resources to prepare staff for new roles. Positive feedback for the online resources used in our hospitals during the pandemic suggests that similar platforms could be utilized as a matter of course in the future. Doctors-in-training, in particular, rotate frequently between departments and access to online resources may aid their preparation for, and effectiveness in, new roles within ICM. Introduction: This study aimed to evaluate the performance of a standard machine learning (ML) technique, namely a multinomial logistic regression model with a ridge estimator, on the task of predicting bacterial resistance to antibiotics in an intensive care unit (ICU). ML techniques are increasingly applied to identify resistance factors and to assist the clinician in selecting empirical antibiotic treatment.

Methods:
We analyzed the microbiological data collected from 345 ICU patients during two years (2017 and 2018), in a public tertiary hospital in Greece. The dataset of 23,067 instances contains the attributes of gender, age, type of the sample (e.g., blood, tracheal aspirate, urine), Gram stain, antibiotics, and the class of antimicrobial susceptibility. We have used the WEKA-Data Mining Software in Java Workbench and, specifically, a multinomial logistic regression model with a ridge estimator [1] with 10-fold cross-validation.

Results:
The multinomial logistic regression model with a ridge estimator weighted average results achieved an F-measure of 0.664, a Precision value of 0.671, a Recall value of 0.667, and an area under the receiver operating characteristic curve (AUC ROC) of 0.726. The experimental results, especially the value of the AUC ROC, indicate that this model could be appropriate for ICU antimicrobial susceptibility forecasting, based on similar datasets with limited attributes, as in our case. The method's performance compares favorably with the best model, a Multilayer Perceptron (an artificial neural network), as has been reported in our previous study [2] based on the same dataset.

Conclusions:
A multinomial logistic regression learning classifier was validated on an ICU antimicrobial susceptibility dataset as a low-cost method to complement the information available to medical doctors before they dispense direct proper treatment.

Introduction:
Acinetobacter baumannii (AB) appears as the most frequent Gram negative micro-organism implicated in post neurosurgical nososcomial meningitis [1]. Intraventricular colistin (IVT) as the last therapeutic resort for the treatment of multidrug-resistant (MDR) and extensively drug-resistant (XDR) Acinetobacter baumannii meningitis. The aim of this study was to determinate the epidemiological characteristics, outcomes end prognosis of patients complicated with XDR Acinetobacter baumannii meningitis.

Methods:
Retrospectively we included patients hospitalized for post neurosurgical nosocomial AB meningitis and treated by IVT colistin (125000 IU) in addition with IV colistin (Loading dose 9 MIU followed by 4.5 MIU every 12h) for a median duration of 21 days. For each patient, the following data were recorded retrospectively: age, sex, primary diagnosis, antimicrobial regimens prescribed, days of sterilization of cerebrospinal fluid, duration of IVT colistin treatment, toxicity, outcome and follow-up.

Results:
A total of 30 patients were found to have been treated with IVT colistin for AB meningitis. The mean age of patients was 54.1 years. In all cases, meningitis was secondary to neurosurgery procedures for the management of various cerebral nervous system diseases. The median time for admission to diagnosis of meningitis was 7 days. Regarding the resistance patter, 30 cases were defined XDR strains. Successful clinical and bacteriological outcome was achieved in 83% of AB meningitis cases. Five patients died of septic shock.

Conclusions:
IVT administration of colistin appears to be a rather safe and efficacious therapeutic option for the treatment of XDR Acinetobacter baumannii meningitis and should be considered an applicable procedure in the neurosurgery setting.  research program is exploring the use of rapid molecular diagnostics to improve the treatment of HAP/VAP patients by more swiftly identifying causative pathogens and their antibiotic resistances. Resulting changes to patient management and antibiotic use potentially have substantial resource implications, quantifying these is important.

