A systematic review of machine learning models for management, prediction and classification of ARDS.

AIM: Acute respiratory distress syndrome or ARDS is an acute, severe form of respiratory failure characterised by poor oxygenation and bilateral pulmonary infiltrates. Advancements in signal processing and machine learning have led to promising solutions for classification, event detection and predictive models in the management of ARDS. METHOD: In this review, we provide systematic description of different studies in the application of Machine Learning (ML) and artificial intelligence for management, prediction, and classification of ARDS. We searched the following databases: Google Scholar, PubMed, and EBSCO from 2009 to 2023. A total of 243 studies was screened, in which, 52 studies were included for review and analysis. We integrated knowledge of previous work providing the state of art and overview of explainable decision models in machine learning and have identified areas for future research. RESULTS: Gradient boosting is the most common and successful method utilised in 12 (23.1%) of the studies. Due to limitation of data size available, neural network and its variation is used by only 8 (15.4%) studies. Whilst all studies used cross validating technique or separated database for validation, only 1 study validated the model with clinician input. Explainability methods were presented in 15 (28.8%) of studies with the most common method is feature importance which used 14 times. CONCLUSION: For databases of 5000 or fewer samples, extreme gradient boosting has the highest probability of success. A large, multi-region, multi centre database is required to reduce bias and take advantage of neural network method. A framework for validating with and explaining ML model to clinicians involved in the management of ARDS would be very helpful for development and deployment of the ML model.

Evaluation of the classifications of severity in acute respiratory distress syndrome in childhood by the Berlin Consensus and the Pediatric Acute Lung Injury Consensus Conference.

OBJECTIVE: To compare two methods for defining and classifying the severity of pediatric acute respiratory distress syndrome: the Berlin classification, which uses the relationship between the partial pressure of oxygen and the fraction of inspired oxygen, and the classification of the Pediatric Acute Lung Injury Consensus Conference, which uses the oxygenation index. METHODS: This was a prospective study of patients aged 0 - 18 years with a diagnosis of acute respiratory distress syndrome who were invasively mechanically ventilated and provided one to three arterial blood gas samples, totaling 140 valid measurements. These measures were evaluated for correlation using the Spearman test and agreement using the kappa coefficient between the two classifications, initially using the general population of the study and then subdividing it into patients with and without bronchospasm and those with and without the use of neuromuscular blockers. The effect of these two factors (bronchospasm and neuromuscular blocking agent) separately and together on both classifications was also assessed using two-way analysis of variance. RESULTS: In the general population, who were 54 patients aged 0 - 18 years a strong negative correlation was found by Spearman's test (ρ -0.91; p < 0.001), and strong agreement was found by the kappa coefficient (0.62; p < 0.001) in the comparison between Berlin and Pediatric Acute Lung Injury Consensus Conference. In the populations with and without bronchospasm and who did and did not use neuromuscular blockers, the correlation coefficients were similar to those of the general population, though among patients not using neuromuscular blockers, there was greater agreement between the classifications than for patients using neuromuscular blockers (kappa 0.67 versus 0.56, p < 0.001 for both). Neuromuscular blockers had a significant effect on the relationship between the partial pressure of oxygen and the fraction of inspired oxygen (analysis of variance; F: 12.9; p < 0.001) and the oxygenation index (analysis of variance; F: 8.3; p = 0.004). CONCLUSION: There was a strong correlation and agreement between the two classifications in the general population and in the subgroups studied. Use of neuromuscular blockers had a significant effect on the severity of acute respiratory distress syndrome.

Biological basis of critical illness subclasses: from the bedside to the bench and back again.

