Virtual patient model for evaluating automated inspired oxygen control.

Area of Science:

• Critical care medicine
• Biomedical engineering
• Computational physiology

Background:

• Physiologic closed-loop control (PCLC) for inspired oxygen (FiO2) requires safety validation.
• Clinical trials may not capture worst-case scenarios for PCLC algorithm assessment.
• Understanding performance sensitivity to physiologic variations is crucial for safe patient care.

Purpose of the Study:

• To develop and validate a virtual patient model for assessing PCLC safety and efficacy.
• To evaluate FiO2 control using a PCLC algorithm under diverse and extreme conditions.
• To identify factors influencing PCLC performance and potential failure modes.

Main Methods:

• A large-scale virtual observational study (3,780,000 simulated patients) was conducted.
• Simulations incorporated clinically relevant scenarios with step changes in patient condition, target SpO2, and PCLC activation.
• A uniform sampling approach evaluated controller performance in worst-case scenarios.

Main Results:

• PCLC significantly reduced the duration and severity of desaturation, preventing it in many cases (e.g., 69.8% to 1.5% occurrence).
• Performance sensitivity was highest for physiologic delays between arterial and peripheral oxygenation.
• Longer delays and positive SpO2 sensor bias increased the likelihood of system oscillations.

Conclusions:

• Virtual patient modeling is a valuable tool for rigorous safety and efficacy evaluation of autonomous medical devices.
• PCLC demonstrates effectiveness in managing oxygenation but requires careful consideration of physiologic delays and sensor accuracy.
• This approach complements clinical trials by exploring underrepresented extreme conditions for enhanced medical device assessment.