We assessed the effect of the alert on the secondary outcome of time until SBP returning to >80 mmHg and subsequently remained ≥80 mmHg for at least 10 minutes using time-to-event analyses as in the primary outcome. For duration of hospitalization, the end point was the time to discharge alive, with patients censored (and considered nonevents) if they expired during the hospital stay and assigned the maximum observed length of hospital stay plus 1 day.
Sample Size Considerations
Pilot data were obtained on 6456 patients from Hillcrest Hospital from July 1, 2012, to December 31, 2012, for purposes of sample size calculation; n = 820 (13%) had at least 1 episode of SBP <80 mmHg for 3 minutes or more. The estimated median time to return to >80 mmHg was 6.1 minutes (95% CI, 5.5–6.8) and 25% and 75% of patients returned by 3.2 and 13.1 minutes, respectively. To have 90% power at the 0.05 significance level to detect a hazard ratio of 0.75 or stronger, a maximum of 711 patients per group was planned. We enrolled about twice that many patients over the 1-year accrual period.
The significance level for both the primary and the secondary outcome was 0.05. SAS statistical software (Cary, NC) was used for all analyses.
Between May 2012 and May 2013, 12,620 patients had 14,506 operations at Hillcrest Hospital. Invasive SBP <80 mmHg for 3 consecutive minutes or the most noninvasive SBP reading <80 mmHg was detected during 3955 surgeries. We excluded 558 operations because they were cardiac procedures (n = 97), cesarean deliveries (n = 382), minor procedures (n = 61), or missing BP records in the database (n =18). We also excluded 232 operations in patients who had previously participated in the trial. We thus included 3165 patients in our analysis: 1598 (50.5%) were randomly assigned to the DSS hypotensive alert and 1567 (49.5%) to nonalert (Fig. 1).
Most baseline (Table 1) and intraoperative (Table 2) characteristics were well balanced between the randomized groups, as indicated by an ASD <0.07, except for race, deficiency anemia, and surgical procedure. Blood pressure was measured solely oscillometrically in about 82% of patients, and 93% of alerts were triggered by oscillometric measurements (Appendices 1 and 2). Number of incorrect (approximately 4%) and missed (approximately 2%) alerts were similar between the randomized groups (Fig. 1; Appendix 2).
Being randomized to our hypotensive, DSS alert did not affect the primary outcome of time to return to SBP ≥80 mmHg after the first alert, with estimated adjusted hazard ratio of 0.99 (95% CI, 0.92–1.06; P = 0.69). The median time [25th, 75th percentiles] to return to SBP ≥80 mmHg was 1 [0, 3] minute in the alert group and 1 [0, 3] minute in the nonalert group (Table 3; Fig. 2). We found no interaction between randomized alert group and arterial line use on primary or secondary outcomes (i.e., all interaction P values >0.10).
The randomized groups did not differ on the secondary outcome of time for SBP to return to ≥80 mmHg and subsequently remain ≥80 mmHg for at least 10 minutes (adjusted hazard ratio [95% CI], 0.99 [0.92–1.06]; P = 0.76). Neither did the groups differ on the duration of hospitalization (adjusted hazard ratio [95% CI], 1.03 [0.96–1.11]; P = 0.35), with median length of hospital stay [25th, 75th percentiles] of 2 [1, 4] days in the alert group and 2 [1, 4] days in the nonalert group (Table 4).
Few clinicians would consider an SBP of 80 mmHg to be acceptable in adults having noncardiac surgery. In fact, even mean arterial pressures <80 mmHg are associated with risk-adjusted mortality.16 We thus assumed that alert clinicians would respond to even a single pressure <80 mmHg and use vasopressors, fluids, or other approaches to improve blood pressure. Our supplemental alerts were thus designed primarily for clinicians who failed to notice or respond to dangerously low blood pressures displayed on their workstations, perhaps by virtue of being distracted.
Serious hypotension (SBP <80 mmHg sustained for at least 3 consecutive minutes) or a single noninvasive blood pressure reading <80 mmHg was fairly common, even in a community hospital without resident anesthesiologists, with 27% (3955/14,506) of cases demonstrating at least 1 episode. That low pressures occurred in about a quarter of patients shows just how unstable anesthetized patients are and the extent to which minute-to-minute management is necessary during surgery.
