Daily Ards Research Analysis
Three studies advanced ARDS science today: a multicenter prospective cohort links delayed intubation and noninvasive ventilation to slower long-term lung recovery in COVID-19 ARDS; a machine-learning model accurately identifies ARDS across health systems using EHR and radiology text; and preclinical work reveals gut bacterial lactate perturbs lung epithelial mitochondria to exacerbate lung injury.
Summary
Three studies advanced ARDS science today: a multicenter prospective cohort links delayed intubation and noninvasive ventilation to slower long-term lung recovery in COVID-19 ARDS; a machine-learning model accurately identifies ARDS across health systems using EHR and radiology text; and preclinical work reveals gut bacterial lactate perturbs lung epithelial mitochondria to exacerbate lung injury.
Research Themes
- Intubation timing and ventilatory modality shape long-term recovery in COVID-19 ARDS
- EHR- and NLP-enabled machine learning for cross-system ARDS identification
- Gut–lung axis: bacterial lactate drives mitochondrial dysregulation in lung epithelium
Selected Articles
1. Longitudinal recovery trajectories and ventilatory modalities in COVID-19 acute respiratory distress syndrome survivors.
In a 52-ICU prospective cohort of 1,854 severe COVID-19 survivors, late intubation was associated with significantly worse percentage-predicted pulmonary function during follow-up, and patients managed with noninvasive mechanical ventilation (NIMV) showed slower lung recovery. Modalities were distributed as HFNC 19.4%, NIMV 15.6%, and IMV 64.9% (early IMV 966 vs late IMV 238).
Impact: This large, multicenter prospective study directly links ventilatory strategy and intubation timing to long-term pulmonary recovery after COVID-19 ARDS, informing post-ICU outcomes and practice patterns.
Clinical Implications: Consider avoiding delayed intubation in decompensating COVID-19 ARDS and exercise caution with prolonged NIMV when lung recovery is a priority. Plan structured pulmonary follow-up (e.g., DLCO) at 3–12 months.
Key Findings
- Among 1,854 ICU patients, HFNC, NIMV, and IMV were used in 19.4%, 15.6%, and 64.9%, respectively.
- Late intubation (n=238) was associated with significantly worse percentage-predicted pulmonary function during follow-up compared with early intubation (n=966).
- Patients initially treated with NIMV exhibited slower lung recovery over time.
Methodological Strengths
- Multicenter, prospective design across 52 ICUs
- Standardized longitudinal follow-up at 3, 6, and 12 months with DLCO measurement
Limitations
- Observational design with potential confounding by indication for ventilatory modality and intubation timing
- Study period limited to early pandemic waves; practice patterns and variants may have evolved
Future Directions: Pragmatic trials and causal inference analyses on intubation timing; standardized NIMV protocols with early failure criteria; integration of pulmonary rehabilitation pathways and biomarker-guided recovery.
BACKGROUND: The impact of different ventilatory support modalities and timing of intubation on longitudinal lung recovery trajectories in patients with severe coronavirus disease 2019 (COVID-19) is unknown. METHODS: This was a multicentre, prospective observational study conducted in 52 Spanish intensive care units (ICUs) involving critically ill COVID-19 patients admitted between 25 February 2020 and 8 February 2021. 1854 COVID-19 patients were followed after hospital discharge at 3, 6 and 12 months with diffusing capacity of the lung for carbon monoxide ( RESULTS: A total of 360 (19.4%) and 290 (15.6%) patients received HFNC and NIMV, respectively. 1204 (64.9%) patients underwent IMV; 966 received early IMV and 238 received late IMV. The latter exhibited a significantly worse percentage predicted CONCLUSIONS: Delay in intubation implies the worst outcomes; however, patients with NIMV exhibited a slower lung recovery in terms of
2. Gut bacterial lactate stimulates lung epithelial mitochondria and exacerbates acute lung injury.
Using gnotobiotic mice, lung epithelial cell assays, and ARDS patient metabolomics, the authors show that gut bacterial metabolites—specifically lactate—stimulate mitochondrial activity in lung epithelium and worsen acute lung injury. Colonization of germ-free mice induced lung mitochondrial gene programs, linking the gut–lung axis to ARDS pathobiology.
Impact: This work uncovers a mechanistic gut–lung axis in ARDS via bacterial lactate and mitochondrial reprogramming, offering tractable metabolic and microbiome targets.
Clinical Implications: While preclinical, findings suggest testing microbiome modulation or lactate transport inhibition to mitigate lung injury. Metabolic profiling may identify at-risk patients with microbiome-driven mitochondrial dysregulation.
Key Findings
- Gut microbiota colonization of germ-free mice increased lung mitochondrial gene expression.
- Gut bacterial metabolites, notably lactate, stimulated mitochondrial activity in lung epithelial cells.
