Extreme value theory applied to postoperative breathing patterns

Abstract

Background. There has been little published work on the statistical features of breath times in postoperative patients. We applied extreme value theory (a statistical method) to the variation in the timing of postoperative breathing. Methods. We observed 49 patients 3-6 h after a variety of surgical procedures, once they had achieved a stable breathing pattern. The breathing patterns could be one of the three types predicted by the extreme value model. 'Finite' breathing patterns (n=30) have a finite upper limit of duration for any apnoea. Patients that displayed one of the other two patterns ('standard' and 'extended') have, potentially, no limit in duration of apnoea. Results. The type of breathing pattern observed in each patient was not reliably identified by most of the commonly used risk factors (age, type of surgery, opioid type, dose, and route of administration). A finite pattern was observed in 13 of 26 patients receiving epidural (vs 17 of 23 parenteral analgesia: P=0.15), and 15 of 19 receiving morphine (vs 15 of 30 other opioids: P=0.05). The patients with 'finite' patterns were also significantly less drowsy (score 1.04 (0.92) vs 1.62 (0.62), P<0.05). Conclusions. The breathing pattern was not related to mean breath times, suggesting that the prevalence of apnoeas cannot be reliably predicted by measurement of the respiratory rate alone.