Ability of the integrated pulmonary index to predict impending respiratory events in the early postoperative period

Abstract

Abstract Background In the early postoperative period, respiratory compromise is a significant problem. Standard-of-care monitoring includes respiratory rate (RR) and pulse oximetry, which are helpful; however, low SpO2 is often a late sign during decompensation. The FDA-approved Capnostream-20p monitor records four variables (SpO2, RR, End-tidal CO2, heart rate), which are combined by fuzzy logic into a single, unit-less value (range 1–10) called the integrated pulmonary index (IPI). No published studies have assessed the performance of a low IPI to predict impending respiratory events. Methods In this investigator-initiated study, adult patients undergoing general anesthesia were monitored with the Capnostream-20p monitor for up to 2 h during their recovery room stay. The study coordinator, who along with clinicians, was blinded to IPI values, recorded the time of any respiratory event, defined a priori as any one of eight respiratory-related interventions/conditions. The primary sensitivity endpoint (early detection success) was defined as at least 80% of events predicted by at least 2 consecutive low IPI (≤ 7) values within 2–15 min before an event occurred. Late detection was defined as low IPI values occurring with 2 min prior to or 2 min after the event occurred. Discussion Of 358 patients, ≥ 1 respiratory event occurred in 183 (51.1%) patients. Of 802 total events, 606 were detected early (within 2–15 min prior to the event), and 653 were detected either early or late. Therefore, the sensitivity for early detection was 75.6% (95% confidence interval [CI]: 72.6–78.5%), which differed significantly from the 80% sensitivity goal by 4.4% (p = 0.0016). Sensitivity for total success (early or late) was 81.4% (95% CI: 78.7–84.1%), which was significantly different from the 90% on time sensitivity goal by 8.6% (p < 0.0001). Conclusions A low IPI was 75.6% sensitive for early detection (within 2–15 min) prior to respiratory events but did not achieve our preset threshold of 80% for success.