Phenylephrine Infusion and Real-Time Alerts
We explored whether the providers learned from SAM notifications to run phenylephrine infusions in a manner designed to avoid alerts. Specifically, mean phenylephrine infusion rate before and after the intervention was compared (Table 2) and found similar (P = 0.491 [CI, 0.019 to −0.009], 2-sample t test). Additionally, no time-trend in mean phenylephrine infusion rate during the postintervention phase was found. To examine whether the providers backdated the phenylephrine infusion rate in the AIMS to avoid alerts, we analyzed the documentation patterns of phenylephrine infusion before and after intervention. No difference was found in the proportion of times providers backdated phenylephrine infusion before (89/543 = 16%) and after (116/670 = 17%) intervention (P = 0.700, Fisher exact test).
Prolonged intraoperative hypotension and hypertension are associated with negative clinical outcome. In this study, we detected 2 scenarios of hypotension and hypertension in near real-time and brought them to the attention of the anesthesia provider via pop-up computer screens. The specific scenarios include hypotension with concurrently high inhaled drug MAC and hypotension with continuing phenylephrine infusion. Notification messages did decrease the instances of prolonged episodes (≥6 minutes) of both Hypotension-high MAC and Hypertension-phenylephrine infusion incidents. Also, notification messages for hypotensive episodes with concurrent high MAC decreased the mean duration of such episodes. However, notification of hypertensive episodes with concurrent phenylephrine infusion did not decrease the mean duration of such episodes. Therefore, the overall efficacy of near real-time notifications was less pronounced for Hypertension-phenylephrine infusion events than with Hypotension-high MAC events. The provider response to both Hypotension-high MAC and Hypertension-phenylephrine infusion events improved with notification messages. For Hypotension-high MAC events, the providers reduced the anesthetic concentration to below 1.25 MAC close to 80% of the time after notification messages were initiated, an increase of 20% from before notifications. Though a similar percentage improvement for correct provider response was also noticed for hypertension-phenylephrine infusion events, the overall compliance to correct provider response was less consistent (37%) even after notification messages.
The goal of this study was to reduce hypotensive and hypertensive episodes through near real-time notification of contributing factors, with the hypothesis that near real-time notifications can elicit near real-time response from anesthesia providers to take actions to mitigate hypotension and hypertension. We used the common definitions of hypotension as systolic blood pressure <80 mm Hg and hypertension as systolic blood pressure higher than 160 mm Hg. As a future enhancement to our rules, we may consider thresholds customized to a patient or case. The median duration of > 6-minute hypotension-high MAC episodes reduced with SAM notifications, while that of hypertension-phenylephrine episodes did not show a statistically significant reduction. Perhaps what may be more clinically significant is the reduction in the frequency of > 6-minute episodes of hypotension-high MAC and hypertension-phenylephrine episodes. Also, it was observed that the maximum duration of hypotensive episodes reduced from 44 to 15 minutes while that of hypertensive episodes reduced from 30 to 21 minutes. Reduction of such prolonged episodes could potentially minimize the risks of intraoperative hypotension and hypertension: myocardial ischemia, infarction, stroke and prolonged hospital stay. Though we did not investigate whether a reduction in unsafe blood pressure episodes actually translated into better patent outcome, near real-time notifications seem to have had a positive impact toward improved intraoperative patient management.
There may not be a single reason why providers did not respond as expected: 19% of the time for hypotension-high MAC episodes and the majority of the time (63%) for hypertension-phenylephrine episodes, even after notification messages were generated. Because messages are displayed on the AIMS screen for 2 minutes (unless the user acknowledges the message), in some instances the providers may not have seen the message if they were busy with patient care. The SAM system did not keep an audit log whether the provider acknowledged a message or not. Thus, it was not possible to determine whether the provider did not take an action after acknowledging a message or whether the provider did not see the message. In many instances, when cuff blood pressure was measured episodically, the notification was based on a single blood pressure measurement being either hypotensive or hypertensive. Some providers may have felt that a single measurement was not clinically relevant to initiate a corrective action and may have waited for a subsequent blood pressure measurement by which time the episode may have ended. Additionally, there may have also been instances when hypertension and hypotension may have had to be deliberatively maintained, in which case the providers may have simply ignored the notification. The reason why the hypertension-phenylephrine alerts were ignored a large number of times could also be related to the relevance of a SAM message that was displayed. A SAM alert recommended stopping the phenylephrine infusion during hypertension-phenylephrine episodes, when in fact most providers felt reducing the phenylephrine infusion rate was a more appropriate intervention (Table 6). We have subsequently changed the SAM message to remove this recommendation and simply highlight the presence of a hypertension-phenylephrine episode.