Critical illness syndromes including sepsis, acute respiratory distress syndrome, and acute kidney injury (AKI) are associated with high in-hospital mortality and long-term adverse health outcomes among survivors. Despite advancements in care, clinical and biological heterogeneity among patients continues to hamper identification of efficacious therapies. Precision medicine offers hope by identifying patient subclasses based on clinical, laboratory, biomarker and 'omic' data and potentially facilitating better alignment of interventions. Within the previous two decades, numerous studies have made strides in identifying gene-expression based endotypes and clinico-biomarker based phenotypes among critically ill patients associated with differential outcomes and responses to treatment. In this state-of-the-art review, we summarize the biological similarities and differences across the various subclassification schemes among critically ill patients. In addition, we highlight current translational gaps, the need for advanced scientific tools, human-relevant disease models, to gain a comprehensive understanding of the molecular mechanisms underlying critical illness subclasses.

Inflammatory subphenotypes previously identified in ARDS are associated with mortality at intensive care unit discharge: a secondary analysis of a prospective observational study.

BACKGROUND: Intensive care unit (ICU)-survivors have an increased risk of mortality after discharge compared to the general population. On ICU admission subphenotypes based on the plasma biomarker levels of interleukin-8, protein C and bicarbonate have been identified in patients admitted with acute respiratory distress syndrome (ARDS) that are prognostic of outcome and predictive of treatment response. We hypothesized that if these inflammatory subphenotypes previously identified among ARDS patients are assigned at ICU discharge in a more general critically ill population, they are associated with short- and long-term outcome. METHODS: A secondary analysis of a prospective observational cohort study conducted in two Dutch ICUs between 2011 and 2014 was performed. All patients discharged alive from the ICU were at ICU discharge adjudicated to the previously identified inflammatory subphenotypes applying a validated parsimonious model using variables measured median 10.6 h [IQR, 8.0-31.4] prior to ICU discharge. Subphenotype distribution at ICU discharge, clinical characteristics and outcomes were analyzed. As a sensitivity analysis, a latent class analysis (LCA) was executed for subphenotype identification based on plasma protein biomarkers at ICU discharge reflective of coagulation activation, endothelial cell activation and inflammation. Concordance between the subphenotyping strategies was studied. RESULTS: Of the 8332 patients included in the original cohort, 1483 ICU-survivors had plasma biomarkers available and could be assigned to the inflammatory subphenotypes. At ICU discharge 6% (n = 86) was assigned to the hyperinflammatory and 94% (n = 1397) to the hypoinflammatory subphenotype. Patients assigned to the hyperinflammatory subphenotype were discharged with signs of more severe organ dysfunction (SOFA scores 7 [IQR 5-9] vs. 4 [IQR 2-6], p < 0.001). Mortality was higher in patients assigned to the hyperinflammatory subphenotype (30-day mortality 21% vs. 11%, p = 0.005; one-year mortality 48% vs. 28%, p < 0.001). LCA deemed 2 subphenotypes most suitable. ICU-survivors from class 1 had significantly higher mortality compared to class 2. Patients belonging to the hyperinflammatory subphenotype were mainly in class 1. CONCLUSIONS: Patients assigned to the hyperinflammatory subphenotype at ICU discharge showed significantly stronger anomalies in coagulation activation, endothelial cell activation and inflammation pathways implicated in the pathogenesis of critical disease and increased mortality until one-year follow up.

Performance of acute respiratory distress syndrome definitions in a high acuity paediatric intensive care unit.