Methods:
Patients, or their representatives, from 4 UK ICUs were approached for involvement if they were either i) starting a course of antibiotics or ii) having a change of antibiotics for the treatment of HAP or VAP. We collected information to allow estimates of: cost of ICU stay [length of stay (LOS) and related health resource group (HRG)]; acquisition cost (from the British National Formulary) of antibiotics used in the 21-days after recruitment; and quality of life (EuroQoL EQ-5D-5L) in those alive at 21-days.
Results: N=143 patients were recruited. They had considerable ICU-associated LOS and hospital costs: their mean stay was 22 days and mean costs were GBP £43,100. Both LOS and costs were heavily right-skewed (most values are low but the remainder take large values). Compared with HAP, VAP caused greater LOS and ICU costs. Antibiotics themselves formed only a tiny fraction of total costs (mean 21-day cost was £321). A total of 43 people completed the EQ-5D-5L: a wide range of utilities resulted, ranging from 0.8 to -0.4 (with negatives indicating states valued worse than death).

Conclusions:
HAP, and particularly VAP, are associated with significant hospital costs. Interventions that could improve the care of individuals with HAP/VAP and reduce their LOS would significantly free up scarce ICU resources, allowing other patients to be treated. Hence, we evaluated BF performance against culture.

Methods:
From January to May 2019, 298 samples were collected consecutively from hospitalized patients with suspected pneumonia. Samples were sent routinely to the Department of Clinical Microbiology for culture and additional testing by BF. Retrospectively, patients were categorized into 'pneumonia', 'probable pneumonia' and "no pneumonia". Analytical performance was evaluated by bacterial pathogen concordance between the two methods. Clinical performance was determined regarding pneumonia/not pneumonia and detection of a positive/ negative bacterial pathogen and evaluated by sensitivity, positive predictive value (PPV), negative predictive value (NPV) and efficacy. Results: 98 patients had pneumonia, 71 had probable pneumonia and 129 had no pneumonia. Positive agreement between culture and BF was 42%. The rate increased to 67% when pathogens in lowest quantity in BF were excluded. Sensitivity of BF was improved from 73% to 89%, and for culture from 43% to 58%, when only high quality samples were included. For BF, PPV: 50%, NPV: 69% and efficacy: 57% were comparable to culture (PPV: 52%; NPV: 62%; efficacy: 58%); this increased slightly for both BF (PPV: 55%; NPV: 76%; efficacy: 61) and culture (PPV: 55%), when only high quality samples were included.

Conclusions:
PPV and NPV of both BF and culture were low and are therefore best used in patients in whom the pneumonia diagnosis has been established clinically. Indiscriminate use may be diagnostically misleading and a cause of improper use of antibiotics.

Methods:
Four focus groups and 34 interviews were conducted with clinicians in four UK ICUs. Focus groups explored perceptions of factors influencing prescribing decisions and interviews explored decision processes using clinical vignettes depicting HAP. Data were analyzed using thematic analysis.

Results:
Prescriber perceptions were key to decision-making. Most clinicians balanced societal risks of antimicrobial resistance (AMR) against individual patients' needs, with the latter generally prioritized. In uncertainty, the default was to prescribe on the basis that antibiotics might prevent patient mortality, with clinicians viewing prescribing as more defensible than withholding. Antibiotic side effects were rarely mentioned. Clinicians were aware of AMR and strove to withhold potentially unnecessary antibiotics. This aim was counterbalanced by previous negative experiences, which motivated prescribing antibiotics 'just in case' of infection. Clinicians' perceptions interacted with the prescribing context. Examples include a lower perceived threshold to prescribe antibiotics out of hours, input from non-ICU team members, and varied local prescribing norms.

Conclusions:
When making prescribing decisions, clinicians' understandable fear of undertreating possible infection often conflicts with AMS aspirations. Prescribers seem to be driven by perceived negative consequences for patients and themselves over more distal issues of AMR. Rapid evidence-based support from more effective diagnostics may help reconcile these competing priorities.

Introduction:
Although viruses are considered the commonest cause of admitted cases of lower respiratory tract infections (LRΤΙs) [1] data for patients with severe LRTI are missing. This study aimed to define the epidemiology of severe LRTI using the Biofire® FilmArray® Pneumonia plus (PNplus) Panel.

