Implications of resolved hypoxemia on the utility of desaturation alerts sent from an anesthesia decision support system to supervising anesthesiologists.

BACKGROUND

Hypoxemia (oxygen saturation <90%) lasting 2 or more minutes occurs in 6.8% of adult patients undergoing noncardiac anesthesia in operating room settings. Alarm management functionality can be added to decision support systems (DSS) to send text alerts about vital signs outside specified thresholds, using data in anesthesia information management systems. We considered enhancing our DSS to send hypoxemia alerts to the text pagers of supervising anesthesiologists. As part of a voluntary application for an investigative device exemption from our IRB to implement such functionality, we evaluated the maximum potential utility of such an alert system.

METHODS

Pulse oximetry values (Spo(2)) were extracted from our anesthesia information management systems for all cases performed in our main operating rooms and ambulatory surgical center between September 1, 2011, and February 4, 2012 (n = 16,870). Hypoxemic episodes (Spo(2) < 90%) were characterized as either (a) lasting one or more minutes or (b) lasting 2 or more minutes. A single simulated "alert" was modeled as having been sent at the timestamp of the first (a) or the second (b) hypoxemic value. The hypoxemic episode was considered resolved at 1, 3, or 5 minutes after the time of the alert if the Spo(2) value was no longer below the 90% threshold. Two-sided 99% conservative confidence limits were calculated for the percentage of unresolved alerts at the 3 evaluation intervals and compared with 70%, the lower limit of an acceptable true alarm rate for clinical utility.

RESULTS

There was at least 1 hypoxemic episode lasting 1 minute or longer in 23% of cases, and at least 1 episode lasting 2 minutes or longer in 8% of cases. Only 7% (99% confidence interval [CI] 6% to 8%) of the 1-minute hypoxemic episodes were unresolved after 3 minutes, and only 8% (99% CI 6%to 9%) of 2-minute episodes after 5 minutes (both P < 10(-6) in comparison with 70% minimum reliability rate).

CONCLUSIONS

Low utility should be expected for a DSS sending hypoxemia alerts to supervising anesthesiologists, because nearly all hypoxemic episodes will have been resolved before arrival of the anesthesiologist in the operating room. These results suggest that the principal research focus should be on developing more sophisticated alerts and processes within rooms for the anesthesia care provider to initiate treatment promptly, to interpret or correct artifacts, and to make it easier to call for assistance via a rapid communication system.