A model to simulate human cardio-respiratory responses to airway obstruction

Xin Jin^1,2Varghese Kurian^1,2Kathy L. Ryan³Anders Wallqvist¹Jaques Reifman^1*Sridevi Nagaraja^1,2

• ¹Department of Defense Biotechnology High Performance Computing Software Applications Institute, Defense Health Agency Research and Development, Medical Research and Development Command, Fort Detrick, MD, United States
• ²Henry M Jackson Foundation for the Advancement of Military Medicine Inc, Bethesda, United States
• ³US Army Institute of Surgical Research, Joint Base San Antonio Fort Sam Houston, United States

Airway obstruction is the second leading cause of potentially survivable death on the battlefield. Managing airway obstruction resulting from severe traumatic injuries to the head and neck, which can distort the airway anatomy, poses significant challenges to combat medics. The medic's ability to make quick and effective interventions to secure the airway in austere, tactical environments is also highly dependent on their training and experience as well as the availability of advanced medical equipment. Artificial intelligence (AI) algorithms can help augment the competency and capability of medics to care for combat casualties by enhancing their training, assessing their skills, and helping identify the most appropriate medical interventions in real time that are likely to result in desired clinical outcomes. However, the training and assessment of AI algorithms require massive amounts of real-world, vital-sign data. Because such data are not currently available for casualties with airway obstruction, an alternative approach is to rely on relevant synthetic data generated by high-fidelity computational physiological models. Here, by adding new respiratory control and respiratory mechanics components, we extended our previously developed and validated human cardio-respiratory (CR) model for representing hemorrhagic injury to account for the physiological effects of airway obstruction on vital signs. We calibrated and validated the extended CR model using data from six human studies and two pig studies, which reported vital-sign changes in response to airway obstruction, changes in arterial oxygen (O2) and carbon dioxide (CO2) pressure, changes in the fraction of inspired O2 and CO2, and hemorrhage followed by ventilation changes. On average, the extended CR model achieved good prediction accuracy, with root mean square errors of 1.3 L/min for minute ventilation, 1.6 breaths/min for respiratory rate, 4.5% for oxygen saturation, 3.5 mmHg for end-tidal CO₂, 11.6 mmHg for systolic blood pressure, 7.8 mmHg for mean arterial pressure, and 9.8 beats/min for heart rate. With this enhancement, the extended CR model can now be used to generate realistic synthetic trauma datasets for the two leading causes of potentially survivable battlefield deaths, hemorrhagic injury and airway obstruction, and help develop AI decision-support tools for combat medics.