Methods:
This was a sub-study of the PROGRESS trial where procalcitonin (PCT)-guidance for early stop of antibiotics was used to prevent infection-associated adverse events in patients with sepsis (Clini-calTrials.gov NCT03333304). PNplus Panel was performed retrospectively in frozen lower respiratory samples of 90 septic patients (Sepsis-3) with LRTI. Primary endpoint was the comparison of the detection rate of pathogens between conventional microbiology (blood, sputum, pleural fluid cultures and urine antigen detection) and PNplus Panel. Secondary endpoints were the association with inflammatory host response and detection of antibiotic resistance.
Results: 56 patients with community-acquired (CAP) and 34 with healthcare-associated pneumonia (HCAP) were studied; median pneumonia severity index was 113 (88-135). PNplus detected at least one pathogen in 65 patients (72.2%) compared to 10% detected by conventional microbiology (p<0.0001); bacteria were the commonest pathogens ( Figure 1). Median PCT was 0.49 ng/ml among patients with ≥105 copies/ml of a bacterial pathogen compared to 0.18 ng/ml in detection at lower loads (p=0.004). SOFA score, serum CRP and white blood cells did not differ in patients with undetected, bacterial, viral or mixed cause of infection. Median PCT was 0.52 ng/ml among patients with bacterial pathogens compared to 0.19 ng/ml with viral (p: 0.045). At least one resistance gene was detected in 14.4% of samples, being more common in HCAP versus CAP (32.2% vs 5.1%; p: 0.001).

Conclusions:
PNplus detects severe pneumonia pathogens at significantly greater rate than conventional microbiology and higher circulating PCT levels reflect their true virulence. Augmented renal clearance (ARC) might lead to subtherapeutic plasma levels of drugs with predominant renal clearance. Early identification of ARC remains challenging for the intensive care unit (ICU) physician. We developed and validated the ARC predictor, a clinical prediction model for ARC on the next day during ICU stay, and made it available via an online calculator. Its predictive performance was compared with that of two existing models for ARC, i.e. the ARC score and the ARCTIC score.

Methods:
A large multicenter database including medical, surgical and cardiac surgery ICU patients (n = 33258 ICU days) from three Belgian tertiary care academic hospitals was used for the development of the prediction model. Development was based on clinical information available during ICU stay. We assessed performance by measuring discrimination, calibration and net benefit. The final model was externally validated (n = 10259 ICU days) in a single-center population. Results: ARC was found on 19.6% of all ICU days in the development cohort. Six clinical variables were retained in the ARC predictor: day from ICU admission, age, sex, serum creatinine, trauma and cardiac surgery. External validation confirmed good performance with an area under the curve of 0.88 (95% CI 0.87 -0.88), and a sensitivity and specificity of 84.1 (95% CI 82.5 -85.7) and 76.3 (95% CI 75.4 -77.2) at the default threshold probability of 0.2, respectively. Conclusions: ARC on the next day can be predicted with good performance during ICU stay, using routinely collected clinical information that is readily available at bedside. The ARC predictor is available at www.arcpredictor.com.

Introduction:
The NIHR-funded INHALE Programme aims to improve antimicrobial stewardship by using molecular microbiology diagnostics for HAP/ VAP in ICUs. The BioFire FilmArray Pneumonia Panel is deployed at point of care in 12 UK ICUs, with patients randomized to FilmArrayguided management or to standard empiric antimicrobials and laboratory microbiology testing. The FilmArray seeks 34 organism and gene targets, with results in 1.5 h; we designed and field-tested an algorithm to guide antimicrobial prescribing based on its findings.

Methods:
The algorithm is based on (i) the organism and resistance gene targets detected, (ii) national resistance prevalence data and (iii) the patient's allergy status. Narrow-spectrum agents are preferred, and good stewardship encouraged. Microbiologists, ICU pharmacists, and ICU clinicians were consulted and local adaptation allowed.

Results:
When single organisms are found, the algorithm favors, e.g. temocillin vs. Enterobacterales, flucloxacillin vs. MSSA and co-amoxiclav vs. H. influenzae; discontinuation is advised if no organism is found and the patient lacks evidence of infection; broader spectrum agents are favored for combinations of organisms. Among 10 adult sites, 4 adopted the algorithm unaltered and 2 with minor variation. Concerns were: unwillingness to adopt: (i) temocillin for Enterobacterales; (ii) ceftazidime vs. Pseudomonas; or (iii) cephalosporins for patients with mild β-lactam allergy. Greater variation was needed at 2 pediatric ICUs. There was debate about infection control implications of rapid ICU-based tests.

