Identification of acute respiratory distress syndrome subphenotypes de novo using routine clinical data: a retrospective analysis of ARDS clinical trials

Identification and Prediction of Clinical Phenotypes in Hospitalized Patients With COVID-19: Machine Learning From Medical Records

BACKGROUND

There is significant heterogeneity in disease progression among hospitalized patients with COVID-19. The pathogenesis of SARS-CoV-2 infection is attributed to a complex interplay between virus and host immune response that in some patients unpredictably and rapidly leads to "hyperinflammation" associated with increased risk of mortality. The early identification of patients at risk of progression to hyperinflammation may help inform timely therapeutic decisions and lead to improved outcomes.

OBJECTIVE

The primary objective of this study was to use machine learning to reproducibly identify specific risk-stratifying clinical phenotypes across hospitalized patients with COVID-19 and compare treatment response characteristics and outcomes. A secondary objective was to derive a predictive phenotype classification model using routinely available early encounter data that may be useful in informing optimal COVID-19 bedside clinical management.

METHODS

This was a retrospective analysis of electronic health record data of adult patients (N=4379) who were admitted to a Johns Hopkins Health System hospital for COVID-19 treatment from 2020 to 2021. Phenotypes were identified by clustering 38 routine clinical observations recorded during inpatient care. To examine the reproducibility and validity of the derived phenotypes, patient data were randomly divided into 2 cohorts, and clustering analysis was performed independently for each cohort. A predictive phenotype classifier using the gradient-boosting machine method was derived using routine clinical observations recorded during the first 6 hours following admission.

RESULTS

A total of 2 phenotypes (designated as phenotype 1 and phenotype 2) were identified in patients admitted for COVID-19 in both the training and validation cohorts with similar distributions of features, correlations with biomarkers, treatments, comorbidities, and outcomes. In both the training and validation cohorts, phenotype-2 patients were older; had elevated markers of inflammation; and were at an increased risk of requiring intensive care unit-level care, developing sepsis, and mortality compared with phenotype-1 patients. The gradient-boosting machine phenotype prediction model yielded an area under the curve of 0.89 and a positive predictive value of 0.83.

CONCLUSIONS

Using machine learning clustering, we identified and internally validated 2 clinical COVID-19 phenotypes with distinct treatment or response characteristics consistent with similar 2-phenotype models derived from other hospitalized populations with COVID-19, supporting the reliability and generalizability of these findings. COVID-19 phenotypes can be accurately identified using machine learning models based on readily available early encounter clinical data. A phenotype prediction model based on early encounter data may be clinically useful for timely bedside risk stratification and treatment personalization.

Clinical Sepsis Phenotypes in Critically Ill Patients

Sepsis, defined as the life-threatening dysregulated host response to an infection leading to organ dysfunction, is considered as one of the leading causes of mortality worldwide, especially in intensive care units (ICU). Moreover, sepsis remains an enigmatic clinical syndrome, with complex pathophysiology incompletely understood and a great heterogeneity both in terms of clinical expression, patient response to currently available therapeutic interventions and outcomes. This heterogeneity proves to be a major obstacle in our quest to deliver improved treatment in septic critical care patients; thus, identification of clinical phenotypes is absolutely necessary. Although this might be seen as an extremely difficult task, nowadays, artificial intelligence and machine learning techniques can be recruited to quantify similarities between individuals within sepsis population and differentiate them into distinct phenotypes regarding not only temperature, hemodynamics or type of organ dysfunction, but also fluid status/responsiveness, trajectories in ICU and outcome. Hopefully, we will eventually manage to determine both the subgroup of septic patients that will benefit from a therapeutic intervention and the correct timing of applying the intervention during the disease process.

ESICM guidelines on acute respiratory distress syndrome: definition, phenotyping and respiratory support strategies

The aim of these guidelines is to update the 2017 clinical practice guideline (CPG) of the European Society of Intensive Care Medicine (ESICM). The scope of this CPG is limited to adult patients and to non-pharmacological respiratory support strategies across different aspects of acute respiratory distress syndrome (ARDS), including ARDS due to coronavirus disease 2019 (COVID-19). These guidelines were formulated by an international panel of clinical experts, one methodologist and patients' representatives on behalf of the ESICM. The review was conducted in compliance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement recommendations. We followed the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) approach to assess the certainty of evidence and grade recommendations and the quality of reporting of each study based on the EQUATOR (Enhancing the QUAlity and Transparency Of health Research) network guidelines. The CPG addressed 21 questions and formulates 21 recommendations on the following domains: (1) definition; (2) phenotyping, and respiratory support strategies including (3) high-flow nasal cannula oxygen (HFNO); (4) non-invasive ventilation (NIV); (5) tidal volume setting; (6) positive end-expiratory pressure (PEEP) and recruitment maneuvers (RM); (7) prone positioning; (8) neuromuscular blockade, and (9) extracorporeal life support (ECLS). In addition, the CPG includes expert opinion on clinical practice and identifies the areas of future research.