- Re-analysis of a large ARDS metabolomics dataset supported links between bacterial metabolites and lung injury severity.
Methodological Strengths
- Integrated multi-system approach (gnotobiotic mice, cell culture, and human metabolomics)
- Mechanistic focus on mitochondrial function linking microbiome metabolites to lung injury
Limitations
- Preprint not peer-reviewed; details on specific taxa and causal human evidence are limited in the abstract
- Translational relevance and therapeutic strategies require validation in clinical studies
Future Directions: Identify causative taxa and metabolite fluxes, test lactate transport/metabolism inhibitors in vivo, and validate microbiome–mitochondria signatures in prospective ARDS cohorts.
Acute respiratory distress syndrome (ARDS) is an often fatal critical illness where lung epithelial injury leads to intrapulmonary fluid accumulation. ARDS became widespread during the COVID-19 pandemic, motivating a renewed effort to understand the complex etiology of this disease. Rigorous prior work has implicated lung endothelial and epithelial injury in response to an insult such as bacterial infection; however, the impact of microorganisms found in other organs on ARDS remains unclear. Here, we use a combination of gnotobiotic mice, cell culture experiments, and re-analyses of a large metabolomics dataset from ARDS patients to reveal that gut bacteria impact lung cellular respiration by releasing metabolites that alter mitochondrial activity in lung epithelium. Colonization of germ-free mice with a complex gut microbiota stimulated lung mitochondrial gene expression. A single human gut bacterial species,
3. Development and External Validation of a Detection Model to Retrospectively Identify Patients With Acute Respiratory Distress Syndrome.
A regularized logistic regression model combining structured EHR data and radiology reports retrospectively identified ARDS with AUROC 0.91 internally and 0.88 externally, calibrated with external ICI 0.13. At a chosen threshold, sensitivity and specificity were both 80% and PPV 64%, identifying cases a median of 2.2 hours after meeting Berlin criteria across two health systems.
Impact: Provides a validated, cross-system approach to ARDS case identification using routinely collected data, enabling registries, quality metrics, and scalable research curation.
Clinical Implications: Hospitals can deploy retrospective ARDS detection for case ascertainment, benchmarking, and research screening; future prospective integration could support earlier recognition and standardized phenotyping.
Key Findings
- Training, internal validation, and external validation cohorts included 1,845, 556, and 199 patients; ARDS prevalence was 19%, 17%, and 31%, respectively.
- EHR-radiology model achieved AUROC 0.91 (internal) and 0.88 (external) with external ICI 0.13.
- At a set threshold, sensitivity and specificity were 80% and PPV 64%; cases were identified a median 2.2 hours after Berlin criteria.
Methodological Strengths
- Physician-adjudicated ARDS labels and external validation across two health systems
- Model calibration reported (ICI) alongside discrimination metrics
Limitations
- Retrospective design; performance depends on data availability and documentation quality
- Generalizability beyond two health systems and portability of NLP for radiology may vary
Future Directions: Prospective, real-time deployment to assess impact on early recognition and outcomes; transportability studies; open sharing of code and definitions to enhance reproducibility.
OBJECTIVE: The aim of this study was to develop and externally validate a machine-learning model that retrospectively identifies patients with acute respiratory distress syndrome (acute respiratory distress syndrome [ARDS]) using electronic health record (EHR) data. DESIGN: In this retrospective cohort study, ARDS was identified via physician-adjudication in three cohorts of patients with hypoxemic respiratory failure (training, internal validation, and external validation). Machine-learning models were trained to classify ARDS using vital signs, respiratory support, laboratory data, medications, chest radiology reports, and clinical notes. The best-performing models were assessed and internally and externally validated using the area under receiver-operating curve (AUROC), area under precision-recall curve, integrated calibration index (ICI), sensitivity, specificity, positive predictive value (PPV), and ARDS timing. PATIENTS: Patients with hypoxemic respiratory failure undergoing mechanical ventilation within two distinct health systems. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: There were 1,845 patients in the training cohort, 556 in the internal validation cohort, and 199 in the external validation cohort. ARDS prevalence was 19%, 17%, and 31%, respectively. Regularized logistic regression models analyzing structured data (EHR model) and structured data and radiology reports (EHR-radiology model) had the best performance. During internal and external validation, the EHR-radiology model had AUROC of 0.91 (95% CI, 0.88-0.93) and 0.88 (95% CI, 0.87-0.93), respectively. Externally, the ICI was 0.13 (95% CI, 0.08-0.18). At a specified model threshold, sensitivity and specificity were 80% (95% CI, 75%-98%), PPV was 64% (95% CI, 58%-71%), and the model identified patients with a median of 2.2 hours (interquartile range 0.2-18.6) after meeting Berlin ARDS criteria. CONCLUSIONS: Machine-learning models analyzing EHR data can retrospectively identify patients with ARDS across different institutions.