Conclusions:
The algorithm aims to ensure that rapid microbiology translates to optimized therapy. It was not possible to impose a single algorithm at all ICUs, but core elements and principles were retained. An early audit of RCT results indicates most test-arm treatments are being guided by the algorithm, illustrating the approach's potential.

Conclusions:
These new insights from daily clinical practice suggest that i.v. fosfomycin is a valuable combination partner for the treatment of IE even in cases of foreign body involvement. Here, we present first real-life experience with i.v. fosfomycin in a subgroup of patients with cSSTI from an ongoing international study.

Methods:
Prospective, non-interventional and monitored European multicenter study (FORTRESS; NCT02979951). The primary objective is clinical success, defined as clinical cure or improvement incl. microbiological cure at end of fosfomycin treatment. cSSTI was defined according to common definitions (US, EU).

Conclusions:
These interim data from clinical practice emphasize that i.v. fosfomycin is a useful combination partner for treatment of cSSTI even in life-threatening cases.

Introduction:
The SOFA score is a reliable tool to describe organ dysfunction following trauma [1]. Serum IL-6 and IL-10 have been shown to rise early following injury and potentially be useful markers in predicting clinical outcomes [2]. This study evaluated the predictive potential of early IL-6 and IL-10 levels on SOFA outcome in patients following major trauma. Methods: One hundred major trauma patients were included in this study. Inclusion criteria were 1) injury severity requiring immediate transfer from the emergency department to the operating theatre or critical care, 2) enrolment within 24 hours of the injury. Principle exclusion criteria were 1) age <18 years, 2) patients on steroids or other immunosuppressive medication. Serum samples collected day 1 and day 3 were evaluated for IL-6 and IL-10 using cytometric bead arrays. SOFA scores were calculated on day 1 and day 5, from which delta SOFA was calculated. All study procedures were approved by the National Research Ethics Committee South Manchester. Results: Day 1 IL-6 did not correlate with day 5 SOFA (r = 0.120, p = 0.21 n = 100), day 3 IL-6 correlated with day 5 SOFA (r = 0.40, p = 0.002 n = 52). The day 1 (r = 0.24, p = 0.018 n = 100,) and day 3 (r = 0.43, p = 0.0013, n = 21,) IL-10 concentrations, however, were both found to significantly correlate with patients' day 5 SOFA scores. Patients with a delta SOFA of < 0, indicating recovery, display a lower IL6:IL10 ratio on both day 1 serum (7.94 ±0.81 compared to 14.15±2.30, p = 0.03) and day 3 serum (5.11 ±0.78 compared to 10.57±1.32, p = 0.0008) ( Figure 1). Conclusions: IL-10 may be a useful biomarker to predict late organ dysfunction following major trauma. Furthermore, the ratio of IL6:IL10 on both day 1 and day 3 post injury, may be of particular value in predicting the risk of deterioration or improvement (delta SOFA) in the first 5 days after trauma.

Introduction:
The impact of mild acute kidney injury (AKI) observed in the immediate postoperative period after major surgery on long term renal function remains poorly studied. According to the "Kidney Disease: Improving Global Outcomes" (KDIGO) classification, a mild injury corresponds to a KIDIGO stage 1, characterized by an increase in creatinine of at least 0.3 mg/dl or 1.5 to 1.9 times the baseline level. We tested the hypothesis that patients who underwent moderate-to high-risk abdominal surgery and developed mild AKI in the following days would be at an increased risk of long-term renal injury compared to patients with no postoperative AKI.

Methods:
This single center retrospective cohort study analyzed all consecutive adult patients without chronic kidney disease who underwent elective moderate to high-risk abdominal surgery at Erasme Hospital between 2014 and 2019 and who had three relevant creatinine measurements: before surgery, during the first seven postoperative days, and long-term (1 year). The study population was divided into three groups according to the postoperative renal function: no renal injury, mild AKI (KIDIGO stage 1) and moderate to severe AKI (KIDIGO stages 2 and 3).

Results:
A total of 815 patients were analyzed. Overall, postoperative AKI (stage 1 to 3) incidence was 13.4% (10% with a mild AKI and 3.4% with a moderate to severe AKI). The median long-term follow-up was 360, 354 and 353 days for the three groups respectively (p=0.190). Compared to patients without postoperative AKI, those developing mild AKI had a higher risk of long-term renal injury (odds ratio [95%CI] of 3.1 [1.7-5.5]; p=0.0001). This effect is even stronger when accounting for other perioperative covariates (adjusted odds ratio [95%CI] of 4.5 [1.8-11.4]; p=0.0001).