BACKGROUND: For years, paediatric critical care practitioners used the adult American European Consensus Conference (AECC) and revised Berlin Definition (BD) for acute respiratory distress syndrome (ARDS) to study the epidemiology of paediatric ARDS (PARDS). In 2015, the paediatric specific definition, Paediatric Acute Lung Injury Consensus Conference (PALICC) was developed. The use of non-invasive metrics of oxygenation to stratify disease severity were introduced in this definition, although this potentially may lead to a confounding effect of disease severity since it is more common to place indwelling arterial lines in sicker patients. We tested the hypothesis that PALICC outperforms AECC/BD in our high acuity PICU, which employs a liberal use of indwelling arterial lines and high-frequency oscillatory ventilation (HFOV). METHODS: We retrospectively collected data from children < 18 years mechanically ventilated for at least 24 h in our tertiary care, university-affiliated paediatric intensive care unit. The primary endpoint was the difference in the number of PARDS cases between AECC/BD and PALICC. Secondary endpoints included mortality and ventilator free days. Performance was assessed by the area under the receiver operating characteristics curve (AUC-ROC). RESULTS: Data from 909 out of 2433 patients was eligible for analysis. AECC/BD identified 35 (1.4%) patients (mortality 25.7%), whereas PALICC identified 135 (5.5%) patients (mortality 14.1%). All but two patients meeting AECC/Berlin criteria were also identified by PALICC. Almost half of the cohort (45.2%) had mild, 33.3% moderate and 21.5% severe PALICC PARDS at onset. Highest mortality rates were seen in patients with AECC acute lung injury (ALI)/mild Berlin and severe PALICC PARDS. The AUC-ROC for Berlin was the highest 24 h (0.392 [0.124-0.659]) after onset. PALICC showed the highest AUC-ROC at the same moment however higher than Berlin (0.531 [0.345-0.716]). Mortality rates were significantly increased in patients with bilateral consolidations (9.3% unilateral vs 26.3% bilateral, p = 0.025). CONCLUSIONS: PALICC identified more new cases PARDS than the AECC/Berlin definition. However, both PALICC and Berlin performed poorly in terms of mortality risk stratification. The presence of bilateral consolidations was associated with a higher mortality rate. Our findings may be considered in future modifications of the PALICC criteria.

Latent Class Analysis Reveals COVID-19-related Acute Respiratory Distress Syndrome Subgroups with Differential Responses to Corticosteroids.

Rationale: Two distinct subphenotypes have been identified in acute respiratory distress syndrome (ARDS), but the presence of subgroups in ARDS associated with coronavirus disease (COVID-19) is unknown. Objectives: To identify clinically relevant, novel subgroups in COVID-19-related ARDS and compare them with previously described ARDS subphenotypes. Methods: Eligible participants were adults with COVID-19 and ARDS at Columbia University Irving Medical Center. Latent class analysis was used to identify subgroups with baseline clinical, respiratory, and laboratory data serving as partitioning variables. A previously developed machine learning model was used to classify patients as the hypoinflammatory and hyperinflammatory subphenotypes. Baseline characteristics and clinical outcomes were compared between subgroups. Heterogeneity of treatment effect for corticosteroid use in subgroups was tested. Measurements and Main Results: From March 2, 2020, to April 30, 2020, 483 patients with COVID-19-related ARDS met study criteria. A two-class latent class analysis model best fit the population (P = 0.0075). Class 2 (23%) had higher proinflammatory markers, troponin, creatinine, and lactate, lower bicarbonate, and lower blood pressure than class 1 (77%). Ninety-day mortality was higher in class 2 versus class 1 (75% vs. 48%; P < 0.0001). Considerable overlap was observed between these subgroups and ARDS subphenotypes. Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) RT-PCR cycle threshold was associated with mortality in the hypoinflammatory but not the hyperinflammatory phenotype. Heterogeneity of treatment effect to corticosteroids was observed (P = 0.0295), with improved mortality in the hyperinflammatory phenotype and worse mortality in the hypoinflammatory phenotype, with the caveat that corticosteroid treatment was not randomized. Conclusions: We identified two COVID-19-related ARDS subgroups with differential outcomes, similar to previously described ARDS subphenotypes. SARS-CoV-2 PCR cycle threshold had differential value for predicting mortality in the subphenotypes. The subphenotypes had differential treatment responses to corticosteroids.

Stratification for Identification of Prognostic Categories In the Acute RESpiratory Distress Syndrome (SPIRES) Score.

