Anesthesiology may be “widely recognized as the pioneering leader in patient safety efforts” but not because of medication safety.1 The Anesthesia Patient Safety Foundation, the Closed Claims Project, and the myriad engineering improvements to the anesthesia machine in the past 30+ years are all laudable. Pioneering work in critical incident analysis in the 1970s led to many improvements to prevent airway- and ventilation-related mishaps—the leading cause of morbidity at the time.2 Monitoring patients (with pulse oximetry and capnography), managing airways (with difficult airway algorithms and video laryngoscopes), and simulation training are impressive advances. However, despite being identified as a significant source of adverse events, the other side of the anesthesia workspace—intravenous (IV) medications—has proven more challenging to improve due to medications that are less tolerant of mistakes than inhaled agents and a more complex provider workflow. Consequently, the anesthesia approach to IV agents remains remarkably primitive, fragmented, and cavalier:
- We do not know the true prevalence of medication errors.
- Dangerous myths persist in our specialty.
- Little prevents providers from making basic errors.
- An entrenched resistance to standardization prevents advances.
Single-point failures—rarely tolerated on the anesthesia machine—are rampant with IV medications, and most error prevention relies on brittle strategies like “provider vigilance” and “reading the label.” Recommendations as recent as this past year still include stipulations like “syringes should be labeled” (with some exceptions)—an indication of how low expectations are in the medication sphere.3 If a bottle of volatile anesthetic was unlabeled, we would never administer it to a patient. An opportunity exists to learn from advances on the anesthesia machine, question assumptions about the IV medication space, and reaffirm our specialty as pioneers in patient safety.
Estimates of medication errors during anesthesia vary in orders of magnitude from <1% of anesthetics to >5% of medication administrations.4–6 Some of this variation can be explained by data gathering methods, but the real issue is the error reporting problem. Active monitoring strategies are not scalable: observers are expensive and cameras are legally problematic. Passive strategies suffer from all the biases and limitations of retrospective self-reporting, and fear of retribution combined with a cultural reluctance to report “minor” errors seriously only exacerbates the problem. As a result, we remain in the dark as a specialty about how often errors occur, and we certainly do not know if recent improvement efforts have had a significant impact on overall safety.
Other industries like transportation and nuclear power with robust reporting mechanisms expect a 10–100:1 ratio of near misses to adverse events.7 Our own self-reported data at Seattle Children’s revealed a much lower 2:6:1 ratio of near misses to medication errors to harm events, while the prospective study by Nanji et al6 found a 1:2:1 ratio. Most errors go unreported; many errors likely go unnoticed. Anesthetized patients—with constant supervision and mechanical ventilation—are remarkably tolerant of many mistakes. Few institutions have reliable methods for collecting medication errors beyond the small minority that result in outcomes poor enough to inspire a formal review. Fewer still have any ability to capture near misses—a critical step toward a more preventive approach to safety design.
A series of dangerous myths persist in our specialty regarding some basic elements of human factors:
- Removing colors will force providers to read labels more carefully and consistently.
- Hand-written labels are reliable in high-stress environments.
- Prefilled syringes have questionable safety benefits.
Lack of medical literature leaves these questions open for debate, but the human factors and interaction design literature indicate that our patients would benefit from dispelling these myths. Even the Institute for Safe Medication Practices vacillates between endorsing and cautioning against color coding, but most of their recommendations are based on anecdotal case reports and expert opinion. We often cite advances in aviation as inspiration for anesthetic improvements, but nobody in aviation encourages removing all colors from a cockpit to enforce closer scrutiny of instruments. While color does have to be used wisely (in a restrained manner with specific attention to meaning) in anesthesia advocates remain of wholesale removal of color from medications in the hopes that it will force users to read labels more carefully. This debate continues despite zero evidence that colorless labels do in fact inspire closer inspection of syringes or, more importantly, reduce errors.8 , 9
Anesthesia work shapes a work setting for distributed cognition. Distributed cognition comprises tasks that require the processing of information that is distributed across the internal mind and the external world.10 The design of the work setting plays a critical role in the safety of operations—designed artifacts provide cognitive support. The world of interaction design already knows that the best way to identify objects during stressful situations is to provide multiple cues—not to remove them. First, visual organization can greatly support the location and identification of critical elements by providing visual patterns for search and recall. Visual organization applies to the shape, color, scale, placement, and orientation of visual elements. By engaging visual memory, hierarchical layout and visual representation are far superior over label indexing and functional descriptions.11 Better representations lead to better understanding and decision making—compare the organization of a diagram with the ambiguity of names listed in a phone book. Second, the visual organization of text as layout greatly enhances the scanability of text, critical during fast-paced work flows. Additional discriminators such as color and textual emphasis (line weights, bolding of text, size contrasts) support the grouping of discrete elements for the processing and organization of information elements in meaningful groups.10 Better yet, adding other tactile and auditory cues takes advantage of multimodal information structures, which is why many oxygen flowmeter knobs are both color coded and have a distinctive tactile form. Syringes, on the other hand, all feel the same.
