Developing A Data Pipeline to Quantify Ventilator Waveforms.

Area of Science:

• Critical Care Medicine
• Biomedical Engineering
• Data Science

Background:

• Understanding complex interactions in mechanical ventilation is limited by data availability.
• Existing datasets lack high-fidelity waveform data integrated with comprehensive electronic health records.

Purpose of the Study:

• To develop an automated pipeline for collecting high-fidelity ventilator waveforms.
• To integrate this data with extensive electronic health record (EHR) data.
• To create a robust dataset for analyzing mechanical ventilation in critically ill patients.

Main Methods:

• Prospective, observational study design.
• Multidisciplinary team (data scientists, engineers, clinicians).
• Automated data pipeline for ventilator waveforms and EHR data integration.

Main Results:

• Collected data from 1,116 patients, generating over 13 ventilator-years of waveform data.
• Analyzed 146 million breaths, with 49 million fitting a single-compartment model.
• Integrated EHR data, providing extensive patient records for each waveform data point.

Conclusions:

• A fully automated pipeline was created to collect mechanical ventilation waveform data.
• This high-fidelity dataset integrates EHR data, crucial for studying ventilation mechanics.
• The dataset enables deeper understanding of lung injury, patient effort, and ventilator dyssynchrony.