Predicting hospital emergency department visits accurately: A systematic review.

OBJECTIVES: The emergency department (ED) is a very important healthcare entrance point, known for its challenging organisation and management due to demand unpredictability. An accurate forecast system of ED visits is crucial to the implementation of better management strategies that optimise resources utilization, reduce costs and improve public confidence. The aim of this review is to investigate the different factors that affect the ED visits forecasting outcomes, in particular the predictive variables and type of models applied. METHODS: A systematic search was conducted in PubMed, Web of Science and Scopus. The review methodology followed the PRISMA statement guidelines. RESULTS: Seven studies were selected, all exploring predictive models to forecast ED daily visits for general care. MAPE and RMAE were used to measure models' accuracy. All models displayed good accuracy, with errors below 10%. CONCLUSIONS: Model selection and accuracy was found to be particularly sensitive to the ED dimension. While ARIMA-based and other linear models have good performance for short-time forecast, some machine learning methods proved to be more stable when forecasting multiple horizons. The inclusion of exogenous variables was found to be advantageous only in bigger EDs.

Antibiotic use in patients with COVID-19: a 'snapshot' Infectious Diseases International Research Initiative (ID-IRI) survey.

BACKGROUND: Antibiotics may be indicated in patients with COVID-19 due to suspected or confirmed bacterial superinfection. OBJECTIVES: To investigate antibiotic prescribing practices in patients with COVID-19. METHODS: We performed an international web-based survey and investigated the pattern of antibiotic use as reported by physicians involved in treatment of COVID-19. SPSS Statistics version 25 was used for data analysis. RESULTS: The survey was completed by 166 participants from 23 countries and 82 different hospitals. Local guidelines for antibiotic use in COVID-19 patients were reported by 61.8% (n = 102) of participants and for 82.9% (n = 136) they did not differ from local community-acquired pneumonia guidelines. Clinical presentation was recognized as the most important reason for the start of antibiotics (mean score = 4.07 and SD = 1.095 on grading scale from 1 to 5). When antibiotics were started, most respondents rated as the highest the need for coverage of atypical pathogens (mean score = 2.8 and SD = 0.99), followed by Staphylococcus aureus (mean score = 2.67 and SD = 1.05 on bi-modal scale, with values 1 and 2 for disagreement and values 3 and 4 for agreement). In the patients on the ward, 29.1% of respondents chose not to prescribe any antibiotic. Combination of ß-lactams and macrolides or fluoroquinolones was reported by 52.4% (n = 87) of respondents. In patients in the ICU, piperacillin/tazobactam was the most commonly prescribed antibiotic. The mean reported duration of antibiotic treatment was 7.12 (SD = 2.44) days. CONCLUSIONS: The study revealed widespread broad-spectrum antibiotic use in patients with COVID-19. Implementation of antimicrobial stewardship principles is warranted to mitigate the negative consequences of antibiotic therapy.

Levels of evidence supporting European and American community-acquired pneumonia guidelines.

Optimal clinical decisions should be supported by clinical practice guidelines (CPG) based on evidence generated from randomized clinical trials (RCT). We aimed to evaluate the class and level of evidence (LOE) supporting the international community-acquired pneumonia (CAP) guidelines and their variation over time. The 2019 Infectious Diseases Society of America/American Thoracic Society (IDSA/ATS) and the 2011 European Respiratory Society/European Society Clinical Microbiology Infectious Diseases (ERS/ESCMID) CPG and its immediate predecessors (2007 and 2005) were evaluated. The number of recommendations and distribution as LOE A (supported by multiple RCT or a single, large RCT), B (supported by data from a single RCT or observational studies) and C (expert opinion, case studies, or standard of care) was identified. Overall, recommendations for diagnosis, management, and prevention were graded as strong in 51.4%, 62.9%, and 23.5% in spite that they were supported by LOE A in 5.7%, 11.1%, and 52.9%, respectively. In the 2019 ATS/IDSA guidelines (39 recommendations), 7.7% (n = 3) recommendations were classified as LOE A, 30.8% (n = 12) as LOE B, and 61.5% (24%) as LOE C. Across the 2011 ERS/ESCMID guidelines (68 recommendations), 21.2% (n = 14) recommendations were classified as LOE A, 4.6% (n = 3) as LOE B, and 74.2% (n = 49) as LOE C. When comparing with prior versions, the proportion of recommendations that were LOE A did not significantly increase in ERS/ESCMID (21.2% vs 20%) and decreased in ATS/IDSA (7.7% vs 32.0%). In conclusion, large randomized trials or network meta-analysis including comparison of regimens to identify high probability of best cure and mortality is an unmet clinical need on CAP.