The aviation industry has entire design guides and studies on the merits and nuances of typography and the design of flight operations that are guided by the design of the flight deck as workspace. The locations and coding of flight information displays and controls are standardized across all aircraft, so that pilots can allocate critical control elements without reading the labels—reading that is not possible during fast-paced response to emergencies. In anesthesia—in contrast—we allow providers to hand scrawl on poorly labeled syringes drawn up from nonstandard, look-alike vials in a distracting environment and organize them in an arbitrary, personalized arrangement. Instead of removing our tenuous safety tools, we need to add more. Syringes should come with more identifiers that extend beyond the syringe itself to the surrounding environment.
We still debate the merits of prefilled syringes despite their superior labeling and ability to eliminate an entire step of the medication cycle (from vial to syringe) along with all of the associated potential errors. A failure modes and effects analysis of the anesthesia medication cycle revealed that the use of prefilled syringes could eliminate 16 medication preparation substeps along with 19 potential-associated failure modes.12 Borrowing these formal methods for evaluating complex tasks in anesthesia can help to guide meaningful change in the medication handling process.
There are no alarms on the medication side of the anesthesia workspace for bolus medications. There is no way to detect errors beyond provider vigilance. Even infusion pumps have little to prevent errant programming beyond generic dosing guardrails. The SEDASYS propofol infusion system for endoscopy (Johnson & Johnson, New Brunswick, NJ) provided automated dosing reduction based on physiologic feedback, but it was taken off the market after <3 years, and the future of similar systems is unclear. Most medication errors are only detected after they have resulted in a significant enough physiologic disturbances to warrant detection while under general anesthesia. This is analogous to the pre-oximeter days when a ventilation mishap had to wait for the patient to turn blue to be appreciated.
Researchers have created sophisticated models for pharmacokinetics and medication interactions, but their availability in clinical practice is limited.13 , 14 Target-controlled infusions are not even a possibility in the United States, and medication visualization tools—like the GE Healthcare Navigator or Draeger SmartPilot—are not widely used. Even the best models are based on an average patient sample and do not take individual variation into account. Medications are most often dosed based on gross patient weight, and there is little consensus about how to dose for obese patients. Attempts have been made to measure propofol concentrations in blood or expired breath in the laboratory.15 , 16 However, there are no commercially available methods for measuring plasma concentrations of any IV medications in real time for anesthesia. We often treat patients as large, uniform fluid buckets where we know the milligrams delivered but little about distribution. Then we “titrate to effect”: giving a dose and seeing what happens.
There are vanishingly few—and mostly brittle—tools to prevent medication errors. Gas lines have Pin-Index systems and anesthetic vaporizers have keyed fillers. Even diesel cars have different sized fueling nozzles, but nothing currently prevents a provider from attaching a potentially lethal dose of local anesthetic—perhaps intended for an epidural—to an IV line. Efforts are currently underway to update the universal Luer connecter, but it has taken decades to apply this lesson from the machine to IV medications.
Levels of mistake proofing exist along a 4-level hierarchy:
- Eliminate the error
- Detect the error
- Detect the defect
- Cognitive aids
Oxygen–nitrous couplers and Diameter-Index Safety Systems are level 1 mistake proofing strategies that make certain errors physically impossible. Yet on the IV side, even 2-provider infusion checks—still awaiting implementation at most institutions—are a much lower form of mistake proofing. From nonstandard vial packaging to nonstandard dilutions to very few mechanisms for dose checking, the medication arena enjoys very little mistake proofing. The medication delivery cycle from vial to syringe to patient to recording is needlessly complex and alarmingly manual, which allows for dozens of potential failure modes for every single medication administration. Each anesthesia provider will administer thousands of medications a year, and even the most conservative estimates suggest each of us likely commits multiple errors annually using existing countermeasures.