OBJECTIVES: To develop a scoring model for stratifying patients with acute respiratory distress syndrome into risk categories (Stratification for identification of Prognostic categories In the acute RESpiratory distress syndrome score) for early prediction of death in the ICU, independent of the underlying disease and cause of death. DESIGN: A development and validation study using clinical data from four prospective, multicenter, observational cohorts. SETTING: A network of multidisciplinary ICUs. PATIENTS: One-thousand three-hundred one patients with moderate-to-severe acute respiratory distress syndrome managed with lung-protective ventilation. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: The study followed Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis guidelines for prediction models. We performed logistic regression analysis, bootstrapping, and internal-external validation of prediction models with variables collected within 24 hours of acute respiratory distress syndrome diagnosis in 1,000 patients for model development. Primary outcome was ICU death. The Stratification for identification of Prognostic categories In the acute RESpiratory distress syndrome score was based on patient's age, number of extrapulmonary organ failures, values of end-inspiratory plateau pressure, and ratio of Pao2 to Fio2 assessed at 24 hours of acute respiratory distress syndrome diagnosis. The pooled area under the receiver operating characteristic curve across internal-external validations was 0.860 (95% CI, 0.831-0.890). External validation in a new cohort of 301 acute respiratory distress syndrome patients confirmed the accuracy and robustness of the scoring model (area under the receiver operating characteristic curve = 0.870; 95% CI, 0.829-0.911). The Stratification for identification of Prognostic categories In the acute RESpiratory distress syndrome score stratified patients in three distinct prognostic classes and achieved better prediction of ICU death than ratio of Pao2 to Fio2 at acute respiratory distress syndrome onset or at 24 hours, Acute Physiology and Chronic Health Evaluation II score, or Sequential Organ Failure Assessment scale. CONCLUSIONS: The Stratification for identification of Prognostic categories In the acute RESpiratory distress syndrome score represents a novel strategy for early stratification of acute respiratory distress syndrome patients into prognostic categories and for selecting patients for therapeutic trials.

Novel criteria to classify ARDS severity using a machine learning approach.

BACKGROUND: Usually, arterial oxygenation in patients with the acute respiratory distress syndrome (ARDS) improves substantially by increasing the level of positive end-expiratory pressure (PEEP). Herein, we are proposing a novel variable [PaO2/(FiO2xPEEP) or P/FPE] for PEEP ≥ 5 to address Berlin's definition gap for ARDS severity by using machine learning (ML) approaches. METHODS: We examined P/FPE values delimiting the boundaries of mild, moderate, and severe ARDS. We applied ML to predict ARDS severity after onset over time by comparing current Berlin PaO2/FiO2 criteria with P/FPE under three different scenarios. We extracted clinical data from the first 3 ICU days after ARDS onset (N = 2738, 1519, and 1341 patients, respectively) from MIMIC-III database according to Berlin criteria for severity. Then, we used the multicenter database eICU (2014-2015) and extracted data from the first 3 ICU days after ARDS onset (N = 5153, 2981, and 2326 patients, respectively). Disease progression in each database was tracked along those 3 ICU days to assess ARDS severity. Three robust ML classification techniques were implemented using Python 3.7 (LightGBM, RF, and XGBoost) for predicting ARDS severity over time. RESULTS: P/FPE ratio outperformed PaO2/FiO2 ratio in all ML models for predicting ARDS severity after onset over time (MIMIC-III: AUC 0.711-0.788 and CORR 0.376-0.566; eICU: AUC 0.734-0.873 and CORR 0.511-0.745). CONCLUSIONS: The novel P/FPE ratio to assess ARDS severity after onset over time is markedly better than current PaO2/FiO2 criteria. The use of P/FPE could help to manage ARDS patients with a more precise therapeutic regimen for each ARDS category of severity.

NT-proBNP levels in the identification and classification of pulmonary transfusion reactions.