Conclusions:
Patients undergoing moderate to high-risk abdominal surgery who develop a mild postoperative AKI more than tripled the odds of having renal injury one year after surgery compared with patients who did not develop any postoperative AKI.

P570
Prediction of cardiac surgery associated-acute kidney injury (CSA-AKI) by healthcare providers, AKI predicting score systems and cell cycle arrest biomarkers. An interim analysis of PREDICTAKI, a prospective observational trial. L Van Laethem, W Vandenberghe, E Hoste UZ Gent, ICU, Gent, Belgium Critical Care 2020, 24(Suppl 2):P570

Introductions:
Cardiac surgery associated acute kidney injury (CSA-AKI) is a frequent complication and associated with increased morbidity and mortality. The aim of this study is to investigate whether healthcare providers (HCP) can predict CSA-AKI. The adequacy of prediction will be compared with 4 existing AKI prediction score systems (AKIpredictor score, Ng score, Crate score and Murphy score) and cell cycle arrest biomarkers (NephroCheck®).

Methods:
This is an interim analysis of a single center prospective study. Patients who underwent elective cardiac surgery with cardiopulmonary bypass (CPB) and without end stage kidney disease or AKI were considered for enrolment. Four hours after CPB was ceased, 3 types of HCP (ICU specialist, ICU nurse, nephrologist) were asked to predict occurrence of AKI stage 2-3, defined according to the KDIGO definition and staging system, and the need for renal replacement therapy (RRT) within 48 hours after admission on ICU, using a structured questionnaire. At the same time we measured cell cycle arrest biomarkers with the Nephrocheck® test. After analysis of the Nephrocheck® the result was communicated to the HCP and the same questionnaire was completed again. The adequacy of prediction by HCP was compared with the Nephrocheck® test results and 4 AKI predicting scoring systems by calculating the area under the ROC curve (AUC).

Results:
Fifty patients were included. AKI stage 2-3 occurred in 19 patients, no patient was treated with RRT. AUC was highest for the Ng, AKIpredictor score, and ICU specialist. We found that the Crate score, the NephroCheck® and nephrologist had poor performance in prediction of CSA-AKI stage 2-3 (Table 1). Providing the NephroCheck® test results did not improve the ability of AKI prediction. Conclusions: ICU specialists, the Ng and AKIpredictor scores had good performance for prediction of CSA-AKI stage 2-3 occurrence within 48-h. Providing info of the NephroCheck® test did not improve this predictive performance.

P571
Critically ill patients with H1N1 pneumonia: two year experience in a tertiary level Indian ICU P Sathe, AK Parathody, R Borse Department of Critical Care, Ruby Hall Clinic, Pune, India Critical Care 2020, 24(Suppl 2):P571

Introduction:
Pandemic H1N1 viral infection is an important issue in India as in the world over. Maharashtra in particular bears a significant brunt of illness. We present the experience of treating H1N1 cases over the period of two calendar years in a 34 bed multi-disciplinary ICU in a tertiary level health care facility in Pune, Maharashtra, India. Our objective was to learn about the clinical profile, outcome and quality of life and factors influencing these, in critically ill patients with H1N1 pneumonia.

Methods:
Retrospective analysis of case files and phone interview of 88 patients with confirmed H1N1 pneumonia.

Introduction:
The acute respiratory distress syndrome (ARDS) can be aggravated by high cycling rates and airway pressures, leading to the application of mechanical power to the respiratory system. Previous estimates [1] of power include energy dissipated into the chest wall, which may not be causal to lung injury, and also include static work from positive end-expiratory pressure (PEEP). We propose and compare three additional estimates of mechanical power based on data from the EPVent study [2].