Perhaps the single most daunting hurdle to improved medication safety in anesthesia is an entrenched cultural resistance to standardization. Time and again we hear that practitioners within and between institutions have their “own special way” of handling and arranging medications within their individual anesthesia workspace. This really speaks to 2 failures: (1) the inability for national bodies to create a broad consensus and (2) the inability of training programs to instill the value of standardization in trainees. Concerns about overly algorithmic medicine are misplaced. Anesthesia is remarkably simple; most cases require <8 medications (at Seattle Children’s about 80% of cases require only 7 medications). Cases that are more complex would benefit even more from a formal system of organization and interaction, which is why standard layouts are sometimes seen in cardiac anesthesia.
A lack of standardization speaks more to a lack of study of what arrangements are safer. Lessons from the design world do suggest certain themes: (1) infrequently used, emergency medications ought to be in a distinctive location; (2) medications that could be easily confused should be separated; (3) a standard layout helps providers create a mental model that makes mistakes less likely; and (4) in team-based practices, a standard setup creates a shared mental model. Standardization constrains the possibilities—a cornerstone of interaction design. It decreases the cognitive load of providers already burdened by distractions. National safety organizations should recruit professional interaction designers and human factors specialists, leverage available literature and lessons from other fields, study what type of medication arrangements are the safest, and develop standards. Perhaps, more importantly, as a specialty, we need to encourage a cultural change to embrace standardization in medication handling rather than cling to outdated ideas that autonomy and flexibility conflict with safety and reliability.
THE WAY FORWARD
Our specialty needs to acknowledge the irony that we are pioneers in anesthesia machine safety, but in terms of medication handling, we are lacking. Anesthesia lacks the redundancy and accountability enjoyed by our nursing colleagues, yet attempts to apply solutions wholesale from other parts of the hospital often prove problematic. The “Five Rights of Medication Administration” (right patient, right drug, right dose, right route, and right time), eg, are broadly stated goals, not procedural solutions. The five rights “fail to acknowledge that human factors and system weaknesses contribute to errors, and that the focus of the five rights on individual performance does little to reflect that safe medication practices are a culmination of both interdisciplinary efforts of many individuals and reliable systems.”17 Bar coding is one example of a technology translated directly from nursing. Originally designed to confirm the “Five Rights” for medications coming from pharmacy, the applicability to anesthesia is less clear when medications are not prescribed, pharmacy is not involved, and medications come from generic vials in a drawer. Elaborate nursing 2-provider checklists are also not easily translatable to solo anesthesia providers working in a fast-paced environment.
We need more creative, anesthesia-specific solutions much like we have engineered for the machine in the past; however, there remain several unique challenges for the medication side of the anesthesia space, including the following: the fragmentation of the pharmaceutical industry and the sheer number of medications (compared to 3 gases), the added cost and complexity caused by the need for sterility, and the dual-use nature of most medications and devices (eg, infusion pumps are used throughout health care, unlike vaporizers). Our blueprint for next steps includes the following:
- Improve data gathering and transparency for medication errors.
- Enlist interaction designers to rethink the medication handling process.
- Engage industry partners to utilize design thinking and embrace basic standards.
- Overcome the misconception that standardization and innovation are at odds.
A first step would be better data. Capturing medication errors is difficult, but trying to measure the impact of improvements without reliable data is like steering a rudderless ship. The leading national safety is limited by the lack of useful, real-time data regarding medication errors. Instead, anecdotal reports and expert opinion drive most policy recommendations. Local institutions need to be more aggressive about gathering data from clinicians—not just for bad outcomes but for errors and near misses as well. Better data sharing and transparency would be helpful. A national data repository and shared error definitions would be a place to start, but the key is to improve error capture at the local level: accessible electronic mechanisms, frequent prompts, a reinforcing culture. Ideally errors would be captured automatically like physiologic data is today, so technology to capture dosing electronically during administration is needed.