There was, however, no benefit from our supplemental alerts. The time required to resolve hypotension once alert conditions were met was low (approximately 1 minute) and was not shortened by decision support alerts. The time required to return to a stable blood pressure (defined by SBP ≥80 mmHg sustained for 10 minutes) also did not differ in the alert and nonalert groups. Because there was no difference in blood pressures between the randomized groups, it is unsurprising that the durations of hospitalization were also comparable.
Our results contrast with those of Nair et al.1 who reported that AIMS-based near real-time decision support reduced the duration and frequency of hypotension in patients given >1.25 minimum alveolar concentration of a volatile anesthetic. However, there are important differences between the studies. Most obviously, Nair et al. used a before-and-after approach that potentially suffered from (1) unrelated time-dependent practice changes; (2) unrecognized confounding factors; and (3) the Hawthorne effect. Their approach was thus inherently weak compared with our randomized trial of >3000 patients. Our results are also consistent with a retrospective analysis by Epstein and Dexter,17 showing that desaturation resolves rapidly in routine clinical practice, presumably because clinicians intervene as necessary.
An important aspect of our study is that it was a trial of alert efficacy rather than a study of any particular blood pressure management approach. Our approach thus differed markedly from a conventional clinical trial in which response to hypotension would normally have been scripted per protocol. Our clinicians were entirely free to act on the decision support hypotension alerts, ignore them, or consider the provided information without acting on it as they considered appropriate. Furthermore, the alert suggested considering raising blood pressure but did not specify any particular approach; clinicians accepting the suggestion might thus do so by giving a vasopressor, reducing anesthetic administration, augmenting vascular volume, or putting the patient into Trendelenburg position—or using a combination of approaches.
That decision support alerts did not shorten the duration of hypotension in our study might thus result for several reasons: (1) the alerts did not provide additional useful information (e.g., the anesthesiologists were already aware of hypotension and doing their best to treat it); (2) the alerts provided additional information (say to a distracted anesthesiologist), but the responses were ineffectual; or (3) the alerts provided additional information that was ignored. Given the completely electronic approach to our trial, we cannot determine which of these possibilities dominated, although the first is obviously most attractive.
Various definitions of hypotension have been used in studies reporting associations between intraoperative blood pressure and adverse outcomes, but an SBP <80 mmHg is a common choice.11 Furthermore, we have shown that a mean arterial pressure <55 mmHg (corresponding roughly to an SBP of 80 mmHg) is strongly associated with acute kidney injury and myocardial infarction.14 We thus considered SBP to be critically low when SBP was <80 mmHg. It is possible that alerts would be more effective at another threshold.
Arterial pressure, by nature, varies over time, and invasive pressures are especially subject to short-term artifact. We thus designed the study to require 3 consecutive minutes with median invasive systolic pressures <80 mmHg (evaluated at 1-minute intervals) before triggering the decision support alerts. In contrast, we triggered alerts with a single oscillometric SBP <80 mmHg because in many cases another value would not be available within 3 minutes. When both invasive and oscillometric pressures were available, the lowest value was used to determine alert triggering. In fact, 93% of the alerts were triggered by oscillometric pressures, meaning that our results largely apply to single oscillometric values.
Our study was conducted in a high-acuity hospital that restricts anesthetic practice to board-certified attending anesthesiologists working alone or with nurse anesthetists. This remains by far the most common practice pattern in the United States, making our results generalizable. It remains possible that alerts for hypotension would be more effective for residents who are perhaps more likely to be distracted or fail to rapidly intervene.
Large randomized trials are generally considered the highest level of clinical evidence. However, large trials are usually expensive and often require prolonged enrollment periods. Consequently, the number of conventional major trials is tiny compared with the number of important clinical questions and probably always will be. Our study design was innovative, and perhaps unique, in using an inexpensive, entirely electronic approach that allowed us to enroll >3000 patients in a single moderate-sized hospital in just a year.
