Abstract

Introduction: Intensivists are routinely required to titrate alveolar minute ventilation based on blood gas information. Survey data show that intensive care trainees feel insufficiently trained in ventilator management. Bedside decision support has the potential to facilitate more efficient ventilator titration and serve as a teaching tool for trainees. Methods: We developed an online ventilator calculator based upon arterial partial pressure of carbon dioxide (PaCO2) and Henderson-Hasselbalch equations, allowing clinicians to input ventilator parameters as well as current and desired blood gas information. We validated predictions among arterial blood gas (ABG) values from a single quaternary academic pediatric intensive care unit (PICU). We included all ABG values obtained between 1/1/2010 and 12/31/2020 among PICU patients receiving invasive pressure control, volume control, or volume targeted ventilation and concurrent pharmacologic muscle relaxation with a repeat blood gas analysis within 1-24 hours. We excluded ABGs from patients on extracorporeal support. We subsequently performed a retrospective in-silico trial on a subset of ABGs meeting criteria for ventilator weaning (pCO2 < 30 and pH > 7.35 or pCO2 < 50 and pH > 7.50) or escalation (pCO2 > 50 and pH < 7.30 or pCO2 > 45 and pH < 7.20), excluding patients with traumatic brain injury. Clinician behavior was compared against calculator recommendations, with a goal of achieving an ABG value that met neither weaning nor escalation criteria. Results: There were 16,835 included ABGs among 664 patients, with a median (IQR) time between ABGs of 3.8 (2.2-4.4) hours. The median (IQR) difference between predicted and actual PaCO2 values was 0.00 (-3.06-3.00) mmHg. 11,460/16,835 (68.1%) PaCO2 values were within 5 mmHg of predicted, and 14,902/16,835 (88.5%) were within 10 mmHg. There were 2,433 ABGs among 350 patients included in the in-silico trial. Calculator recommendations were preferable to clinician behavior in 1,149/1,534 (74.9%) weaning scenarios and 605/899 (67.3%) escalation scenarios. Conclusions: Computer decision support outperformed clinician actions in this retrospective in-silico trial and may provide a useful teaching tool for trainees. Prospective study is needed to determine feasibility and user satisfaction.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call