Levels of evidence supporting clinical practice guidelines on invasive aspergillosis.

Invasive aspergillosis (IA) is a severe life-threatening infection with challenges in therapy. The aim was to evaluate the level of evidence (LOE) supporting recommendations in clinical practice guidelines (CPGs) of IA and changes over time. Search on CPG on IA released between 2000 and 2019 was done. Last versions were evaluated and compared with previous versions. Recommendations were classified by LOE as A (multiple randomized controlled trial (RCT) or meta-analysis), B (data from a single RCT or observational studies), or C (observational studies with limitations, case series, or expert opinion). Diagnosis recommendations were excluded. Five CPG from three groups of scientific societies were identified: the 2016 Infectious Diseases Society of America/American Thoracic Society (IDSA/ATS), 2017 European Society of Clinical Microbiology Infectious Diseases/European Confederation of Medical Mycology/European Respiratory Society (ESCMID/ECMM/ERS), 2018 Spanish Society of Infectious Diseases and Clinical Microbiology (SEIMC) CPGs, and their previous versions (2008 IDSA/ATS and 2011 GEMICOMED/SEIMC). ECMID/ECMM/ERS have not published any previous version. From 511 recommendations analyzed, 80 were classified as LOE A (15.7%), 223 LOE B (43.6%), and 208 LOE C (40.7%). Among 238 strong recommendations, only 57 (24.0%) were supported by LOE A. When comparing recent CPGs with previous versions, the proportion of recommendations supported by LOE A did not significantly increase over time (IDSA/ATS: 13.3% [2016] vs. 14.8% [2008], p = 0.798; and SEIMC: 22.6% [2018] vs. 19% [2011], p = 0.568). In conclusion, IA is a condition with an urgent unmet clinical need for more high-quality randomized trials.

Evidence supporting recommendations from international guidelines on treatment, diagnosis, and prevention of HAP and VAP in adults.

Clinical practice guidelines (CPGs) are intended to support clinical decisions and should be based on high-quality evidence. The objective of the study was to evaluate the quality of evidence supporting the recommendations issued in CPGs for therapy, diagnosis, and prevention of hospital-acquired and ventilator-associated pneumonia (HAP/VAP). CPGs released by international scientific societies after year 2000, using the Grades of Recommendation, Assessment, Development, and Evaluation (GRADE) methodology, were analyzed. Number and strength of recommendations and quality of evidence (high, moderate, low, and very low) were extracted and indexed in the aforementioned sections. High-quality evidence was based on randomized control trials (RCT) without important limitations and exceptionally on rigorous observational studies. Eighty recommendations were assessed, with 7 (8.7%), 24 (30.0%), 29 (36.3%), and 20 (25.0%) being supported by high, moderate, low, and very low-quality evidence, respectively. Highest evidence degree was reported for 26 prevention recommendations, with 7 (26.9%) supported by high-quality evidence and no recommendation based on very low-quality evidence. In contrast, among 9 recommendations for diagnosis and 45 for therapy, none was supported by high-quality evidence, in spite of being recommended as strong in 33.3% and 46.7%, respectively. Among HAP/VAP diagnosis recommendations, the majority of evidence was rated as low or very low-quality (55.6% and 22.2%, respectively) whereas among HAP/VAP therapy recommendations, 4/5 were rated as low and very low-quality (40% each). In conclusion, among HAP/VAP international guidelines, most recommendations, particularly in therapy, remain supported by observational studies, case reports, and expert opinion. Well-designed RCTs are urgently needed.