BACKGROUND: Consensus definitions for transfusion-related acute lung injury (TRALI) and transfusion-associated circulatory overload (TACO) have recently been revised; however, pulmonary transfusion reactions remain difficult to diagnose. We hypothesized that N-terminal pro-brain natriuretic peptide (NT-proBNP) levels could have utility in the identification and classification of pulmonary transfusion reactions. STUDY DESIGN AND METHODS: We performed a secondary analysis of a case-control study of pulmonary transfusion reactions at four academic hospitals. We evaluated clinical data and measured NT-proBNP levels prior to and following transfusion in patients with TACO (n = 160), transfused acute respiratory distress syndrome (ARDS) [n = 51], TRALI [n = 12], TACO/TRALI [n = 7], and controls [n = 335]. We used Wilcoxon Rank-Sum tests to compare NT-proBNP levels, and classification and regression tree (CART) algorithms to produce a ranking of covariates in order of relative importance for differentiating TACO from transfused controls. RESULTS: Pre-transfusion NT-proBNP levels were elevated in cases of transfused ARDS and TACO (both P < .001) but not TRALI (P = .31) or TACO/TRALI (P = .23) compared to transfused controls. Pre-transfusion NT-proBNP levels were higher in cases of transfused ARDS or TRALI with a diagnosis of sepsis compared to those without (P < .05 for both). CART analyses resulted in similar differentiation of patients with TACO from transfused controls for models utilizing either NT-proBNP levels (AUC 0.83) or echocardiogram results (AUC 0.80). CONCLUSIONS: NT-proBNP levels may have utility in the classification of pulmonary transfusion reactions. Prospective studies are needed to test the predictive utility of pre-transfusion NT-proBNP in conjunction with other clinical factors in identifying patients at risk of pulmonary transfusion reactions.

Prevalence of phenotypes of acute respiratory distress syndrome in critically ill patients with COVID-19: a prospective observational study.

BACKGROUND: In acute respiratory distress syndrome (ARDS) unrelated to COVID-19, two phenotypes, based on the severity of systemic inflammation (hyperinflammatory and hypoinflammatory), have been described. The hyperinflammatory phenotype is known to be associated with increased multiorgan failure and mortality. In this study, we aimed to identify these phenotypes in COVID-19-related ARDS. METHODS: In this prospective observational study done at two UK intensive care units, we recruited patients with ARDS due to COVID-19. Demographic, clinical, and laboratory data were collected at baseline. Plasma samples were analysed for interleukin-6 (IL-6) and soluble tumour necrosis factor receptor superfamily member 1A (TNFR1) using a novel point-of-care assay. A parsimonious regression classifier model was used to calculate the probability for the hyperinflammatory phenotype in COVID-19 using IL-6, soluble TNFR1, and bicarbonate levels. Data from this cohort was compared with patients with ARDS due to causes other than COVID-19 recruited to a previous UK multicentre, randomised controlled trial of simvastatin (HARP-2). FINDINGS: Between March 17 and April 25, 2020, 39 patients were recruited to the study. Median ratio of partial pressure of arterial oxygen to fractional concentration of oxygen in inspired air (PaO2/FiO2) was 18 kpa (IQR 15-21) and acute physiology and chronic health evaluation II score was 12 (10-16). 17 (44%) of 39 patients had died by day 28 of the study. Compared with survivors, patients who died were older and had lower PaO2/FiO2. The median probability for the hyperinflammatory phenotype was 0·03 (IQR 0·01-0·2). Depending on the probability cutoff used to assign class, the prevalence of the hyperinflammatory phenotype was between four (10%) and eight (21%) of 39, which is lower than the proportion of patients with the hyperinflammatory phenotype in HARP-2 (186 [35%] of 539). Using the Youden index cutoff (0·274) to classify phenotype, five (63%) of eight patients with the hyperinflammatory phenotype and 12 (39%) of 31 with the hypoinflammatory phenotype died. Compared with matched patients recruited to HARP-2, levels of IL-6 were similar in our cohort, whereas soluble TNFR1 was significantly lower in patients with COVID-19-associated ARDS. INTERPRETATION: In this exploratory analysis of 39 patients, ARDS due to COVID-19 was not associated with higher systemic inflammation and was associated with a lower prevalence of the hyperinflammatory phenotype than that observed in historical ARDS data. This finding suggests that the excess mortality observed in COVID-19-related ARDS is unlikely to be due to the upregulation of inflammatory pathways described by the parsimonious model. FUNDING: US National Institutes of Health, Innovate UK, and Randox.

Reclassifying severity after 48 hours could better predict mortality in acute respiratory distress syndrome.

