Previous Abstract | Next Abstract
Printable Version
A922
October 18, 2010
9:00:00 AM - 11:00:00 AM
Room Hall B1-Area G
Modeling the Impact of Alert Thresholds of 'Triple Low' Conditions on 90-Day Postoperative Mortality
  *  Wolf H. Stapelfeldt, M.D., Scott D. Greenwald, Ph.D., Paul J. Manberg, Ph.D., Nassib Chamoun, M.S., Daniel I. Sessler, M.D.
Anesthesiology Institute, Cleveland Clinic, Cleveland, Ohio; Technology, Research and Clinical Development, Covidien, Mansfield, Massachusetts
We previously demonstrated that simultaneous intraoperative presence of low BIS, low MAC and low MAP ([start_en]0027;Triple Low') was an independent predictor of poor postoperative outcome; moreover, vasopressor (VP) administration within 5 minutes of entering a 'Triple Low' state may improve outcome. In preparation for implementing real-time clinical decision support alerts we sought to identify optimal trigger parameters (threshold values) for potentially improving 90-day mortality.

Methods: With IRB approval, BIS, MAP, and end-tidal volatile anesthetic concentrations (in MAC equivalents: Iso=1.17; Sevo=1.8; Des=6.6) were extracted from our registry. A total of 23,999 non-cardiac patients were classified into 1 of 4 groups based on whether or not patients experienced a 'Triple Low' state for >=1 minutes and whether or not ephedrine or phenylephrine (VP) were given at all, were given early ([start_en]003C; 5 minutes) or late ([start_en]003E;= 5 minutes) after entering a 'Triple Low' state. 90-day mortality was risk-adjusted with a Cox Proportional Hazards model which included age, gender, race, BMI, ASA physical status, and a composite derived from ICD-9 codes. The modeled consequence of responding to alert conditions for various threshold combinations was assessed by estimating the impact of early VP administration. Metrics estimated were: 1) fraction of patients receiving an alert; 2) risk-adjusted mortality rate for each group; 3) the number of potential lives saved by early VP administration; and 4) the 'alert efficiency' (i.e., the number of patients expected to receive an alert for each life potentially saved).

Results: Based on an overall risk-adjusted mortality of 2.3% the table below tabulates modeled performance metrics for a range of 'Triple Low' alert threshold conditions (BIS: 40-45, MAP:70-80, MAC: 0.70-0.90):[table1]The threshold combination previously reported (MAP=75, BIS=45, MAC=0.70) provided a modeled 67 additional lives potentially saved with 30% of patients receiving at least 1 alert, for an overall efficiency of 108 patients alerted per potential additional life saved. The threshold combination offering greatest efficiency (MAP=75, BIS=40,MAC=0.90) yielded a modeled 93 potentially additional lives saved with 27% of patients receiving an alert, for an overall efficiency of 70 patients alerted per potentially additional life saved.

Conclusion: Modeled relative benefit of early VP administration following '[start_en]0027;Triple Low' state detection varied in magnitude and efficiency as a function of the various MAP, BIS and MAC alert threshold combinations tested.

From Proceedings of the 2010 Annual Meeting of the American Society Anesthesiologists.
Risk-Adjusted 90-Day Mortality
MAP70707575758080
BIS40454040454045
MAC0.70.70.70.90.80.90.8
No Alert2.22.22.22.22.02.22.0
Had Alert2.82.72.72.52.62.42.6
No Alert & No VP2.42.22.32.32.12.42.0
No Alert & Had VP1.91.81.92.01.91.91.9
Had Alert & Early VP1.41.60.90.81.41.11.6
Had Alert & No/Late VP3.33.13.23.12.92.72.8
Percent Patients with Alert (%)14231927373445
Potential Lives Saved (in 23999)33416293909284
Number of Patients with Alerts per Potential Life Saved10113174709888128