The results of this particular trial were negative, and some might say predictably so, because anesthesiologists should not require an additional alert for serious hypotension. However, we note that it is easy to develop various alerts and that alerts and alarms for various conditions are increasingly incorporated into anesthesia information systems. Many address similarly “obvious” patient responses.
An important conclusion we draw is that various alerts being incorporated into clinical systems should not be assumed to be beneficial. They may even be harmful by inducing alarm fatigue or diverting clinicians from more important issues, both of which are increasingly likely as the number of alerts increases. Using methodology similar to ours, the value of proposed clinical alerts can be formally tested inexpensively and quickly. And they should be!
The methodology we used can also be extended to important questions where the results are certainly nonobvious. For example, our method could be used to evaluate an intraoperative handover checklist or a subtle combination of patient responses that would be difficult for clinicians to otherwise continuously monitor, such as triple lows of mean arterial pressure, minimum alveolar concentration, and bispectral index.18
In summary, using decision support to detect low SBP and generate an additional warning did not reduce the duration of intraoperative hypotension or length of hospital stay. An additional warning for severe hypotension thus did not reduce the duration of hypotension or hospitalization. Decision support alarms may be more useful for more complicated situations or when the alarm event is triggered by a combination of physiologic variables.
Appendix 1. Summary of Arterial Line Usage for SBP
Appendix 2. Evaluation of DSS Algorithm
Name: Krit Panjasawatwong, MD.
Contribution: This author helped write the manuscript.
Attestation: Krit Panjasawatwong reviewed the analysis of the data and approved the final manuscript.
Conflicts: Krit Panjasawatwong reported no conflicts of interest.
Name: Daniel I. Sessler, MD.
Contribution: This author helped design the study, conduct the study, and write the manuscript.
Attestation: Daniel I. Sessler reviewed the analysis of the data and approved the final manuscript.
Conflicts: Daniel I. Sessler reported no conflicts of interest.
Name: Wolf H. Stapelfeldt, MD.
Contribution: This author helped design the study, conduct the study, write the manuscript, and Designer of Clinical Decision Support System.
Attestation: Wolf H. Stapelfeldt has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.
Conflicts: Wolf H. Stapelfeldt worked for Talis Clinical, LLC, and received royalties from Talis Clinical, LLC. He is the inventor of the Clinical Decision Support system used for this study (eligible for receiving royalties). He is also a founder, officer (CMO), and member of the Board of Directors of the company operating this decision support system (Talis Clinical, LLC).
Name: Douglas B. Mayers, MD, PhD.
Contribution: This author helped conduct the study and write the manuscript.
Attestation: Douglas B. Mayers has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.
Conflicts: Douglas B. Mayers reported no conflicts of interest.
Name: Edward J. Mascha, PhD.
Contribution: This author helped design the study, analyze the data, and write the manuscript.
Attestation: Edward J. Mascha has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.
Conflicts: Edward J. Mascha reported no conflicts of interest.
Name: Dongsheng Yang, MS.
Contribution: This author helped analyze the data and write the manuscript.
Attestation: Dongsheng Yang has seen the original study data, reviewed the analysis of the data, and approved the final manuscript.
Conflicts: Dongsheng Yang reported no conflicts of interest.
Name: Andrea Kurz, MD.
Contribution: This author helped design the study, conduct the study, and write the manuscript.
Attestation: Andrea Kurz reviewed the analysis of the data, approved the final manuscript, and is the author responsible for archiving the study files.
Conflicts: Andrea Kurz reported no conflicts of interest.
This manuscript was handled by: Maxime Cannesson, MD, PhD.
a Elixhauser A, Steiner C, Palmer L. Clinical Classifications Software (CCS), 2014. U.S. Agency for Healthcare Research and Quality. Available at: http://www.hcup-us.ahrq.gov/toolssoftware/ccs/ccs.jsp. Accessed June 18, 2015.