The 2 blood pressure management issues chosen for the study were deliberate. For the hypotension-high MAC issue, both the systolic blood pressure and the inhaled drug concentration data were automatically acquired by the AIMS and thus by SAM. However, for the hypertension-phenylephrine issue, only systolic blood pressure was automatically acquired, while the information related to phenylephrine infusion (the time of dose adjustment and the actual dose) was entered manually by the anesthesia provider. In our AIMS, as is the case with most other AIMS, manual entry of information often occurs retrospectively. This explains why the SAM messages were not generated in some cases (Table 5) when retrospective analysis revealed that there were episodes of hypertension with continuing phenylephrine infusion. On closely analyzing the audit log of the AIMS, we were able to determine that in such instances the users actually entered the phenylephrine infusion information after the hypertension episode has ended. Conversely, we also found instances where a SAM notification message was generated, but we could not find a corresponding hypertension-phenylephrine episode during retrospective analysis. In such cases, the anesthesia provider had edited the phenylephrine infusion information retrospectively such that it was documented as being stopped before the hypertension episode.
SAM alerts did not seem to have influenced the manner in which providers administered and documented phenylephrine infusion. This may be due to the fact that SAM alerts related to hypotension and hypertension are infrequent (<1% of cases), and the anesthesia personnel changes at our institution are fairly frequent. However, the number of cases that had phenylephrine infusion was higher during the postintervention period when compared with preintervention (Table 2). The reasons for this observation are not quite clear. It was also noticed that hypotensive events decreased during the postintervention period. It could be that providers used more phenylephrine to avoid hypotension. Additionally, we noticed that there were slightly more elderly patients (>60 years) during the postintervention period that may also have influenced the use of more phenylephrine. Yet another influencing factor could be the changes in anesthesia personnel during the course of the study and the associated differences in their anesthetic management. It is possible that SAM notifications could also have indirectly contributed to the increased use of phenylephrine. However, as SAM hypotension messages are infrequent and are directed toward reduction of inhaled drug MAC, this possibility seems unlikely.
The SAM decision support program is an add-on module separate from AIMS. This constrains SAM to perform conservatively without negatively impacting the performance of AIMS. Additionally, its sampling of AIMS data is limited by how fast the AIMS database refreshes its data from the AIMS workstations. For the AIMS installed in our hospital, the data refresh rate is variable and can be as slow as every 5 minutes. To accommodate these considerations, SAM was configured to sample the AIMS database every 6 minutes. A 6-minute sampling period meant that hypotension-high MAC and hypertension-phenylephrine episodes lasting less that 6 minutes could not be reliably detected (Table 5) as they could occur within the sampling period without being detected. Additionally, a variable refresh rate of the AIMS meant that data in the AIMS database could be as old as 5 minutes. For this reason, even hypotension/hypertension episodes of up to 11 minutes (6 minutes sampling period + 5-minute data refresh period) are sometimes not detected by SAM (Table 5).
Real-time detection of continuing patient safety issues is where AIMS-based decision support will have the most significant impact on patient care. In this study, we attempted to detect and notify 2 unsafe clinical scenarios: hypotension with a high concentration of inhaled drugs and hypertension with continuing phenylephrine infusion. Though near real-time detection and notification of these scenarios decreased the episodes of these scenarios lasting >6 minutes, near-instantaneous detection and notification are possible only by increasing the data sampling rate and decreasing data latencies in the system.20 It is suggested that the decision support module be integrated as part of an AIMS rather than function as an add-on module. This can promote faster sampling of an AIMS database without risking system failure. Data latency can be reduced by more frequent refreshing of an AIMS database with data from the AIMS workstation. Alternately, for AIMS configuration that follows a thick client model (majority of the computing occurs on the client workstation), the time-critical decision support logic can run on the data collected by the client module (or OR workstation) instead of waiting for these data to reach the server database. Our study also revealed that the efficacy of real-time detection and notification of unsafe events is dependent on how data are acquired in an AIMS. Automated acquisition of relevant data meant that detection of hypotension-high inhaled drug concentration events were more effective when compared with hypertension-phenylephrine events for which phenylephrine infusion data were manually entered. For effective clinical decision support, all data relevant for decision support should be automatically acquired to eliminate the possibility of retrospective data entry or delayed data corrections. Integration with infusion pumps and other monitoring equipment, bar-coded data entry of medication, and blood products are mechanisms to increase automated data entry into an AIMS. Increased automation and frequency of data acquisition in an AIMS ensure real-time availability of a larger information source, which in turn can lead to effective and near-instantaneous clinical decision support.
Name: Bala G. Nair, PhD.
Contribution: The author is the primary investigator. The author is the primary individual who performed design and conduct of the study, data analysis, and manuscript preparation.
Attestation: Bala G. Nair approved the final manuscript, also attests to the integrity of the original data and the analysis reported in this manuscript, and is the archival author.
Name: Mayumi Horibe, MD.
Contribution: The author is the coinvestigator. The author is the secondary individual who performed data preparation, analysis, and manuscript preparation.
Attestation: Mayumi Horibe approved the final manuscript and also attests to the integrity of the original data and the analysis reported in this manuscript.
Name: Shu-Fang Newman, MS.
Contribution: The author is the coinvestigator and Smart Anesthesia Manager programmer.
Attestation: Shu-Fang Newman approved the final manuscript.
Name: Wei-Ying Wu, PhD.