Ventilator-associated pneumonia diagnosis: a prioritization exercise based on multi-criteria decision analysis.

The aim was to provide global experts ranking on priorities in diagnostic tools for VAP in clinical practice. A multiple criteria decision analysis (MCDA) was performed to identify diagnosis tools for VAP diagnosis. Priority factors were identified after literature review. An international, multidisciplinary expert panel reviewed variables and ranked diagnostic tools. Experts from ten European hospitals participated. Regarding bedside clinical practices, seven required chest X-ray use in all patients, whereas six reported the use of blood cultures and endotracheal aspirate in all patients. Invasive techniques were routinely performed in seven sites. CRP, PCT, and Gram stains were performed in all patients by 5, 2, and 8, respectively. Impact on patient outcomes, safety, and impact on the decision to start antibiotic therapy were ranked as the top three relevant concerns (7.7/10, 7/10, and 6.9/10, respectively). Chest X-ray was ranked as the most important imaging technique to diagnose VAP (score 251.7). Apart from blood cultures, endotracheal aspirate culture was identified as the main collection method for the microbiological testing (scores of 274.8 and 246.8, respectively). Mini-BAL was the preferred invasive technique with a score of 208. Top three biomarkers were CRP (score 184.3), PCT (181.3), and WBC (166.4). Gram stain (192.5) was prioritized among laboratory diagnostic techniques. Using MCDA, it is recommended to perform a combination of diagnostic techniques including images (chest X-ray), culture of clinical specimens (blood cultures and endotracheal aspirate), and biomarkers (CRP or PCT) for VAP diagnosis at the bedside. Gram stain was ranked as the preferred laboratory technique.

Outcome and Management of Refractory Respiratory Failure With Timely Extracorporeal Membrane Oxygenation: Single-Center Experience With Legionella Pneumonia.

OBJECTIVE:: To analyze the management and outcome of patients with refractory respiratory failure complicating severe Legionella pneumonia rescued with extracorporeal membrane oxygenation (ECMO) in our Center. DESIGN AND SETTING:: Observational study of patients with refractory respiratory failure treated with ECMO in Hospital S.João (Porto, Portugal), between November 2009 and September 2016. PARTICIPANTS:: A total of 112 patients rescued with ECMO, of which 14 had Legionella pneumonia. RESULTS:: Patients with Legionella pneumonia were slightly older than patients with acute respiratory failure of other etiologies (51 [48-56] vs 45 [35-54]), but with no significant differences in acute respiratory failure severity between groups: Pao2/Fio2 ratio 67 (60-75) versus 69 (55-85) and Respiratory Extracorporeal Membrane Oxygenation Survival Prediction score 4 (1-5) versus 2 (-1-4), respectively. Legionella pneumonia was associated with earlier ECMO initiation (days of invasive mechanical ventilation [IMV] before ECMO: 2.0 [1.0-4.0] vs 5.0 [2.0-9.5]). After IMV adjustment to "lung rest" settings, this group presented higher respiratory system (RS) static compliance (28.7 [18.8-37.4] vs 16.0 [10.0-20.8] mL/cmH2O) but required higher ECMO support (blood flow 5.0 [4.3-5.4] vs 4.2 [3.6-4.8]). Patients with Legionella pneumonia had shorter IMV (16 [14-23] vs 27 [20-42] days) and lower incidence of intensive care unit nosocomial infections (35.7% vs 64.3%), with a trend to higher hospital survival (85.7% vs 62.2%; P = .13). CONCLUSION:: In Legionella pneumonia complicated by refractory respiratory failure, ECMO support allowed patient stabilization under lung protective ventilation and high survival rates. Timely ECMO referral should be considered for Legionella pneumonia failing conventional treatment.

