Frequency, Timing, and Types of Medication Ordering Errors Made by Residents in the Electronic Medical Records Era

Ari Garber, MD, EdD; Amy S. Nowacki, PhD; Alexander Chaitoff, MPH; Andrei Brateanu, MD; Colleen Y. Colbert, PhD; Seth R. Bauer, PharmD; Zubin Arora, MD; Ali Mehdi, MD; Simon Lam, PharmD; Abby Spencer, MD; Michael B. Rothberg, MD, MPH


South Med J. 2019;112(1):25-31. 

In This Article


In this study of medication orders during a 4-year period in one large academic medical center, we found that approximately 4% of all medication orders placed by internal medicine trainees contained pharmacist-identified errors. The most common type of ordering error was failure to adjust dosing for renal impairment, and the class of medication most commonly involved in errors was antimicrobials. The rate of medication ordering errors varied by residents' level of training, shift (day vs night), and month of year. Specifically, the highest rate of errors occurred during the PGY1 year, particularly during the month of August. The rate of errors then declined throughout the year. PGY2s committed the fewest errors. Our hypothesis that errors would be more frequent at night and during transitions of care was not borne out. In fact, errors were more frequent during the day shift than during any other period.

Few studies have directly examined the association between medication order errors and resident training level. Results of prior studies have been conflicting. For example, many studies used differing definitions of medication errors and examined the association in cohorts of residents from differing specialties. Furthermore, most studies were conducted before the EMR era, and the results of many are limited by small sample sizes (ie, containing <5% of the number of prescriptions analyzed in the present study), making findings difficult to generalize. For example, Pacheco et al[25] examined 180 prescriptions written in a pediatric emergency department setting and found that senior residents were more likely to make medication errors than were interns. Alternatively, Pippins et al[26] analyzed approximately 2000 medication errors that occurred on an inpatient medicine service and found that increasing experience level was associated with significantly fewer errors in medication reconciliation. Finally, Garbutt et al[27] reported that among medical students and residents using paper orders, some self-reported aspects of safe prescribing were related to experience, whereas others were not.

The decrease in medication order errors observed from PGY1 to PGY2 may represent an increased familiarity with the EMR, because effective use of EMRs has been shown to reduce medication errors.[28] Alternatively, it may represent increased knowledge, because medication-related knowledge deficiencies are associated with medication errors.[29] Our finding that the least-often prescribed medications were many times associated with higher order error rates supports this interpretation. More puzzling is the increase in order errors from PGY2 to PGY3. For PGY3 residents, who placed the fewest orders, knowledge decay and/or the complexity of patients—and thus medication orders—may have contributed to this increase.[30] It is also possible that PGY3s are the least likely to consult with others before placing an order.

We did not find evidence of a July effect. In fact, order error rates were lower in July before peaking in August, after which they gradually declined. In 2010, Philips and Barker[31] found that fatal medication errors increased by 10% in July in US counties with teaching hospitals; however, they addressed only fatal medication errors and included data going back as far as 1979, a time with minimal supervision of house staff and no EMR. Other studies have offered conflicting findings. One systematic review found evidence of increased mortality and decreased work efficiency in July, but could not determine whether medication errors were implicated.[32] Other studies found no difference in outcomes among patients seen by medicine teams and time of year, although medication errors were not reported specifically.[33,34] Our finding that error rates in July were among the lowest in the year may be caused by heightened supervision during new residents' first month.[35] Alternatively, new residents may be particularly careful in their first month and become less so as they become more comfortable in their role or realize that their orders are being checked by a pharmacist. Although autonomy is integral to resident learning,[36] the sharp spike in August could indicate that supervision from any number of members of the interprofessional care team is being withdrawn prematurely. Ultimately, additional research is needed to confirm the mechanism behind the association that we report here.

Similarly, we hypothesized that nighttime may be particularly hazardous because of limited supervision and fatigue,[11,37–39] a finding that was supported by one study of interns in an emergency department.[17] That study was conducted before the introduction of the EMR, however. In contrast, smaller inpatient studies, both conducted before[18] and during the EMR era,[40] report the highest rate of medication errors during the day. These results are similar to our findings, but the reasons for this are unclear. It may be that the volume of orders, the type of orders, and the competing demands (eg, rounds, family meetings, educational conferences) during the day contribute to increased risk.

Another vulnerable period is patient handoffs.[41–44] Inadequate signouts have been associated with adverse events,[45] but not specifically with medication errors. We did not find any increase in order errors during handoff periods, perhaps because errors that originate from inadequate handoffs manifest after the transition time; however, it also is possible that the use of handoff tools helped to limit errors.[41–43] At our institution, residents are taught the situation, background, assessment, recommendation (SBAR) mnemonic, which is one of the most cited and recommended handoff techniques,[46,47] as well as the more recently published patient information, active hospital course, status, supporting data, overnight to do, nursing, summary by received (PASSONS) mnemonic.[48] Similar to the July effect, awareness of the dangers of handoffs may have actually made them safer, at least concerning risk of a medication order error.

Our findings have several important implications. First, the timing of errors suggests the need for increased supervision in the months of August and September, not simply in July. Third-year residents may need to take more care and occasionally consult with others, because these residents write fewer orders and may become less familiar with entering orders into the EMR. Second, the high frequency of order errors, despite use of an EMR that detects common errors such as allergies and duplicate orders, underscores the need for clinical pharmacist review, given that EMR alerts often are ignored.[22] Similarly, the frequency of error types highlights the need for better attention to dosing for patients with renal impairment. Others have proposed that EMRs could prevent such errors,[49] but despite evidence suggesting computerized support for renal dose adjustments improves order appropriateness,[50] most EMRs do not yet integrate measures of creatinine clearance with drug prescribing. Third, medications with the highest rates of ordering errors were generally those that were prescribed infrequently, pointing to the need to be cautious when prescribing or filling less common medications. There were exceptions to this, however, such as antimicrobials and anticoagulants. Future research may explore why this was the case. For example, it is possible that if residents are aware that a pharmacist will be checking their orders, they may be less careful when writing them. In that case, the redundancy incorporated into the system as a safety measure could produce the opposite effect. No matter the reason, the differing types of errors seen with commonly versus less commonly prescribed medications underscore the importance of resident education about specific kinds of errors that are common when ordering certain types of medications.

Limitations to this study include the retrospective nature of this study and the focus upon residents within a single institution, which limits generalizability of findings. We also included only certain types of errors in our analysis (eg, errors categorized into predefined iVENT communication categories); thus, some ordering errors may not have been included in our analysis. As such, even though the rate of errors in this study is comparable to the rate of medication errors reported in more general epidemiologic studies,[51] absolute frequencies should be interpreted with caution, although trends in ordering errors across resident and order characteristics should be more accurate. Furthermore, only data on ordering errors detected and classified by pharmacists were used. Pharmacist staffing levels and experience may have affected the rate of iVents generated, and the positive predictive value of pharmacist use of an iVent is likely not 1.00. Moreover, because iVents are created before medications are dispensed, it is likely that none of the order errors in this study resulted in harm. Real-time pharmacist surveillance has been used previously to study medication error rates,[52,53] and patterns of near misses are commonly used in safety reporting systems and have been credited with improving safety in a number of industries.[54]

Despite these limitations, our study had a number of strengths, including the large sample size of both medication orders and residents, our ability to accurately track medication ordering errors over time, and being one of the few studies conducted in the era of widespread EMR use.