Existing expertise from human factors and interaction design should be exploited. We need to think about ways to engineer mistakes away even though medications involve a lot more human process than the machine: prefilled syringes remove vial-related mistakes; tagged or bar-coded syringes help prevent programming errors; incompatible Luer connections prevent wrong-route administrations. Expertise from outside our specialty will be crucial: engineers, designers, and cognitive psychologists all have valuable insights to reimagine our medication handling process. In some circles, quality improvement has engendered resistance by failing to consider provider workflow. Airport security often falls into a similar trap: a single incident spawns a dramatic response to prevent a specific event, and now everyone must remove their shoes, which slows down the whole security process and alienates customers. Similarly, many improvement efforts are driven by single, past events and end up complicating providers’ workflows to avoid a specific error. Instead, a more proactive approach addressing the systemic causes of errors and a philosophical realignment to improving clinician performance can yield solutions that make clinical care easier rather than harder and facilitate provider acceptance of change. The Anesthesia Medication Template—a device for organizing and identifying medication syringes atop an anesthesia cart—is one example of applying interaction design principles to anesthesia (designed by the authors).18
Industry engagement is critical. Oxygen cylinders and hoses did not turn green on their own. The Global Enteral Device Supplier Association is an example of engaging industry to engineer away mistakes, and the forthcoming NRFit standard involves a new Luer-type connector to prevent local anesthetics from being inadvertently injected into IV lines. The variety of vials that arrive from different pharmaceutical manufacturers certainly increases the risk of anesthesia errors. Requiring bar codes was a start. Creating labeling standards using visual design principles would be better. Prefilled syringe companies tend to put more thought into label design and help to eliminate the myriad issues surrounding medication vials. Pharmacokinetic models—available in the laboratory or outside the United States—should be incorporated into infusion pumps. Devices able to measure plasma levels of medications would help customize dosing to individual patients, and a more interactive anesthetic record that analyzes incoming data could model dosing and identify medication interactions.
Interoperability is a major obstacle. Infusion pumps cannot talk to anesthesia records. Syringes cannot talk to anything. Anesthesia records are just sinks of information that gather and absorb information and attention but rarely provide any useful data back to the user. The creation of a Smart Anesthesia Manager (University of Washington, Seattle, WA) that mines the anesthetic record and provides feedback for the provider foreshadows a future where information systems will be much more interactive and useful with real-time decision support.19 Bar-coding machines suggest a future where syringes can “talk,” and the theory is sound, but it is too early to know if current implementations are a durable solution.20
Changing large industries with multiple players that serve health care customers outside of anesthesia is no easy task. A coherent voice from national societies will be the key. But first we can look inward to our own specialty and examine our own willingness to standardize and embrace change. The Anesthesia Patient Safety Foundation talks about improving the culture of error reporting and instilling a culture of safety. An equally important shift is away from the “standardization versus innovation” duality. Throughout history, many disruptive innovations—money, railroads, the Internet—thrived because of, not despite, standardization. The enemy of innovation is the individual unwillingness to embrace change—not the ability to reach a consensus. The goal is not to make everyone practice the same way. The goal is to engender a readiness to question assumptions and seek out better ways of doing things.
We need to demand more of our industry partners and more of ourselves. Anesthesiology is already a model for other specialties due to our approach to risk analysis and safety engineering on the machine. Making similar gains in medication handling would have even farther-reaching effects because there is so much more overlap with the rest of health care. There are hints of scattered innovations slowly chipping away at our medication vulnerabilities, but there is currently no coherent vision.21–23 A place to start would be a machine analogy: consistently designed vials and syringes with keyed connections and an ability to measure dosing both on delivery and after inside the patient. Eventually a singular vision of an “anesthesia cockpit” is needed where interaction design principles and the best available data enable medications and tools and providers to work in concert to minimize predictable human limitations and maximize clinician performance. Eventually—with enough real-time information at our disposal—we may even be able to develop artificial intelligence tools to identify potential errors before they happen.
Sponsoring a national competition for new solutions could generate innovative ideas. Perhaps someday, we will load medication cartridges into a machine not unlike an automatic coffee maker and then type (or speak) doses that will be delivered with real-time medication identification, dose calculation, automatic documentation, and physiologic feedback. Bayesian feedback loops may automate some medication dosing based on continuously measured plasma levels. Perhaps a visual dashboard will prominently display which medications have been given with visual cues for weight-based dosing and patient-specific kinetic curves. Or perhaps some other solution is waiting to be conceived by a new generation of designers and clinicians. In the 1970s and 1980s, our specialty underwent a safety-related soul searching and emerged smarter and more reliable by reengineering the way we approach the anesthesia machine and airway management. It is time to do the same for IV medications in the operating room.