Contribution: The author helped design and analyze the statistical study and assisted prepare the manuscript.
Attestation: Wei-Ying Wu approved the final manuscript and also attests to the integrity of the statistical analysis reported in this manuscript.
Name: Gene N. Peterson, MD, PhD.
Contribution: The author is the clinical advisor for research study and manuscript preparation.
Attestation: Gene N. Peterson approved the final manuscript.
Name: Howard A. Schwid, MD.
Contribution: The author is the clinical advisor for decision support rules, research study, and manuscript preparation.
Attestation: Howard A. Schwid approved the final manuscript.
This manuscript was handled by: Franklin Dexter, MD, PhD.
1. Bijker JB, Persoon S, Peelen LM, Moons KGM, Kalkman CJ, Kappelle JL, van Klei WA. Intraoperative hypotension and perioperative ischemic stroke after general surgery. Anesthesiology. 2012;116:658–64
2. Sharma D, Brown MJ, Curry P, Noda S, Chesnut RM, Vavilala MS. Prevalence and risk factors for intraoperative hypotension during craniotomy for traumatic brain injury. J Neurosurg Anesthesiol. 2012;24:178–84
3. Frank M, Radtke FK, Prahs C, Seeling M, Papkalla N, Wernecke KD, Spies CD. Documented intraoperative hypotesion according to the three most common definitions does not match the application of antihypotensive Medication. J International Med Res. 2011;39:846–56
4. 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
5. Bijker JB, van Klei WA, Kappen TH, van Wolfswinkel L, Moons KGM, Kalkman CJ. Incidence of intraoperative hypotension as a function of the chosen definition. Anesthesiology. 2007;107:213–20
6. Lien SF, Bisognano JD. Perioperative hypertension: defining at-risk patients and their management. Curr Hypertens Rep. 2012;14:432–41
7. Reich DL, Bennet-Guerrero E, Bodian CA, Hossain S, Winfree W, Krol M. Intraoperative tachycardia and hypertension are independently associated with adverse outcome in noncardiac surgey of long duration. Anesth Analg. 2002;95:273–7
8. Reich DL, Bodian CA, Krol M, Kuroda M, Osinski T, Thys DM. Intraoperative hemodynamic predictors of mortality, stroke, and myocardial infarction after coronary artery bypass surgery. Anesth Analg. 1999;89:814–22
9. 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
10. 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
11. Nair BG, Newman SF, Peterson GN, Schwid HA. Smart Anesthesia Manager™ (SAM)–a real-time decision support system for anesthesia care during surgery. IEEE Trans Biomed Eng. 2013;60:207–10
12. Nair BG, Peterson GN, Newman S, Wu W, Kolios-Morris V, Schwid HA. Improved documentation of β-blocker quality measure through anesthesia information management system and real-time notification of documentation errors. Jt Comm J Qual Saf. 2012;38:283–8
13. Rothman B, Leonard JC, Vigoda MM. Future of electronic health records: implications for decision support. Mt Sinai J Med. 2012;79:757–68
14. 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
15. Ehrenfeld JM, Epstein RH, Bader S, Kheterpal S, Sandberg WS. Automatic notifications mediated by anesthesia information management systems reduce the frequency of prolonged gaps in blood pressure documentation. Anesth Analg. 2011;113:356–63
16. Chau A, Ehrenfeld JM. Using real-time clinical decision support to improve performance on perioperative quality and process measures. Anesthesiol Clin. 2011;29:57–69
17. Wanderer JP, Sandberg WS, Ehrenfeld JM. Real-time alerts and reminders using information systems. Anesthesiol Clin. 2011;29:389–96
18. Nair BG, Newman SF, Peterson GN, Schwid HA. Automated electronic reminders to improve redosing of antibiotics during surgical cases: comparison of two approaches. Surg Infect (Larchmt). 2011;12:57–63
19. 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
20. Epstein RH, Dexter F, Ehrenfeld JM, Sandberg WS. Implications of event entry latency on anesthesia information management decision support systems. Anesth Analg. 2009;108:941–7
21. Dexter F, Epstein RH, Lee JD, Ledolter J. Automatic updating of times remaining in surgical cases using bayesian analysis of historical case duration data and “instant messaging” updates from anesthesia providers. Anesth Analg. 2009;108:929–40
22. 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
23. Kheterpal S, Gupta R, Blum JM, Tremper KK, O’Reilly M, Kazanjian PE. Electronic reminders improve procedure documentation compliance and professional fee reimbursement. Anesth Analg. 2007;104:592–7
24. Epstein RH, Dexter F, Piotrowski E. Automated correction of room location errors in anesthesia information management systems. Anesth Analg. 2008;107:965–71
25. Spring SF, Sandberg WS, Anupama S, Walsh JL, Driscoll WD, Raines DE. Automated documentation error detection and notification improves anesthesia billing performance. Anesthesiology. 2007;106:157–63
26. 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
© 2014 International Anesthesia Research Society
27. Newcombe RG. Interval estimation for the difference between independent proportions: comparison of eleven methods. Stat Med. 1998;17:873–90