Methods:
Mechanical power was calculated as product of respiratory rate and inspiratory work (the sum of inspiratory pressure*Δvolume products): (1) respiratory system-directed power, based on airway pressure (Pao), including static work from PEEP; (2) respiratory system-directed driving power, based on Pao-PEEP, excluding static work; (3) lungdirected power, based on transpulmonary pressure (PL) and (4) lungdirected driving power, based on ΔPL from end-expiration. We hypothesized that these estimates differ at baseline and that they are affected by PEEP titration to achieve positive PL in the EPVent study.  (Table 1). PL-guided PEEP titration in the EPVent intervention group differentially affected the four estimates of power (Table 1): respiratory system-directed and lung-directed power increased, while respiratory system-directed driving power and lungdirected driving power did not change (p for interaction 0.008).

Conclusions:
Proposed estimates of mechanical power differ based on whether static work from PEEP and chest-wall directed energy are included and they are differentially affected by changes in PL and PEEP.

Introduction:
Bedside assessment of lung recruitability is essential to screen ARDS patients.to improve lung function and to guide the treatment. A high recruitability index, for instance, would favor the balance towards a recruitment maneuver, followed by individualized PEEP. The purpose of this study was to measure the gain of respiratory system compliance after a brief and "light" maneuver of PEEP challenge. The goal of the maneuver was to predict a meaningful reduction in DP.

Methods:
The "light" recruitability maneuver was preformed through a tool developed by Nihon Koden Mechanical Ventilator (NKV550, California, EUA). The applicative program (Recruitability-Assessment-APP) available at the mechanical ventilator performs the recruitability maneuver automatically. The APP provides gain of compliance percentage and shows the difference between pulmonary compliance at the same PEEPs. This study was realized using PEEP at 5 cmH 2 O and Ideal PEEP according of PEEP titration maneuver performed by electrical impedance tomography (EIT) and CT. Results: 5 animals (38.7±2.8 kg) with severe lung injury (PaO 2 /FiO 2 <100 mmHg) were studied. By applying different lung-histories and PEEPs,

Introduction:
Critical illness may result in muscle weakness and decreased cardiopulmonary fitness. To avoid this, mobilization is started in the early stage of intensive care unit admission. Cardiopulmonary exercise testing can be used for diagnostic purposes. An accepted method of interpreting exercise data is by presenting them as 9-panel plots, as proposed by Wasserman [1], and analyzing them systematically. We set out to explore whether gas exchange data obtained during exercise in intubated, critically ill patients could be analyzed as standard 9-panel plots.

Methods:
Mechanically ventilated patients recovering from critical illness were subjected to an incremental exercise protocol using a bedside cycle ergometer (MOTOmed). Respiratory gases were analyzed with a Cosmed Quark, applied to an endotracheal tube. Blood gases were sampled from arterial catheters. Data were analyzed using MS Excel.

Results:
Exercise data of seven patients were analyzed individually, providing insight in their respiratory physiology. The cumulative data are shown in Figure 1. Basal metabolic rate was increased in 6/7 patients. Median oxygen uptake (VO 2 ) increased from 408 to 489 ml/min during unloaded cycling, corresponding to an intrinsic workload of 10 W. Median maximum extrinsic workload during loaded cycling was 7 W, resulting in a median peak VO 2 of 34.3 % of predicted VO 2 max. This was accompanied by an increase in CO 2 production, respiratory minute volume and heart rate. Three patients passed the anaerobic threshold. Median fractional dead space at rest was 44%, decreasing to 42% during exercise, accompanied by a similar decrease in arterial-end tidal PCO 2 difference, and improved respiratory efficiency for O 2 and CO 2 . Median arterio-alveolar oxygen difference was 9.9 kPa, decreasing to 7.2 during exercise.

Conclusions:
This research demonstrated the feasibility of standard 9-panel plots to present and analyze exercise data in mechanically ventilated, critically ill patients.

Introduction:
Respiratory management of patients with moderate-severe ARDS remains challenging, regardless of cause. We must individualize management with each patient, and for this it is essential to have adequate respiratory monitoring to support decision-making. Sapsford and Jones described the SpO 2 -FiO 2 Diagram, a useful tool for bedside, which allows estimating the shunt and areas with low V / Q due to the closure of the small airway [1,2].

Methods:
Male patient, 37 years old, with severe ARSD due to H1N1 influenza, requiring intubation in the first 24 hours of admission to the ICU. Protective ventilation is started, with a tidal volume of 7 cc/kg ideal weight, with RR to keep pCO 2 in the normal range. A SpO 2 -FiO 2 Diagram is made after connection to MV, after performing lung recruitment maneuvers and after recruitment with prone position. For this, FiO 2 is increased to 100% and drops of 10% are made every 2 minutes, evaluating the pulse oximetry value.

