Daily Ards Research Analysis
Machine learning using routine EHR data accurately classified pediatric ARDS subphenotypes, potentially enabling early precision care without biomarker assays. A meta-analysis suggests prone positioning during VV-ECMO improves early survival and gas exchange but not longer-term survival. An international PICU cohort showed PARDS is common in intubated LRTI infants and is associated with higher mortality and prolonged ventilation.
Summary
Machine learning using routine EHR data accurately classified pediatric ARDS subphenotypes, potentially enabling early precision care without biomarker assays. A meta-analysis suggests prone positioning during VV-ECMO improves early survival and gas exchange but not longer-term survival. An international PICU cohort showed PARDS is common in intubated LRTI infants and is associated with higher mortality and prolonged ventilation.
Research Themes
- Precision phenotyping in ARDS using EHR and machine learning
- Adjunctive strategies during VV-ECMO (prone positioning)
- Epidemiology and outcomes of pediatric ARDS in LRTI
Selected Articles
1. Development and Validation of an Electronic Health Record-Based, Pediatric Acute Respiratory Distress Syndrome Subphenotype Classifier Model.
Using two temporally distinct PARDS cohorts, the authors trained an EHR-only machine learning model that accurately classified biomarker-defined hyperinflammatory vs hypoinflammatory subphenotypes. A parsimonious 5-laboratory-variable model maintained high performance (AUC ~0.92) at 24 hours after diagnosis, suggesting feasibility for early precision stratification without biomarker assays.
Impact: This work bridges biomarker-defined biology and point-of-care data, enabling early subphenotyping that could guide targeted therapies and trial enrollment in PARDS.
Clinical Implications: Clinicians could use EHR-derived lab values within 24 hours to identify hyperinflammatory PARDS, informing escalation, adjunctive therapies, and enrollment into phenotype-specific trials.
Key Findings
- EHR-only classifier achieved AUC 0.93 (95% CI 0.87–0.98) with 88% sensitivity and 83% specificity for hyperinflammatory PARDS.
- A parsimonious model using five laboratory variables achieved AUC 0.92 (95% CI 0.86–0.98) with 76% sensitivity and 87% specificity.
- Subphenotypes were defined via biomarker-guided latent class analysis, providing a biological reference standard.
- Classification feasible at 24 hours after PARDS diagnosis, supporting early risk stratification.
Methodological Strengths
- Temporal external validation in a separate cohort
- Biomarker-guided latent class analysis as a biologically grounded reference
- Feature selection to derive a parsimonious, clinically implementable model
- Performance reported with AUC, sensitivity, specificity, and 95% confidence intervals
Limitations
- Single-center study limits generalizability
- Retrospective design with potential unmeasured confounding
- External validation was temporal but not multicenter
- Clinical impact on outcomes not tested prospectively
Future Directions: Multicenter, prospective validation and testing of phenotype-guided therapy or trial stratification using the parsimonious model.
OBJECTIVE: To determine if hyperinflammatory and hypoinflammatory pediatric acute respiratory distress syndrome (PARDS) subphenotypes defined using serum biomarkers can be determined solely from electronic health record (EHR) data using machine learning. DESIGN: Retrospective, exploratory analysis using data from 2014 to 2022. SETTING: Single-center quaternary care PICU. PATIENTS: Two temporally distinct cohorts of PARDS patients, 2014-2019 and 2019-2022. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: Patients in the derivation cohort ( n = 333) were assigned to hyperinflammatory or hypoinflammatory subphenotypes using biomarkers and latent class analysis. A machine learning model was trained on 165 EHR-derived variables to identify subphenotypes. The most important variables were selected for inclusion in a parsimonious model. The model was validated in a separate cohort ( n = 114). The EHR-based classifier achieved an area under the receiver operating characteristic curve (AUC) of 0.93 (95% CI, 0.87-0.98), with a sensitivity of 88% and specificity of 83% for determining hyperinflammatory PARDS. The parsimonious model, using only five laboratory values, achieved an AUC of 0.92 (95% CI, 0.86-0.98) with a sensitivity of 76% and specificity of 87% in the validation cohort. CONCLUSIONS: This proof-of-concept study demonstrates that biomarker-based PARDS subphenotypes can be identified using EHR data at 24 hours of PARDS diagnosis. Further validation in larger, multicenter cohorts is needed to confirm the clinical utility of this approach.
2. Efficacy and safety of prone positioning in patients undergoing extracorporeal membrane oxygenation (ECMO): A systematic review and meta-analysis.
Across 17 studies of ARDS patients on VV-ECMO, prone positioning improved 30-day and in-hospital survival and enhanced oxygenation/PaCO2, but did not improve 60- or 90-day survival, ICU survival, or ECMO weaning. Earlier and more frequent proning shortened mechanical ventilation and ICU stays, with greater early survival benefit in non-COVID cohorts.
Impact: This synthesis informs a common bedside dilemma—whether to add proning during ECMO—by demonstrating early survival and physiologic benefits but limited longer-term effects.
Clinical Implications: Consider early and repeated prone positioning during VV-ECMO to improve oxygenation and early survival, especially in non-COVID ARDS, while counseling that long-term survival benefits remain uncertain.
Key Findings
- ECMO+prone positioning improved 30-day and in-hospital survival.
- No significant benefit for 60-day, 90-day, ICU survival, or ECMO weaning rates.
- Significant improvements in oxygenation and reductions in PaCO2 with ECMO+proning.
- Earlier and more frequent proning sessions shortened mechanical ventilation and ICU length of stay; non-COVID cohorts derived greater early survival benefit.