BACKGROUND: Disease severity may change in the first week after acute respiratory distress syndrome (ARDS) onset. The aim of this study was to evaluate whether the reclassification of disease severity after 48 h (i.e. day 3) of ARDS onset could help in predicting mortality and determine factors associated with ARDS persistence and mortality. METHODS: We performed a secondary analysis of a 3-year prospective, observational cohort study of ARDS in a tertiary care referral center. Disease severity was reclassified after 48 h of enrollment, and cases that still fulfilled the Berlin criteria were regarded as nonresolving ARDS. RESULTS: A total of 1034 ARDS patients were analyzed. Overall hospital mortality was 57.7% (56.7%, 57.5%, and 58.6% for patients with initial mild, moderate, and severe ARDS, respectively, p = 0.189). On day 3 reclassification, the hospital mortality rates were as follows: resolved (42.1%), mild (47.9%), moderate (62.4%), and severe ARDS (76.1%) (p < 0.001). Patients with improving severity on day 3 had lower mortality (48.8%), whereas patients with the same or worsening severity on day 3 had higher mortality (62.7% and 76.3%, respectively). Patients who were older, had lower PaO2/FiO2, or higher positive end-expiratory pressure on day 1 were significantly associated with nonresolving ARDS on day 3. A Cox regression model with ARDS severity as a time-dependent covariate and competing risk analysis demonstrated that ARDS severity was independently associated with hospital mortality, and nonresolving ARDS had significantly increased hazard of death than resolved ARDS (p < 0.0001). Cumulative mortality curve for ARDS severity comparisons demonstrated significantly different (overall comparison, p < 0.001). CONCLUSIONS: Reclassification of disease severity after 48 h of ARDS onset could help to divide patients into subgroups with greater separation in terms of mortality. The reviews of this paper are available via the supplemental material section.

Subphenotypes in critical care: translation into clinical practice.

Despite progress in the supportive care available for critically ill patients, few advances have been made in the search for effective disease-modifying therapeutic options. The fact that many trials in critical care medicine have not identified a treatment benefit is probably due, in part, to the underlying heterogeneity of critical care syndromes. Numerous approaches have been proposed to divide populations of critically ill patients into more meaningful subgroups (subphenotypes), some of which might be more useful than others. Subclassification systems driven by clinical features and biomarkers have been proposed for acute respiratory distress syndrome, sepsis, acute kidney injury, and pancreatitis. Identifying the systems that are most useful and biologically meaningful could lead to a better understanding of the pathophysiology of critical care syndromes and the discovery of new treatment targets, and allow recruitment in future therapeutic trials to focus on predicted responders. This Review discusses proposed subphenotypes of critical illness syndromes and highlights the issues that will need to be addressed to translate subphenotypes into clinical practice.

Machine Learning Classifier Models Can Identify Acute Respiratory Distress Syndrome Phenotypes Using Readily Available Clinical Data.