Is there a C-reactive protein value beyond which one should consider infection as the cause of acute heart failure?

BACKGROUND: Heart Failure (HF) is a low grade inflammatory condition. High sensitivity C-reactive protein (hsCRP) is an established marker of inflammation. A cut-off value of hsCRP beyond which an infection should be sought has never been studied in HF. We aimed to determine the best hsCRP cut-off for infection prediction in acute HF. METHODS: We analyzed patients included in an acute HF registry - EDIFICA (Estratificação de Doentes com InsuFIciência Cardíaca Aguda). Admission hsCRP measurement was available as part of the registry's protocol. Patients with acute coronary syndrome as the cause of acute HF were excluded from the registry. Infection was considered according to the diagnosis registered in the discharge record. A receiver-operating characteristic (ROC) curve was used to determine the best hsCRP cut-off for infection prediction. RESULTS: We studied 615 patients. Mean age was 76 years, 45.2% were male, 60.3% had systolic dysfunction. Median admission hsCRP was 20.3 (9.5-55.5)mg/L; in 41.6% the cause of decompensation was an infection. The area under the ROC curve for admission hsCRP in the prediction of infection was 0.79 (0.76-0.83); the best hsCRP cut-off was 25 mg/L with a sensitivity of 72.7%, specificity 77.2%, positive predictive value 69.4% and negative predictive value 79.9%. Age and elevated hsCRP independently associated with an infection as the precipitant of acute HF. CONCLUSIONS: We suggest 25 mg/L as a cut-off beyond which an infection should be sought underlying acute HF. Almost 80% of the patients with hsCRP< 25 mg/L are not infected and 69.4% of those with higher hsCRP have a concomitant infection.

Dynamics of growth differentiation factor 15 in acute heart failure.

AIMS: Risk stratification in acute heart failure (HF) patients can help to decide therapies and time for discharge. The potential of growth differentiation factor 15 (GDF-15) in HF has been previously shown. We aimed to study the importance of GDF-15-level variations in acute HF patients. METHODS AND RESULTS: We retrospectively evaluated a cohort of patients hospitalized due to acute HF. GDF-15 was measured both at admission and on the discharge day. Patients were followed-up during a 3 year period. The endpoint under analysis was all-cause mortality. GDF-15 variation is equal to [(admission GDF-15 - discharge GDF-15)∕admission GDF-15] × 100. Variation was categorized in levels of increase or decrease of GDF-15. Patients were cross-classified according to admission and discharge GDF-15 cut-off points. A Cox regression analysis was used to assess the prognostic impact of GDF-15 variation and the impact of both admission and discharge GDF-15 according to the cross-classification. We studied a group of 249 patients with high co-morbidity burden. Eighty-one patients died at 1 year and 147 within 3 years. There was a modest decrease in GDF-15 during hospitalization from a median value of 4087 to 3671 ng/mL (P = 0.02). No association existed between GDF-15 variation and mortality. In multivariate analysis, patients with admission GDF-15 ≥ 3500 ng/mL and discharge GDF-15 ≥ 3000 ng/mL had a significantly higher 1 year death risk when compared with the remaining-hazard ratio = 2.59 (95% confidence interval: 1.41-4.76)-and a 3 year 1.76 (95% confidence interval: 1.08-2.87) higher death risk compared with those with both values below the cut-off. CONCLUSIONS: Growth differentiation factor 15 decreased during an acute HF hospitalization, but its variation had no prognostic implications. The knowledge of both admission and discharge GDF-15 added meaningful information to patients' risk stratification.

Anxiety among front-line health-care workers supporting patients with COVID-19: A global survey.

