Vice Chair of Research Cleveland Clinic Foundation
Disclosure information not submitted.
Introduction: ARDS carries significant morbidity and mortality. Early classification using ventilator-derived metrics—PaO₂/FiO₂ (P/F), P/FP, Oxygenation Index (OI), OSI (Oxygen Saturation Index), Mechanical Power (MP), Ventilatory Ratio (VR), and Driving Pressure (DP)—is vital for prognostication and research. Manual data abstraction is time-consuming. We implemented a ventilator dashboard-based capture process with an automated calculator for real-time ARDS severity classification and ventilator analytics in < 5 minutes per patient. This abstract reports the method’s completeness, efficiency, and clinical correlations.
Methods: We analyzed ventilator data from 54 ICU patient-days involving cases of ARDS or acute hypoxic respiratory failure due to other causes. Ventilator data were captured directly from a ventilator dashboard, including FiO₂, PaO₂, PEEP, plateau pressure, tidal volume, respiratory rate, PaCO₂, P/F, and S/F ratio. Derived calculations (P/FP, OI, OSI, MP, VR, MP, DP) were automated. Time required for data entry was recorded and compared to historical manual chart abstraction (~20 minutes/patient). ARDS severity was defined using P/F ratio.
Results: We evaluated 54 patient-days of ventilator data. Completeness: P/F (46%), S/F (52%), Combined P/F or S/F (96%), OI (41%), VR (40%), MP (42%), DP (30%). ARDS severity by P/F or S/F ratio: Mild (16), Moderate (23), Severe (14), Unknown (1). Dashboard-based capture saved 900 minutes vs. manual chart abstraction. Manual review had higher Day 1 P/F availability (93%) but dropped to 61.5% by Day 3. In contrast, our dashboard maintained 96% oxygenation classification using P/F or S/F with consistent performance. Completeness for derived variables via dashboard vs. manual charting—DP (30% vs. 26%) and MP (42% vs. 87.6%)—was comparable. OSI was calculated in 44% using our method but was not captured manually. Our method was also over 4 times faster per patient.
Conclusions: This pilot demonstrates the feasibility of a rapid ventilator data tool for ARDS prognostication. With high completeness and substantial time savings, this model is scalable and facilitates early risk stratification and research. Broader implementation and EHR integration may standardize ventilator phenotyping across centers.