Results:
Protective ventilation is carried out, maintaining plateau pressure <30 cmh 2 o, driving pressure <15 at all times. AutoPEEP <2 cmH 2 O. After recruitment maneuvers, PEEP is titrated in 13 cmH 2 O for better static compliance. Initially, it requires 100% FiO 2 , which can be reduced to 60% in both RM and Prone Position. The initial PaO 2 is 65mmHg, with a subsequent increase to 81 mmHg (post-MR) and 168 mmHg (after 2 hours of prone position). The increase in PaO 2 / FiO 2 is as follows: 65 -135 -281. Figure 1 shows the SpO 2 -FiO 2 diagram with the results obtained.

Conclusions:
Respiratory monitoring is key in the management of patients with severe ARDS, the SpO 2 -FiO 2 diagram is easy to perform and at bedside, providing useful information. Consent to publish: written informed consent for publication was obtained from the patient.

P580
Mortality rates in mild versus severe abbreviated injury score (AIS) graded burns inhalation injury: a systematic review WN Charles 1 , A Dutt 2 , D Collins 3 , S Singh 4

Introduction:
The Abbreviated Injury Score (AIS) grading system is the most widely used bronchoscopic classification for assessing the severity of burns inhalation injury. However, there is a lack of clarity regarding the effect of AIS-graded inhalation injury on mortality. This systematic review evaluated whether more severe injury grades were associated with increased mortality. Methods: OVID MEDLINE, EMBASE and CENTRAL were searched from inception to 5th April 2020. Clinical studies utilizing the AIS system to grade inhalation injury were deemed eligible if mortality data were reported by AIS grade. For comparison, AIS grades 0, 1 and 2 constituted a low-grade (milder) injury and AIS grades 3 and 4 constituted a highgrade (severe) injury. The level of evidence of each study was assessed as per the Oxford Centre for Evidence-Based Medicine guidelines. This systematic review adhered to the PRISMA statement.

Results:
The search identified 177 papers, of which 13 underwent full-text review. Six single-center, comparative studies (1 prospective and 5 retrospective designs) were included for descriptive analysis. The total number of patients was 715. Inter-grade differences in age and total body surface area burned were non-significant in five studies. Two studies demonstrated a statistically significant increase in mortality for more severe grades, with a further two demonstrating a non-significant trend (p ≤ 0.10). Five studies facilitated the comparison of low-grade (n=513) versus high-grade (n=122) inhalation injury. Mortality rates ranged from 13% (range 4-26%) in low-grade injuries to 27% (range 15-32%) in high-grade injuries. The use of AIS was inconsistent between studies, with all papers performing some further stratification to minimize clinician bias or for statistical purposes. The median level of evidence of included studies was 3.

Conclusions:
Mortality rates were higher in more severe AIS-graded burns inhalation injury. Refinement of the AIS is recommended to enable standardized use. Introduction: Mechanical insufflation-exsufflation (MI-E) possibly improves cough efficacy but there are limited data in terms of cough assist techniques, modes of device, safety and efficacy of MI-E in critically ill patients. So we want to compare cough efficacy (secretion clearance) and cough strength (maximum expiratory pressure; MEP), and safety between implementation of volume mode and pressure mode of MI-E in postextubated patients.

Methods:
We conducted a prospective crossover study. The postextubated patients with history of MV for ≥24 hours were enrolled. We excluded patients with severe COPD, lung bleb, pneumothorax, hemoptysis, recent history of thoracoabdominal surgery, increased intracranial pressure, impaired consciousness and pregnancy. The patients were randomized into 2 groups. The patients received cough assistance with either volume mode (group A) or pressure mode of MI-E (group B) of MI-E during the first period. The sputum volume, respiratory and hemodynamic parameters, and adverse events were recorded. One day after the first period, the patients received cough assistance with the other mode during the second period. The primary outcome was the amount of secretion after MI-E implementation and the secondary outcomes were MEP changes and adverse events.

