An Electronic Health Record-Based Classifier for Moderate-to-Severe Acute Respiratory Distress Syndrome With...

Area of Science:

• Critical Care Medicine
• Health Informatics
• Pulmonary Medicine

Background:

• Persistent moderate-to-severe acute respiratory distress syndrome (ARDS) presents diagnostic challenges.
• Accurate identification is crucial for timely intervention and research.

Purpose of the Study:

• To evaluate the performance of an electronic health record (EHR)-based classifier for identifying persistent moderate-to-severe ARDS.
• To assess the utility of this classifier in clinical settings.

Main Methods:

• Retrospective observational study in six ICUs across two health systems.
• Included adults on invasive mechanical ventilation (IMV) meeting specific hypoxemia criteria.
• Evaluated classifier performance on development and validation cohorts using EHR data and clinician adjudication.

Main Results:

• The EHR classifier identified persistent moderate-to-severe ARDS with acceptable performance in both cohorts (PPV 71% and 66%).
• Alternative definitions using SpO2/FiO2 showed similar classification but lower positive predictive values.
• Clinician-adjudicated ARDS was present in 55% and 53% of patients in the development and validation cohorts, respectively.

Conclusions:

• An EHR-based classifier using readily available data demonstrates acceptable performance for identifying persistent moderate-to-severe ARDS.
• This open-source tool offers potential for retrospective and prospective identification of ARDS patients for research and quality improvement.