Name: Eliot B. Grigg, MD.
Contribution: This author helped in conceptualization of the data and writing the manuscript.
Conflicts of Interest: E. B. Grigg designed the Anesthesia Medication Template, which is referenced in the article.
Name: Axel Roesler, PhD.
Contribution: This author helped in conceptualization of the data and writing the manuscript.
Conflicts of Interest: A. Roesler designed the Anesthesia Medication Template, which is referenced in the article.
This manuscript was handled by: Ken B. Johnson, MD.
2. Cooper JB, Newbower RS, Long CD, McPeek B. Preventable anesthesia mishaps: a study of human factors. Anesthesiology. 1978;49:399–406.
3. Whitaker D, Brattebø G, Trenkler S, et al; European Section and Board of Anaesthesiology of the UEMS. The European Board of Anaesthesiology recommendations for safe medication practice: first update. Eur J Anaesthesiol. 2017;34:4–7.
4. Cooper L, Nossaman B. Medication errors in anesthesia: a review. Int Anesthesiol Clin. 2013;51:1–12.
5. Wahr JA, Abernathy JH 3rd, Lazarra EH, et al. Medication safety in the operating room: literature and expert-based recommendations. Br J Anaesth. 2017;118:32–43.
6. Nanji KC, Patel A, Shaikh S, Seger DL, Bates DW. Evaluation of perioperative medication errors and adverse drug events. Anesthesiology. 2016;124:25–34.
7. Barach P, Small SD. Reporting and preventing medical mishaps: lessons from non-medical near miss reporting systems. BMJ. 2000;320:759–763.
8. Christ RE. Review and analysis of color coding research for visual displays. Human Factors. 1975;17:542–570.
9. Grissinger M. Color-coded syringes for anesthesia drugs-use with care. P T. 2012;37:199–201.
10. Hutchins E. Cognition in the Wild. 1995.Cambridge, MA: MIT Press.
11. Degani A, Wiener EL. On the Design of Flight Deck Procedures. NASA Contractor Report 177642. 1994.Mountain View, CA: NASA Ames Research Center
12. Martin LD, Grigg EB, Verma S, Latham GJ, Rampersad SE, Martin LD. Outcomes of a failure mode and effects analysis for medication errors in pediatric anesthesia. Paediatr Anaesth. 2017;27:571–580.
13. Schnider TW, Minto CF, Gambus PL, et al. The influence of method of administration and covariates on the pharmacokinetics of propofol in adult volunteers. Anesthesiology. 1998;88:1170–1182.
14. Cortínez LI, Anderson BJ, Penna A, et al. Influence of obesity on propofol pharmacokinetics: derivation of a pharmacokinetic model. Br J Anaesth. 2010;105:448–456.
15. Wang X, Zhou Q, Jiang D, Gong Y, Li E, Li H. Ion mobility spectrometry as a simple and rapid method to measure the plasma propofol concentrations for intravenous anaesthesia monitoring. Sci Rep. 2016;6:37525.
16. Colin P, Eleveld DJ, van den Berg JP, et al. Propofol breath monitoring as a potential tool to improve the prediction of intraoperative plasma concentrations. Clin Pharmacokinet. 2016;55:849–859.
17. Matthew G. The five rights: a destination without a map. P T. 2010;35:542.
18. Grigg EB, Martin LD, Ross FJ, et al. Assessing the impact of the anesthesia medication template on medication errors during anesthesia: a prospective study. Anesth Analg. 2017;124:1617–1625.
19. 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;60:207–210.
20. Jelacic S, Bowdle A, Nair BG, Kusulos D, Bower L, Togashi K. A system for anesthesia drug administration using barcode technology: the Codonics Safe Label System and Smart Anesthesia Manager. Anesth Analg. 2015;121:410–421.
21. Jensen LS, Merry AF, Webster CS, Weller J, Larsson L. Evidence-based strategies for preventing drug administration errors during anaesthesia. Anaesthesia. 2004;59:493–504.
22. Merry AF, Anderson BJ. Medication errors–new approaches to prevention. Paediatr Anaesth. 2011;21:743–753.
23. Orser BA, U D, Cohen MR. Perioperative medication errors: building safer systems. Anesthesiology. 2016;124:1–3.