Rationale: Two distinct phenotypes of acute respiratory distress syndrome (ARDS) with differential clinical outcomes and responses to randomly assigned treatment have consistently been identified in randomized controlled trial cohorts using latent class analysis. Plasma biomarkers, key components in phenotype identification, currently lack point-of-care assays and represent a barrier to the clinical implementation of phenotypes.Objectives: The objective of this study was to develop models to classify ARDS phenotypes using readily available clinical data only.Methods: Three randomized controlled trial cohorts served as the training data set (ARMA [High vs. Low Vt], ALVEOLI [Assessment of Low Vt and Elevated End-Expiratory Pressure to Obviate Lung Injury], and FACTT [Fluids and Catheter Treatment Trial]; n = 2,022), and a fourth served as the validation data set (SAILS [Statins for Acutely Injured Lungs from Sepsis]; n = 745). A gradient-boosted machine algorithm was used to develop classifier models using 24 variables (demographics, vital signs, laboratory, and respiratory variables) at enrollment. In two secondary analyses, the ALVEOLI and FACTT cohorts each, individually, served as the validation data set, and the remaining combined cohorts formed the training data set for each analysis. Model performance was evaluated against the latent class analysis-derived phenotype.Measurements and Main Results: For the primary analysis, the model accurately classified the phenotypes in the validation cohort (area under the receiver operating characteristic curve [AUC], 0.95; 95% confidence interval [CI], 0.94-0.96). Using a probability cutoff of 0.5 to assign class, inflammatory biomarkers (IL-6, IL-8, and sTNFR-1; P < 0.0001) and 90-day mortality (38% vs. 24%; P = 0.0002) were significantly higher in the hyperinflammatory phenotype as classified by the model. Model accuracy was similar when ALVEOLI (AUC, 0.94; 95% CI, 0.92-0.96) and FACTT (AUC, 0.94; 95% CI, 0.92-0.95) were used as the validation cohorts. Significant treatment interactions were observed with the clinical classifier model-assigned phenotypes in both ALVEOLI (P = 0.0113) and FACTT (P = 0.0072) cohorts.Conclusions: ARDS phenotypes can be accurately identified using machine learning models based on readily available clinical data and may enable rapid phenotype identification at the bedside.

Acute respiratory failure in COVID-19: is it "typical" ARDS?

In December 2019, an outbreak of coronavirus disease 2019 (COVID-19) was identified in Wuhan, China. The World Health Organization (WHO) declared this outbreak a significant threat to international health. COVID-19 is highly infectious and can lead to fatal comorbidities especially acute respiratory distress syndrome (ARDS). Thus, fully understanding the characteristics of COVID-19-related ARDS is conducive to early identification and precise treatment. We aimed to describe the characteristics of COVID-19-related ARDS and to elucidate the differences from ARDS caused by other factors. COVID-19 mainly affected the respiratory system with minor damage to other organs. Injury to the alveolar epithelial cells was the main cause of COVID-19-related ARDS, and endothelial cells were less damaged with therefore less exudation. The clinical manifestations were relatively mild in some COVID-19 patients, which was inconsistent with the severity of laboratory and imaging findings. The onset time of COVID-19-related ARDS was 8-12 days, which was inconsistent with ARDS Berlin criteria, which defined a 1-week onset limit. Some of these patients might have a relatively normal lung compliance. The severity was redefined into three stages according to its specificity: mild, mild-moderate, and moderate-severe. HFNO can be safe in COVID-19-related ARDS patients, even in some moderate-severe patients. The more likely cause of death is severe respiratory failure. Thus, the timing of invasive mechanical ventilation is very important. The effects of corticosteroids in COVID-19-related ARDS patients were uncertain. We hope to help improve the prognosis of severe cases and reduce the mortality.

ARDS Subphenotypes: Understanding a Heterogeneous Syndrome.

This article is one of ten reviews selected from the Annual Update in Intensive Care and Emergency Medicine 2020. Other selected articles can be found online at https://www.biomedcentral.com/collections/annualupdate2020. Further information about the Annual Update in Intensive Care and Emergency Medicine is available from http://www.springer.com/series/8901.

Development and validation of parsimonious algorithms to classify acute respiratory distress syndrome phenotypes: a secondary analysis of randomised controlled trials.