OBJECTIVE: We aimed to explore anxiety status across a broad range of HCWs supporting patients with COVID-19 in different global regions. METHOD: This was an international online survey in which participation was on voluntary basis and data were submitted via Google Drive, across a two-week period starting from March 18, 2020. The Beck Anxiety Inventory was used to quantify the level of anxiety. RESULTS: 1416 HCWs (70.8% medical doctors, 26.2% nurses) responded to the survey from 75 countries. The distribution of anxiety levels was: normal/minimal (n = 503, 35.5%), low (n = 390, 27.5%); moderate (n = 287, 20.3%), and severe (n = 236, 16.7%). According to multiple generalized linear model, female gender (p = 0.001), occupation (ie, being a nurse dealing directly with patients with COVID-19 [p = 0.017]), being younger (p = 0.001), reporting inadequate knowledge on COVID-19 (p = 0.005), having insufficient personal protective equipment (p = 0.001) and poor access to hand sanitizers or liquid soaps (p = 0.008), coexisting chronic disorders (p = 0.001) and existing mental health problems (p = 0.001), and higher income of countries where HCWs lived (p = 0.048) were significantly associated with increased anxiety. CONCLUSIONS: Front-line HCWs, regardless of the levels of COVID-19 transmission in their country, are anxious when they do not feel protected. Our findings suggest that anxiety could be mitigated ensuring sufficient levels of protective personal equipment alongside greater education and information.

Burden of Community-Acquired Pneumonia and Unmet Clinical Needs.

Community-acquired pneumonia (CAP) is the leading cause of death among infectious diseases and an important health problem, having considerable implications for healthcare systems worldwide. Despite important advances in prevention through vaccines, new rapid diagnostic tests and antibiotics, CAP management still has significant drawbacks. Mortality remains very high in severely ill patients presenting with respiratory failure or shock but is also high in the elderly. Even after a CAP episode, higher risk of death remains during a long period, a risk mainly driven by inflammation and patient-related co-morbidities. CAP microbiology has been altered by new molecular diagnostic tests that have turned viruses into the most identified pathogens, notwithstanding uncertainties about the specific role of each virus in CAP pathogenesis. Pneumococcal vaccines also impacted CAP etiology and thus had changed Streptococcus pneumoniae circulating serotypes. Pathogens from specific regions should also be kept in mind when treating CAP. New antibiotics for CAP treatment were not tested in severely ill patients and focused on multidrug-resistant pathogens that are unrelated to CAP, limiting their general use and indications for intensive care unit (ICU) patients. Similarly, CAP management could be personalized through the use of adjunctive therapies that showed outcome improvements in particular patient groups. Although pneumococcal vaccination was only convincingly shown to reduce invasive pneumococcal disease, with a less significant effect in pneumococcal CAP, it remains the best therapeutic intervention to prevent bacterial CAP. Further research in CAP is needed to reduce its population impact and improve individual outcomes.

Quality of evidence supporting Surviving Sepsis Campaign Recommendations.

INTRODUCTION: The Surviving Sepsis Campaign (SSC) guidelines, released in 2017, are a combination of expert opinion and evidence-based medicine, adopted by many institutions as a standard of practice. The aim was to analyse the quality of evidence supporting recommendations on the management of sepsis. METHODS: The strength and quality of evidence (high, moderate, low-very low and best practice statements) of each recommendation were extracted. Randomised controlled trials were required to qualify as high-quality evidence. RESULTS: A total of 96 recommendations were formulated, and 87 were included. Among thirty-one (43%) strong recommendations, only 15.2% were supported by high-quality evidence. Overall, thirty-seven (42.5%) recommendations were based on low-quality evidence, followed by 28 (32.2%) based on moderate-quality, 15 (17.2%) were best practice statements and only seven (8.0%) were supported by high-quality evidence. Randomised controlled trials supported 21.4%, 9.5% and 8.6% recommendations on mechanical ventilation, resuscitation, and management/adjuvant therapy, respectively. In contrast, none high-quality evidence recommendation supported antimicrobial/source control (82.4% were low-very low evidence or best practice statements), and nutrition. CONCLUSIONS: In the SSC guidelines most recommendations were informed by indirect evidence and non-systematic observations. While awaiting trials results, Delphi-like approaches or multi-criteria decision analyses should guide recommendations.