Results:
Totally, 25 patients were enrolled. The majority of them were female, medical ICU patients with the mean age of 64 years. The amount of secretion after MI-E implementation was not significantly different between both modes. In addition, MEP, respiratory events and hemodynamic instability were also not significantly different between both modes. Remarkably, regardless of modes of MI-E, MEP tended to increase after multiple sessions of MI-E. Two patients were reintubated due to volume overload and vocal cord edema.

Conclusions:
In postextubated patients, cough efficacy after MI-E implementation was not different between volume and pressure modes. Additionally, MI-E was safe and tended to improve cough strength after extubation.

P582
Knockdown of bone morphogenetic protein type II receptor leads to decreased aquaporin 1 expression and function in human pulmonary microvascular endothelial cells Introduction: The aim of our retrospective analysis is to evaluate the results of conservative management of acquired tracheoesophageal fistulae (TEF). TEF are rare but potentially life-threatening emergencies which can be of either spontaneous or iatrogenic origin. Spontaneous ones can be congenital or secondary to malignancy. For acquired ones numerous causes have been documented, the most common of which are endotracheal and tracheostomy tube-related injuries.

Methods:
From February 2017 to March 2020 seven patients (5 men; 2 women) with acquired TEF were diagnosed in our intensive care unit (ICU). The injury occurred after dilational percutaneous tracheostomy in 3 patients, after esophageal endoscopy in 1 patient, after cuff-related ruptures in the 3 intubated and mechanically ventilated patients. Our patients had no particular medical history. Mean age: 46 years. Mean duration of signs before diagnosis: 8 hours. All our patients had a Level III B lesion. The median length of the injury was 1.4 cm. The mean duration of hospitalization in the ICU was 31 days.

Results:
All patients underwent conservative management: antibiotic therapy, close bronchoscopic controls, percutaneous endoscopic gastrostomy and tomographic investigation. No mediastinitis was observed. Two patients died from causes unrelated to the tracheal injury.

Conclusions:
Successful management of acquired TEF requires a fast and straightforward diagnostic evaluation. According to our experience conservative management of TEF may be a save option in patients with uncomplicated ventilation and moderate and nonprogressive emphysema. Consent to publish: Written informed consent for the publication of these details was obtained from the participants. Introduction: Current ICU sedation guidelines recommend strategies using nonbenzodiazepine sedatives. Since the development of halogenated anesthetic reflectors, inhaled ICU sedation has become increasingly popular. This survey was undertaken to explore the use of volatile agents for ICU sedation in France.

Methods:
Adult ICUs of the Société Française d'Anesthésie et Réanimation (SFAR) database were contacted by phone or email between July and August 2019. Heads of ICU were questioned about the characteristics of their department, their knowledge on inhaled sedation, and practical aspects of inhaled sedation use in their department.

Results:
Among the 374 ICUs contacted, 187 provided responses (50%). Most ICU directors (73%) knew about the use of inhaled ICU sedation and 21% used inhaled sedation in their unit, mostly with the Anesthetic Conserving Device (AnaConDa, Sedana Medical). Most intensivists had used volatile agents for sedation for <5 years (63%) and in <20 patients per year (75%), with their main indications being: failure of intravenous sedation, severe asthma, and acute respiratory distress syndrome. Sevoflurane and isoflurane were mainly used (88% and 20%, respectively). The main reasons for not using inhaled ICU sedation were: "device not available" (40%), "lack of medical interest" (37%), "lack of familiarity or knowledge about the technique" (35%) and "elevated cost" (21%). Most respondents (80%) were overall satisfied with the use of inhaled sedation (Figure 1). Almost 75% stated that inhaled sedation was a seducing alternative to intravenous sedation.

Conclusions:
This survey highlights the widespread knowledge about inhaled ICU sedation in France but shows its limited use to date. Differences in education and knowledge, as well as the recent and relatively scarce literature on the use of volatile agents in the ICU, might explain the diverse practices that were observed. The low rate of mild adverse effects and the users' satisfaction are promising for this potentially important tool for ICU sedation. Fig. 1 (abstract P583). Tidal volume. Ventral: ventral regional tidal volume; Dorsal: dorsal regional tidal volume; S1-4: Supine Registration Point 1-4; AU: Arbitrary Unit; Filled boxes: Prone group; Open circles: Supine group