Doing the Math: Computation Errors During Critical Drug Administration

See Article, p Drug errors in medicine are ubiquitous, underreported, costly, and dangerous. These concerns are equally applicable to the operating room where anesthesia providers err in roughly 5% of medication administrations, with a large majority of them being judged as preventable.1 Interestingly, medication errors (and near-miss) rates appear pervasive and relatively consistent around the world from New Zealand (1:133 anesthetics) to South Africa (1:274 anesthetics) to Japan (1:450 anesthetics).2 Anesthesia professionals are especially vulnerable as they are usually solely responsible for diagnosing, prescribing, identifying, preparing, dosing, dispensing, and administering the right medication at the right dose to the right patient via the correct route at the right time. This process is further complicated if the required drug needs to be diluted or mixed before administration and/or needs to be delivered by controlled IV infusion device in a highly concentrated solution. There are many opportunities along the complex drug administration pathway in the operating room for a drug error to occur. According to a study based on incident reporting in New Zealand hospitals, the most frequent drug errors were dose errors (20%) and drug substitutions (20%) with most errors involving IV boluses and IV infusions.3 In a related retrospective study out of Japan, the authors also found overdose to be the primary cause of anesthetic medication error, with mistaking units, misuse of the infusion pump, and misunderstanding of the drug concentration among the most common causes for these errors.4 Indeed, mistakes during drug computations; that is, miscalculations, are an important subgroup of all drug errors, but seldom achieve the notoriety of other drug administration errors. Fortunately, in this issue of the Journal, Black et al5 report the findings of a 15-question standardized examination designed to determine the frequency of computational drug errors specific to anesthesia residents and faculty in teaching programs. The standardized test was handed out during grand rounds at 7 different US academic institutions to all faculty and residents in attendance to a total of 209 residents and 162 faculty. Examinees were allowed unlimited time and use of any calculation devices on their person at the time. The only demographic data collected were years of postgraduate training and hours of sleep the night before the test. Any answer outside of 5% greater or lesser than the correct answer was deemed incorrect. The results were alarming as only 20% of residents and 25% of faculty scored 100% on the test. Interestingly, there was no correlation between error rates and hours of sleep preceding the examination, but errors predictably increased with the number of computational steps needed to derive the correct answer.5 The authors opine that there may be an “underappreciated deficiency in the computational skills of both anesthesia residents and faculty.” We agree! Indeed, computational proficiency (used here in the context of one’s ability to do mathematic calculations correctly) is seldom tested before medical school or during residency selection, and this omission usually continues during medical training and the anesthesiology 3-year continuum. Nonetheless, anesthesiology residents and faculty are responsible for calculating and administering potent medications in specific doses at titrated rates for patients of varying weight multiple times a day, and often during time-pressured events. Neonatal and pediatric patients present special challenges where the majority of drugs have to be diluted 10-fold or greater, with even smaller margins for potentially harmful or lethal errors. Kaushal et al6 compared pediatric and neonatal medication errors to a previous adult inpatient study and found that the potential adverse drug event rate was 3 times higher in pediatric patients. Neonates in the neonatal intensive care unit were at greatest risk with 34% of the potential adverse drug events involving incorrect dosing. Although many studies have shown the positive impact computerized physician order entry has had on the prevention of medication dosing errors in both adults and pediatric patients,7,8 anesthesiologists seldom have the luxury of ordering a medication via a computerized order entry system with built-in error-detection algorithms. Instead, one provider is solely responsible for the identification, dilution, labeling, calculating, and administering the correct drug. But some might ask: “Given the ubiquitous presence of smart phones with their specialty applications, computerized anesthesia work stations, and satellite pharmacies (and pharmacists) in the operating rooms, are human computational skills really critical in 2019?” First, we acknowledge the reality that our clinical practice has changed—25 years ago, we all compounded our own vasopressor and vasodilator infusions, calculated the infusion rates based on patient weight, and were solely responsible to monitor the accuracy of the administration (sometimes by methods as basic as counting microdrops per minute titrated by a “Dial-a-Flow” restrictor device). But while acknowledging that reality no longer exists, we must also recognize the limits of current technology and resources. Smart phones can be lost or forgotten. Batteries become depleted. Pharmacists rarely work 24/7 in satellite locations like the operating rooms. And most automated infusion or syringe pumps will have a limited drug library of the most commonly used medications. When patients require “specialty drugs” like