To the Editor:
The recent article by Nanji et al.1 concerning errors related to anesthetic drug administration is interesting and raises a number of provocative questions. However, we are concerned that the manner in which the data are presented and interpreted may lead readers to conclusions that may not be warranted.
Nanji et al. have utilized a very broad definition of drug administration error. For example, “significant hypotension (mean arterial pressure < 55 mmHg) that is not treated”1 is listed as a drug error in table 2. We would argue that depending upon the circumstances, this is not an error of drug administration (it may be an error in anesthetic management) and may not be an error at all. We would also argue that an unattended syringe of hydromorphone (table 5) is a not a drug administration error, although it may be a violation of a hospital policy for handling controlled substances. The authors have given other examples of their definitions of drug administration error but have not provided a complete list of all drug error definitions or a list of the errors observed in this study. Thus, it is difficult to know what was actually measured. This is important because their reported rate of error is at least an order of magnitude greater than reported by other investigators.
Nanji et al. have also utilized a very broad definition of adverse drug events. We would argue that the example of adverse drug events listed in table 2, “a patient with > 4/10 pain on emergence that is not treated until after arriving in the recovery room,”1 is not an adverse drug event. It has to do with the strategy for perioperative pain management rather than drug administration per se.
Webster et al.2 performed a key study of anesthetic drug administration error using prospective facilitated incident monitoring (self-reporting) by anesthesiologists in New Zealand. We replicated this study at the University of Washington Medical Center, Seattle, Washington, using similar methodology and obtained similar results,3 as did Zhang et al.4 in a study of over 24,000 anesthetics in China. The rate of drug administration error in the study by Webster et al. was 0.75% of anesthetics; it is important to note that the rate of error was expressed as the percentage of patients who were subject to at least one error (the rate “per anesthetic”), not as a fraction of the total drugs administered. It is also important to note that Webster et al. were concerned with actual performance errors of drug administration that reached the patient, such as administering the wrong drug, not process errors such as errors in labeling or record keeping.
Nanji et al. state that the rate of error determined by Webster et al. (as confirmed by us in the United States and Zhang et al. in China) is “markedly lower than the rates that we found, which may be due to provider reluctance to self-report errors or failure of providers to recognize errors they have made.”1
If we examine table 5, we find a classification of errors as defined by Nanji et al. These include five types of errors that did not reach patients—labeling errors (37 to 24% of total errors), documentation errors (26 to 17%), monitoring errors (10 to 6.5%), wrong timing (5 to 3.3%), and other (2 to 1.3%). These errors account for 52% (80/153) of the total errors. Webster et al. did not classify any of these error types to be drug administration performance errors (such as giving the wrong drug), and these error types were not reported in their study.
In order to compare the results of Webster et al. to that of Nanji et al., we should first subtract the 80 errors that are not directly related to the performance of drug administration, leaving 73 errors directly related to drug administration (error rate, 73/3297 = 0.022 or 2.2%). The rate of errors per anesthetic (Nanji et al. did not specify the number of patients effected by errors, but an assumption of no more than one error per patient is a reasonable approximation) would be 73/277 = 0.26, i.e., 26% of anesthetics would have been affected by an error. This is 35 times greater than the rate reported by Webster et al., which was 0.75% of anesthetics. How are we to explain this enormous difference in results? Nanji et al. suggest that this is due to dramatic underreporting of errors in studies where providers report their own errors.
However, Merry et al.5 (the same group in New Zealand who reported the study by Webster et al.) also performed a direct observation study. In that study, the rate of drug administration error was 0.32% of drugs administered (table 2), or 3.2% of anesthetics (0.0032 × 5084 = 16; 16/509 = 0.032). Thus, the rate of drug administration error (comparing apples to apples, using the error classification of Merry et al.) is about 10 times higher in the Nanji et al.’s direct observation study than in the Merry et al.’s direct observation study.
We agree that self-reporting underestimates the rate of error. However, comparing the direct observation data from the study by Merry et al. to the self-reporting data from the study by Webster et al. (both from New Zealand), the magnitude of the difference is about 4- to 5-fold (3.2 vs. 0.75%), not 35-fold as when comparing the study by Nanji et al. to the study by Webster et al. (26 vs. 0.75%). Moreover, Nanji et al. found a rate of error about 10 times greater than Merry et al. when both used direct observation.
We would speculate that the higher rates found by Nanji et al. have to do with what appears to be overly broad definitions of error; however, it is impossible to know from their article because the actual drug error data are not provided. We believe that this is important, and that Nanji et al. should provide their actual data so that readers are able to make their own interpretations of what constitutes a drug administration error and what does not.
Also, we were disappointed that Nanji et al. did not provide data describing their use of the Codonics Safe Label System (Codonics Inc., USA) to produce syringe labels or their use of the MetaVision anesthesia information system (iMDSoft, USA) to scan bar codes on syringes before administration, since there is some evidence that proper labeling and scanning bar codes may reduce drug administration errors. Without having more information, it is not possible to know whether these technologies had a significant impact on the errors that they have reported. Simply having these systems in place does not ensure that they are properly used, as reported previously by our group6 and by Merry et al.5 The fact that 24% of the errors reported in table 5 involved labeling suggests that Nanji et al. are not obtaining the full potential benefits of the Codonics Safe Label System. Clearly, education, a culture of safety, and the details of the implementation are all important when it comes to technology that is employed to mitigate medical errors.
The authors declare no competing interests.