The Science Journal of the American Association for Respiratory Care

1995 OPEN FORUM Abstracts

INTELLIGENT CO2 MONITORING

Jigish D. Trivedi M.S., George Thomsen M.D., Jeff Anderson M.E., Thomas D. East Ph.D. Pulmonary Division, LDS Hospital and the University Of Utah, Salt Lake City, Utah 84143

Introduction : Reliable non-invasive methods to determine PaCO2 are not presently available. End-tidal partial pressure of CO2(PetCO2) is only useful as monitor PaCO2 in patients with little gas exchange derangement. The difference between PetCO2 and PaCO2 may be quite large and trends can be misleading. The purpose of this work was to create a dependable non-invasive monitor of PaCO2 for use with mechanical ventilation.

Methods: A Macintosh IIci computer was used to measure the airway pressure, gas flow and expiratory timing signals from a Siemens 900C ventilator, and the partial pressure of expired CO2 from a Novametrix CO2 monitor. These variables were sampled at 100 Hz. PetCO2, the slope of phase III of the capnogram, VCO2, tidal volume, and ventilatory rate were calculated on a breath-by-breath basis and used as input parameters to a mathematical model of CO2 uptake, distribution and elimination. With each breath, the model predicted values for all of the input parameters were compared to actual values and the differences (errors) used in a Bayesian modified Chi-Squared parameter estimation to predict what the values should be for PaCO2 for each lung compartment. Essentially this system provided a non-invasive estimate of PaCO2 for each of the two compartments. The accuracy and precision of the data acquisition system and calculated parameters was tested by using precision CO2 flows (Tylan Flow Controllers) into a mechanical lung model (Michigan Instruments). To test the feasibility of the concept, we collected at least two hours of data, from five patients with the Adult Respiratory Distress Syndrome. PaCO2 values from arterial blood gases were compared with estimated PaCO2 values predicted by our system.

Results: The laboratory testing of the data collection and analysis software showed accuracy of 0.4% and precision of 0.6% in measurement of VCO2. Four patients had relatively stable PaCO2 during the data collection period. The system was able to very accurately predict PaCO2 under these circumstances (error less than 5%). One patient had a long record of fluctuating PaCO2 which was adequate to challenge the feasibility of the system. In this patient, 14 ABG samples were collected during 48 hours. For ABG samples obtained during controlled mechanical ventilation, the error between measured and predicted PaCO2 was 1.9 ± 2.2 mmHg (mean ± SD). For ABG samples obtained during assisted ventilation the error was -6.8 ± 17.8 mm Hg. Conclusion : A computerized system designed provide non-invasive estimates of PaCO2 in a two compartment model was developed and the system was shown to be accurate and precise in a carefully controlled laboratory setting. A small clinical feasibility study indicates that the system can accurately predict PaCO2; however, accuracy is influenced by noise in measurement of CO2 and flow. This effect is more pronounced in patients who are assisting. In the future, digital signal processing techniques must be used to extract representative values from the "clinically noisy" environment prior to making estimates of PaCO2. With such improvements this system should be capable of providing robust non-invasive predictions of PaCO2.

SEE ORIGINAL IMAGE

Acknowledgment: We sincerely thank Siemens Life Support Systems, Solna, Sweden and Deseret Foundation, Salt Lake City, UT for their support.

OF-95-185

You are here: RCJournal.com » Past OPEN FORUM Abstracts » 1995 Abstracts » INTELLIGENT CO2 MONITORING