Physiological and clinical variables identify ARDS classes and therapeutic heterogeneity to glucocorticoids: a retrospective study

OBJECTIVE

We aimed to identify new classes in acute respiratory distress syndrome (ARDS) using physiological and clinical variables and to explore heterogeneity in the effects of glucocorticoid therapy between classes.

METHODS

Using the Medical Information Mart for Intensive Care-IV database, we identified patients with ARDS. Potential profile analysis was used to identify classes with physiological and clinical data as delineating variables. Baseline characteristics and clinical outcomes were compared between classes. The effect of glucocorticoid treatment was explored by stratifying by class and glucocorticoid treatment.

RESULTS

From 2008 to 2019, 1104 patients with ARDS were enrolled in the study. The 2-class potential analysis model had the best fit (P < 0.0001), with 78% of patients falling into class 1 and 22% into class 2. Additional classes did not improve the model fit. Patients in class 2 had higher anion gap, lactate, creatinine, and glucose levels and lower residual base, blood pressure, and bicarbonate compared with class 1. In-hospital mortality and 28-day mortality were significantly higher among patients in class 2 than those in class 1 (P < 0.001). Heterogeneity of glucocorticoid treatment was observed, stratified by class and treatment, with no significant effect in class 1 (P = 0.496), increased mortality in class 2 (P = 0.001), and a significant interaction (P = 0.0381). In class 2, 28-day survival was significantly lower with glucocorticoid treatment compared with no hormone treatment (P = 0.001).

CONCLUSION

We used clinical and physiological variables to identify two classes of non-COVID-19-associated ARDS with different baseline characteristics and clinical outcomes. The response to glucocorticoid therapy varied among different classes of patients.

Research Progress on Phenotypic Classification of Acute Respiratory Distress Syndrome: A Narrative Review

Acute respiratory distress syndrome (ARDS) is a clinical syndrome that is characterized by an acute onset and refractory hypoxemia. It remains an important contributor to high mortality in critically ill patients, and the majority of clinical randomized controlled trials on ARDS provide underwhelming findings, which is attributed in large part to its pathophysiological and clinical heterogeneity, among other aspects. It is now widely accepted that ARDS is highly heterogeneous, growing evidences support this. ARDS phenotypic and subphenotypic studies aim to further differentiate and identify ARDS heterogeneity in the hope that clinicians can benefit from it, then can diagnose ARDS faster and more accurately and provide targeted treatments. This review collates and evaluates the major phenotype-related research advances of recent years, with a specific focus on ARDS biomarkers and clinical factors.

Mortality associated with acute respiratory distress syndrome, 2009-2019: a systematic review and meta-analysis

Background: Acute respiratory distress syndrome (ARDS) occurs commonly in intensive care units. The reported mortality rates in studies evaluating ARDS are highly variable. Objective: To investigate mortality rates due to ARDS from before the 2009 H1N1 influenza pandemic began until the start of coronavirus disease 2019 (COVID-19) pandemic. Design: We performed a systematic search and then ran a proportional meta-analysis for mortality. We ran our analysis in three ways: for randomised controlled trials only, for observational studies only, and for randomised controlled trials and observational studies combined. Data sources: MEDLINE and Embase, using a highly sensitive criterion and limiting the search to studies published from January 2009 to December 2019. Review methods: Two of us independently screened titles and abstracts to first identify studies and then complete full text reviews of selected studies. We assessed risk of bias using the Cochrane RoB-2 (a risk-of-bias tool for randomised trials) and the Cochrane ROBINS-1 (a risk-of-bias tool for non-randomised studies of interventions). Results: We screened 5844 citations, of which 102 fully met our inclusion criteria. These included 34 randomised controlled trials and 68 observational studies, with a total of 24 158 patients. The weighted pooled mortality rate for all 102 studies published from 2009 to 2019 was 39.4% (95% CI, 37.0-41.8%). Mortality was higher in observational studies compared with randomised controlled trials (41.8% [95% CI, 38.9-44.8%] v 34.5% [95% CI, 30.6-38.5%]; P = 0.005). Conclusions: Over the past decade, mortality rates due to ARDS were high. There is a clear distinction between mortality in observational studies and in randomised controlled trials. Future studies need to report mortality for different ARDS phenotypes and closely adhere to evidence-based medicine. PROSPERO registration: CRD42020149712 (April 2020).

Lessons learned in mechanical ventilation/oxygen support in COVID19

The clinical spectrum of severe acute respiratory syndrome coronavirus-2 infection ranges from asymptomatic infection or mild respiratory symptoms to pneumonia, with severe cases leading to acute respiratory distress syndrome with multiorgan involvement. The clinical management of patients with coronavirus disease 2019 (COVID-19) acute respiratory distress syndrome (ARDS) changed over the course of the pandemic, being adjusted as more evidence became available. This article will review how the ventilatory management of COVID-19 ARDS evolved and will conclude with current evidence-based recommendations.