Abstract
Physiological monitoring in the operating room is needed to follow the patient's state of ventilation, circulation, etc. Parameters such as heart rate, blood pressure, and respiratory gas content are observed with devices of uncertain reliability. These provide speedy information in the form of cautions and alarms, which may indicate that corrective action is needed. In practice the large number of alarms when there is no hazard to the patient (false alarms) is a considerable problem. We describe a method of comparing and validating instrument readings in this situation involving a knowledge base whose core is a set of 36 rules. This was applied to 7803 warnings (6287 cautions and 1516 alarms) from 68 day surgery patients undergoing 115 hours of surgery. Most of the cautions were validated by our analysis, but 734 of the 1516 alarms were invalidated while 419 were validated and 363 left indeterminate. This translates to a potential reduction from one alarm every 4 minutes to one every 16 minutes.