C-reactive protein decrease associates with mortality reduction only in heart failure with preserved ejection fraction.

AIMS: The prognostic role of high-sensitivity C-reactive protein (hsCRP) in acute heart failure is less well established than for chronic heart failure and the impact of its variation is unknown. We studied the impact of hsCRP variation in acute heart failure and whether it differed according to left ventricular function. METHODS: We analyzed patients prospectively included in an acute heart failure registry. Admission and discharge hsCRP were evaluated as part of the registry's protocol and its relative variation (ΔhsCRP) was assessed. ΔhsCRP during hospitalization =  [(admission hsCRP - discharge hsCRP)/admission hsCRP] × 100. Endpoint: all-cause death; follow-up: 3 years. A multivariate Cox-regression model was used to assess the prognostic value of ΔhsCRP (continuous and categorical variable: cut-off 40% decrease); analysis was stratified according to ventricular function. RESULTS: We studied 439 patients: mean age 75 years, 50.1% men and 69.2% had heart failure with reduced ejection fraction (HFrEF). Median discharge hsCRP was 12.4 mg/l and median ΔhsCRP was ∼40%. During follow-up 247 patients (56.3%) died: 73 (54.1%) heart failure with preserved ejection fraction (HFpEF) patients and 174 (57.2%) HFrEF patients. The multivariate-adjusted hazard ratio of 3-year mortality in HFpEF patients with hsCRP decrease of at least 40% during hospitalization was 0.56 (95% CI 0.32-0.99). A decrease of at least 40% in hsCRP was not mortality-associated in HFrEF patients. There was interaction between ΔhsCRP and left ventricular ejection fraction. CONCLUSION: A decrease of at least 40% in hsCRP in acute heart failure was associated with a 44% decrease in 3-year death risk in HFpEF patients. No association between ΔhsCRP and prognosis existed in HFrEF patients. Inflammation appears to play a different role according to left ventricular function.

Towards a multi-marker prognostic strategy in acute heart failure: a role for GDF-15.

AIMS: Growth differentiation factor (GDF)-15 mirrors inflammation and oxidative stress in cardiovascular diseases. Brain natriuretic peptide (BNP) is associated with cardiomyocyte stretch in heart failure (HF). The objective of this study was to evaluate the prognostic impact of plasma GDF-15 and BNP in acute HF. METHODS AND RESULTS: We studied a subgroup of patients prospectively recruited in an acute HF registry (follow-up: 2 years; endpoint: all-cause mortality). Cox regression multivariate models were built to study the association of GDF-15 and mortality. Further cross-classification according to discharge GDF-15 (mean) and BNP (mean) and association with mortality was studied. We studied 158 patients: seventy-nine were male, mean age was 75 years, 55.1% had left ventricular ejection fraction < 40%, mean discharge BNP was 1000 pg/mL, and mean GDF-15 was 3013 ng/mL. Higher BNP and GDF-15 predicted 2-year mortality. Patients with GDF-15 ≥ 3000 ng/mL had a multivariate adjusted 2-year death risk of 1.86 (1.08-3.18). Patients discharged with both BNP and GDF-15 above the mean had an adjusted hazard ratio of 4.33 (2.07-9.06) when compared with those with both

Liver cytolysis in acute heart failure: What does it mean? Clinical profile and outcomes of a prospective hospital cohort.