lepirudin or treprostinil, anesthesia professionals will be charged with the task of independently programming an infusion device using the “generic label.” Provider calculations must be correct. But alarmingly, the findings of Black et al5 suggest that we err 12%–22% of the time when faced with this vital task, and anesthesiology residents commit serious errors (defined as a 100-fold miscalculation) twice that of faculty. Perhaps a society where citizens rely so heavily on smart phones and similar devices for simple calculations and unit conversions will discover “doing the math” even more tense and prone to error in the future as clinicians struggle with drug calculations that were never “hard-wired” by years of repeated experience. Thus, we affirm that the ultimate responsibility for correct administration of drugs in the perioperative period still rests with the anesthesiologist, and computational skills remain critical to that task as we ensure patient safety. Furthermore, the American Board of Anesthesiology upholds this position within their requirements for the In-training, Part 1, Basic and Advanced Examinations. This topic is specifically identified on page 42 of the American Board of Anesthesiology Content Outline in Section II.E.4.g.2: Medication Errors: Assessment and Prevention.9 Moreover, candidates for American Board of Anesthesiology certification may again be faced with the topic of drug errors in the Objective Structured Clinical Examination as noted in the course outline A.3: “Peri-procedural complications.”10 For example, the candidate may be challenged to discuss the cause and consequences of a serious medication error with a simulated patient. If we stipulate that medication errors, specifically computational drug errors, continue to be a real threat to our patients in an era where anesthesiology residents are neither proficient nor routinely tutored on these skills—and that we cannot always rely on computers, smart phones, or satellite pharmacists—what can we do to mitigate the risk? System solutions are feasible and viable. For instance, why are we still mixing and diluting drugs while simultaneously providing care for patients? If computational drug errors are as abundant as the authors of this study suggest, then we should urge routine use of ready-to-administer drugs and syringes in the operating room. Ready-to-administer drugs would reduce the need to admix and dilute medications for use in the operating room, are noted for bright clear labeling, and could decrease the rate of drug administration errors. In addition, ready-to-administer syringes and packages would decrease drug preparation and administration time allowing more attention for other key patient-related responsibilities. While the current literature is still quite limited on this topic, a recent study in the Journal of Patient Safety found ready-to-administer products to be associated with significantly fewer errors than our traditional IV push practice (22% vs 79%, respectively).11 What else can be done? We urge current program directors and academic anesthesia chairs to reflect on the findings of Black et al5 and implement regular didactic sessions and testing for computational expertise, and if necessary, update their curriculum to improve resident performance. Indeed, one might consider computational competency as a core entrustable professional activity for entering residency or an obligatory milestone in the early development of an anesthesiology house officer. Moreover, the current lapse in skills provides an opportunity for educators to explore and define improved pathways for adult learners to enhance these special computational abilities. While multiple approaches exist to improve fluency in computational skills, practice is always a necessary component. But beyond that, experts partition effective computational thinking into key elements such as decomposition, pattern recognition, abstraction, and specific algorithms. Decomposition partitions complex problems into simpler core elements. Pattern recognition improves learning from earlier, similar problems or experience. Abstraction separates the “wheat from the chaff” to ignore irrelevant details. And finally, key algorithms may be imprinted to utilize similar solutions to recurring problems. But beyond simply improving our “math skills,” anesthesia leaders and students will hopefully affirm a principle observed in other scientific disciplines—that expert computational thinking is a diffuse and valuable skill that imparts significant ancillary benefits in terms of complex problem solving.12 As we applaud the unique findings of Black et al,5 we look forward to further efforts to categorize and understand all causes of medication errors. More progress is needed, not just in human computational performance, but perhaps with technological advances such as those advocated by Dr Alan F. Merry13–15 and others. In the meantime, we must continue efforts to address the performance of anesthesia residents and professionals in calculating, dosing, and administering potent anesthesia and life support medications. Our patients deserve no less. DISCLOSURES Name: Ilana R. Fromer, MD. Contribution: This author helped design, formulate, prepare, and construct the editorial. Conflicts of Interest: None. Name: Richard C. Prielipp, MD, MBA, FCCM. Contribution: This author helped design, formulate, prepare, and construct the editorial. Conflicts of Interest: R. C. Prielipp is a member of the Speakers Bureau for Merck & Co, Kenilworth, NJ, and a consultant for Fresenius Kabi USA, LLC, Lake Zurich, IL. This manuscript was handled by: Edward C. Nemergut, MD.