Methodological Strengths
- PRISMA-compliant systematic review and meta-analysis
- Multiple databases searched (MEDLINE, Embase, Cochrane Library)
- Clinically meaningful outcomes and subgroup insights (COVID vs non-COVID)
Limitations
- Predominantly observational studies with risk of residual confounding
- Heterogeneity in proning protocols, timing, and patient populations
- Potential publication bias and lack of randomized trials
Future Directions: Prospective randomized trials to quantify causal effects of proning during ECMO, optimal timing/frequency, and identification of subgroups most likely to benefit.
BACKGROUND: Acute respiratory distress syndrome (ARDS) is a severe lung condition characterized by diffuse alveolar damage and hypoxemia. Venovenous extracorporeal membrane oxygenation (vv-ECMO) supports gas exchange and reduces ventilator-induced injury, while prone positioning (PP) improves oxygenation by optimizing ventilation-perfusion matching. METHODS: We conducted a systematic review and meta-analysis, following PRISMA guidelines, using MEDLINE, Embase, and the Cochrane Library. Studies included were on adult ARDS patients undergoing vv-ECMO with PP. Outcomes measured were survival rates, ECMO weaning, duration of ECMO support, mechanical ventilation, ICU and hospital stays, and complications. RESULTS: Seventeen studies met inclusion criteria. While ECMO+PP improved 30-day and hospital survival rates, there was no significant improvement in 60-day survival, 90-day survival, ICU survival, or ECMO weaning rates. ECMO+PP significantly enhanced oxygenation parameters and reduced PaCO2 levels. Earlier and more frequent PP sessions shortened mechanical ventilation and ICU stays. Non-COVID patients had better 30-day survival with ECMO+PP than COVID patients. CONCLUSIONS: Combining PP with ECMO improves early but not long-term survival in severe ARDS, especially in non-COVID patients, while extending duration of mechanical support and stays. Further prospective randomized trials are needed to confirm these findings.
3. Acute Respiratory Distress Syndrome in Children With Lower Respiratory Tract Infection Requiring Invasive Mechanical Ventilation: Post Hoc Analysis of the 2019-2020 Bronchiolitis and Codetection Cohort.
In 571 intubated infants with acute LRTI across 48 PICUs, 42% met PARDS criteria on day 1, with higher mortality, longer invasive ventilation, and prolonged PICU stays. Competing risk modeling suggested day-1 PARDS causally prolongs ventilation, with 7-day extubation probability 49% vs 64% without PARDS.
Impact: This multicenter analysis quantifies the burden and prognostic impact of PARDS in intubated LRTI, informing resource planning and trial design in pediatric critical care.
Clinical Implications: Early recognition of PARDS in intubated LRTI infants is crucial; anticipate longer ventilation and PICU stay and consider tailored strategies to mitigate risks.
Key Findings
- Day-1 PARDS prevalence was 42% among 571 intubated LRTI infants across 48 PICUs.
- PARDS was associated with higher mortality (7.9% vs 2.7%; p=0.023).
- PARDS increased invasive ventilation duration (median 165 vs 135 hours; p<0.001) and PICU length of stay (11 vs 8 days; p<0.001).
- Competing risk modeling showed 7-day extubation probability 49% with PARDS vs 64% without PARDS.
Methodological Strengths
- International, multicenter prospective cohort as data source
- Standardized day-1 PARDS criteria application
- Multivariable competing risk modeling for extubation probability
Limitations
- Post hoc analysis may introduce bias and residual confounding
- Limited to children <2 years with LRTI; generalizability to older children or other etiologies is uncertain
- PARDS status assessed on day 1 only; trajectory phenotypes not evaluated
Future Directions: Prospective studies to test targeted management strategies for PARDS in bronchiolitis/LRTI and to evaluate trajectory-based phenotypes.
OBJECTIVES: Bronchiolitis and other lower respiratory tract infections (LRTIs) are the most common causes of pediatric respiratory failure. There is insufficient evidence characterizing pediatric acute respiratory distress syndrome (PARDS) in young children with LRTI to inform clinical management. We aimed to describe the prevalence and clinical characteristics of children intubated for LRTI and meeting PARDS criteria. DESIGN: We performed a post hoc analysis of data from the Bronchiolitis And COdetectioN (BACON) study, an international prospective observational study of critical bronchiolitis. We compared PARDS subjects (meeting criteria the first full calendar day following intubation) to non-PARDS subjects. SETTING: Forty-eight international PICUs recruiting to the BACON study, from December 2019 to November 2020. PATIENTS: Children younger than 2 years old, requiring mechanical ventilation for acute LRTI. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: Complete data were available for 571 children. Day 1 PARDS was diagnosed in 240 subjects(42%) and associated with increased mortality (7.9% vs. 2.7%; p = 0.023), greater duration of invasive ventilation (165 hr [interquartile range, 112-251 hr] vs. 135 hr [76-204 hr]; p < 0.001), and PICU length of stay (11 d [7-16 d] vs. 8 d [5-13 d]; p < 0.001). In our multivariable competing risk model, the presence of PARDS on day 1 was causally related to a prolonged duration of mechanical ventilation with the probability of extubation at 7 days for those with PARDS equal to 49% (44-54%) compared with 64% (59-69%) for those without PARDS. CONCLUSIONS: PARDS development was common in this critical bronchiolitis cohort, resulted in a longer duration of mechanical ventilation, and was associated with increased mortality and PICU length of stay. Prospective studies are needed to elucidate the optimal management of critical bronchiolitis.