1. Nair BG, Horibe M, Newman SF, Wu WY, Peterson GN, Schwid HA. Anesthesia information management system-based near real-time decision support to manage intraoperative hypotension and hypertension. Anesth Analg. 2014;118:206–14
2. Nair BG, Horibe M, Newman SF, Wu WY, Schwid HA. Near real-time notification of gaps in cuff blood pressure recordings for improved patient monitoring. J Clin Monit Comput. 2013;27:265–71
3. Nair BG, Peterson GN, Newman SF, Wu WY, Kolios-Morris V, Schwid HA. Improving documentation of a beta-blocker quality measure through an anesthesia information management system and real-time notification of documentation errors. Jt Comm J Qual Patient Saf. 2012;38:283–8
4. Nair BG, Peterson GN, Neradilek MB, Newman SF, Huang EY, Schwid HA. Reducing wastage of inhalation anesthetics using real-time decision support to notify of excessive fresh gas flow. Anesthesiology. 2013;118:874–84
5. Nair BG, Newman SF, Peterson GN, Wu WY, Schwid HA. Feedback mechanisms including real-time electronic alerts to achieve near 100% timely prophylactic antibiotic administration in surgical cases. Anesth Analg. 2010;111:1293–300
6. Sandberg WS, Sandberg EH, Seim AR, Anupama S, Ehrenfeld JM, Spring SF, Walsh JL. Real-time checking of electronic anesthesia records for documentation errors and automatically text messaging clinicians improves quality of documentation. Anesth Analg. 2008;106:192–201
7. Wax DB, Beilin Y, Levin M, Chadha N, Krol M, Reich DL. The effect of an interactive visual reminder in an anesthesia information management system on timeliness of prophylactic antibiotic administration. Anesth Analg. 2007;104:1462–6
8. Monk TG, Saini V, Weldon BC, Sigl JC. Anesthetic management and one-year mortality after noncardiac surgery. Anesth Analg. 2005;100:4–10
9. Aronson S, Phillips-Bute B, Stafford-Smith M, Fontes M, Gaca J, Mathew JP, Newman MF. The association of postcardiac surgery acute kidney injury with intraoperative systolic blood pressure hypotension. Anesthesiol Res Pract. 2013;2013:174091
10. Morris RW, Watterson LM, Westhorpe RN, Webb RK. Crisis management during anaesthesia: hypotension. Qual Saf Health Care. 2005;14:e11
11. Bijker JB, van Klei WA, Kappen TH, van Wolfswinkel L, Moons KG, Kalkman CJ. Incidence of intraoperative hypotension as a function of the chosen definition: literature definitions applied to a retrospective cohort using automated data collection. Anesthesiology. 2007;107:213–20
12. Bijker JB, van Klei WA, Vergouwe Y, Eleveld DJ, van Wolfswinkel L, Moons KG, Kalkman CJ. Intraoperative hypotension and 1-year mortality after noncardiac surgery. Anesthesiology. 2009;111:1217–26
13. Bijker JB, Persoon S, Peelen LM, Moons KG, Kalkman CJ, Kappelle LJ, van Klei WA. Intraoperative hypotension and perioperative ischemic stroke after general surgery: a nested case-control study. Anesthesiology. 2012;116:658–64
14. Walsh M, Devereaux PJ, Garg AX, Kurz A, Turan A, Rodseth RN, Cywinski J, Thabane L, Sessler DI. Relationship between intraoperative mean arterial pressure and clinical outcomes after noncardiac surgery: toward an empirical definition of hypotension. Anesthesiology. 2013;119:507–15
15. Austin PC. Balance diagnostics for comparing the distribution of baseline covariates between treatment groups in propensity-score matched samples. Stat Med. 2009;28:3083–107
16. Mascha EJ, Yang D, Weiss S, Sessler DI. Intraoperative mean arterial pressure variability and 30-day mortality in patients having noncardiac surgery. Anesthesiology. 2015;123:79–91
17. Epstein RH, Dexter F. Implications of resolved hypoxemia on the utility of desaturation alerts sent from an anesthesia decision support system to supervising anesthesiologists. Anesth Analg. 2012;115:929–33
© 2015 International Anesthesia Research Society
18. Sessler DI, Sigl JC, Kelley SD, Chamoun NG, Manberg PJ, Saager L, Kurz A, Greenwald S. Hospital stay and mortality are increased in patients having a “triple low” of low blood pressure, low bispectral index, and low minimum alveolar concentration of volatile anesthesia. Anesthesiology. 2012;116:1195–203