BACKGROUND: Using latent class analysis (LCA) in five randomised controlled trial (RCT) cohorts, two distinct phenotypes of acute respiratory distress syndrome (ARDS) have been identified: hypoinflammatory and hyperinflammatory. The phenotypes are associated with differential outcomes and treatment response. The objective of this study was to develop parsimonious models for phenotype identification that could be accurate and feasible to use in the clinical setting. METHODS: In this retrospective study, three RCT cohorts from the National Lung, Heart, and Blood Institute ARDS Network (ARMA, ALVEOLI, and FACTT) were used as the derivation dataset (n=2022), from which the machine learning and logistic regression classifer models were derived, and a fourth (SAILS; n=715) from the same network was used as the validation test set. LCA-derived phenotypes in all of these cohorts served as the reference standard. Machine-learning algorithms (random forest, bootstrapped aggregating, and least absolute shrinkage and selection operator) were used to select a maximum of six important classifier variables, which were then used to develop nested logistic regression models. Only cases with complete biomarker data in the derivation dataset were used for variable selection. The best logistic regression models based on parsimony and predictive accuracy were then evaluated in the validation test set. Finally, the models' prognostic validity was tested in two external ARDS clinical trial datasets (START and HARP-2) by assessing mortality at days 28, 60, and 90 and ventilator-free days to day 28. FINDINGS: The six most important classifier variables were interleukin (IL)-8, IL-6, protein C, soluble tumour necrosis factor receptor 1, bicarbonate, and vasopressor use. From the nested models, three-variable (IL-8, bicarbonate, and protein C) and four-variable (3-variable plus vasopressor use) models were adjudicated to be the best performing. In the validation test set, both models showed good accuracy (AUC 0·94 [95% CI 0·92-0·95] for the three-variable model and 0·95 [95% CI 0·93-0·96] for the four-variable model) against LCA classifications. As with LCA-derived phenotypes, the hyperinflammatory phenotype as identified by the classifier model was associated with higher mortality at day 90 (87 [39%] of 223 patients vs 112 [23%] of 492 patients; p<0·0001) and fewer ventilator-free days (median 14 days [IQR 0-22] vs 22 days [0-25]; p<0·0001). In the external validation datasets, three-variable models developed in the derivation dataset identified two phenotypes with distinct clinical features and outcomes consistent with previous findings, including differential survival with simvastatin versus placebo in HARP-2 (p=0·023 for survival at 28 days). INTERPRETATION: ARDS phenotypes can be accurately identified with parsimonious classifier models using three or four variables. Pending the development of real-time testing for key biomarkers and prospective validation, these models could facilitate identification of ARDS phenotypes to enable their application in clinical trials and practice. FUNDING: National Institutes of Health.

Host-Response Subphenotypes Offer Prognostic Enrichment in Patients With or at Risk for Acute Respiratory Distress Syndrome.

OBJECTIVES: Classification of patients with acute respiratory distress syndrome into hyper- and hypoinflammatory subphenotypes using plasma biomarkers may facilitate more effective targeted therapy. We examined whether established subphenotypes are present not only in patients with acute respiratory distress syndrome but also in patients at risk for acute respiratory distress syndrome (ARFA) and then assessed the prognostic information of baseline subphenotyping on the evolution of host-response biomarkers and clinical outcomes. DESIGN: Prospective, observational cohort study. SETTING: Medical ICU at a tertiary academic medical center. PATIENTS: Mechanically ventilated patients with acute respiratory distress syndrome or ARFA. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: We performed longitudinal measurements of 10 plasma biomarkers of host injury and inflammation. We applied unsupervised latent class analysis methods utilizing baseline clinical and biomarker variables and demonstrated that two-class models (hyper- vs hypoinflammatory subphenotypes) offered improved fit compared with one-class models in both patients with acute respiratory distress syndrome and ARFA. Baseline assignment to the hyperinflammatory subphenotype (39/104 [38%] acute respiratory distress syndrome and 30/108 [28%] ARFA patients) was associated with higher severity of illness by Sequential Organ Failure Assessment scores and incidence of acute kidney injury in patients with acute respiratory distress syndrome, as well as higher 30-day mortality and longer duration of mechanical ventilation in ARFA patients (p < 0.0001). Hyperinflammatory patients exhibited persistent elevation of biomarkers of innate immunity for up to 2 weeks postintubation. CONCLUSIONS: Our results suggest that two distinct subphenotypes are present not only in patients with established acute respiratory distress syndrome but also in patients at risk for its development. Hyperinflammatory classification at baseline is associated with higher severity of illness, worse clinical outcomes, and trajectories of persistently elevated biomarkers of host injury and inflammation during acute critical illness compared with hypoinflammatory patients. Our findings provide strong rationale for examining treatment effect modifications by subphenotypes in randomized clinical trials to inform precision therapeutic approaches in critical care.