UNLABELLED: Abnormal liver function tests (LFTs) are a common manifestation of heart failure (HF). Our purpose was to characterize patients hospitalized for acute HF (AHF) with liver cytolysis, analyze cytolysis predictors and explore its prognostic implications. METHODS: In a prospective cohort study, we enrolled patients with AHF consecutively admitted to the Internal Medicine Department of University Hospital between January 2009 and December 2010, and recorded demographic, clinical, laboratory and echocardiogram parameters. A logistic regression was done to identify cytolysis predictors. In survival analysis primary endpoints were all-cause death, readmission due to AHF, and the combined event of all-cause death and readmission for AHF at 90days of follow-up. RESULTS: Fifty-eight patients had cytolysis at admission. AHF attributed to atrial fibrillation (OR 3.235), higher heart rate at admission (OR 1.028), cold/wet profile at admission (OR 7.12) and ejection fraction <30% (OR 2.316) were independent predictors of cytolysis. Death occurred more frequently during follow-up in the cytolysis group (27.6 vs. 15.1%, p=0.014, respectively). On survival analysis, cytolysis remained an independent predictor of death at 90days when adjusted to age (HR 1.066), male gender (HR 1.884), valvular etiology (HR 2.365), neurologic status at admission (sleepy HR 3.854; confusion HR 3.176) and cardiac output (HR 0.762). CONCLUSION: Cytolysis may be a marker of systemic hypoperfusion, so strategies to improve hemodynamic profile should be considered, especially in the presence of cold/wet clinical profile, AHF attributed to AF, tachycardia, and EF<30%. Cytolysis is associated with higher mortality at 90days in patients with AHF.

PaO2/FiO2 Deterioration During Stable Extracorporeal Membrane Oxygenation Associates With Protracted Recovery and Increased Mortality in Severe Acute Respiratory Distress Syndrome.

BACKGROUND: During extracorporeal membrane oxygenation (ECMO), arterial oxygen partial pressure to fractional inspired oxygen (PaO2/FiO2; PF ratio reflects native and artificial lung blood oxygenation). In this study we analyzed PF ratio during ECMO support and its association with clinical outcome. METHODS: This was a single-center observational study of adult patients (n = 81) undergoing veno-venous ECMO support for severe acute respiratory distress syndrome. RESULTS: In 37 patients (46%) PF ratio decreased from ECMO-day 1 to ECMO-day 7 (PF ratio deterioration [PF-d]; -37 ± 6.1 mm Hg), whereas in 44 patients PF ratio improved (PF-i; 65 ± 10.8 mm Hg). PF-d group required prolonged ECMO (median 21 days [interquartile range (IQR)]:14-35 days] versus 13 days [IQR: 10-20 days]) and invasive mechanical ventilation (median 33 days [IQR: 24-52 days] versus 26 days [IQR: 22-34 days]), longer intensive care unit (median 44 days [IQR: 32-74 days] versus 30 days [IQR: 25-47 days]), and hospital (median 66 days [IQR: 39-95 days] versus 36 days [IQR: 28-54 days]) lengths of stay, with higher hospital mortality rates (48.7% versus 22.7%). ECMO oxygenation did not explain PF ratio variation that remained stable in PF-d and decreased in PF-i (198 ± 12.7 mL/min versus 171 ± 8.8 mL/min). Pre-ECMO PF ratio, neuromuscular blockade, and prone position, as well as ventilatory variables did not differ between groups. The PF-d group was older (49 ± 2.1 years versus 41 ± 1.8 years) and presented lower Respiratory Extracorporeal Membrane Oxygenation Survival Prediction (RESP) scores (0.57 ± 0.63 versus 2.2 ± 0.52). With the use of logistic regression, PF ratio variation remained an independent predictor of hospital mortality after adjusting for age or RESP score. CONCLUSIONS: In severe acute respiratory distress syndrome, PF ratio deterioration during stable ECMO associates with protracted recovery and increased mortality, not accounted for by patient baseline characteristics, acute respiratory distress syndrome severity, or pre-ECMO management.