An urban medical system's exploratory study of medication errors

Abstract Aims This study sought to identify patterns of medication errors with respect to shifts, day of week, unit involved, severity, medication class and cause of errors and to propose possible solutions. Design This was a retrospective explorative study using a database containing 605 medication events from two medical centres. Variables assessed include medication type, the error severity, and time the medication was ordered, the unit that the error occurred on and the day of the week of the errors. Methods Simple percentages were used to report the results, and point‐biserial correlation was employed to test for significant differences between the day and night shifts. Results There were no statistically significant findings when comparing event severity against the a.m. or p.m. shifts. The medication classes with the most errors were antibiotics, and the most common reason cited for errors was dose omission. The most commonly reported severity level was a 2 which requires increased patient monitoring.


| BACKG ROUND AND LITER ATURE RE VIE W
With the increasing focus on nurse sensitive metrics and their concomitant impact on outcomes and Centers for Medicare and Medicaid Services (CMS) reimbursements, it follows that all aspects of nursing practice are subject to intense scrutiny. Carlton and Blegen (2006) examined medication errors through the lens of quality management and categorized medication errors as either active or latent. Both categories represented errors which reached the patient. An active error was caused by dose miscalculation and dose omission while latent errors were promoted, at least in part, by extraneous environmental factors such as fatigue, interruptions during the medication procurement and administration process and staffing issues. Other causes of latent errors seem to be influenced by poor nurse educational preparation or lack of basic pharmacological knowledge (Bower, Jackson, & Manning, 2015;Frith, Anderson, Tseng, & Fong, 2012). Emphasizing and managing conditions which may lead to errors can be highly complex mainly because one cannot control for or eliminate all intrinsic or extrinsic factors which have an impact on medication administration. In 2012, Frith et al. conducted a retrospective correlational study to determine the effect that staffing variables and skill mix have on medication errors. It was found that the more complex the medical diagnosis, the more likely it was that the patient would experience a medication error during their hospitalization. Tzeng, Yin, and Schneider (2013) reported that 9.2% of all admitted people will experience an adverse event related to medication administration. While that number seems very high, it is likely significantly below the actual number. For reasons ranging from inattentional blindness to the desire to avoid disciplinary action, many nurses do not report or even recognize medication errors. These same researchers also identified that medication administration incidents represented 53% of all reported clinical errors. Of these, 58% were medication doses that were missed and 21% because the wrong dose was administered (Tzeng et al., 2013). Other notable takeaways from this study include the finding that decreasing the frequency of licensed practical nurse (LPN) use resulted in an overall decrease in errors. This finding held for all inpatient nursing units where LPNs were used. The effect of complexity of diagnosis and increased errors has been found in other research (Breckenridge-Sproat, Johantgen, & Patrician, 2012;Hall, Doran, & Pink, 2004).
The issue of nurse interruptions while administering medications continues to surface in contemporary nursing research. Studies have linked interruptions during the medication administration process with increased errors (Bower et al., 2015;Harkanan, Turunen, Saano, & Julkenen, 2013;Malone, 2016). More ominously, Raban and Westbrook (2014) found that errors committed because of interruptions were more likely to result in serious patient harm and death. These researchers identified three types of interruptions which impinge on the nurse's time and ultimately affect the process of medication delivery. The first are sudden and unpredictable changes in the patient's condition which merit rapid intervention and may interrupt the medication administration process mid-stream. This type of interruption makes it difficult to maintain the ordered timeliness of medication delivery since the nurse will have to manage the existing emergency and then go back to the patient and either initiate the medication administration process all over again or return in the middle of the process, thus increasing the chance that a critical step will be omitted. In one study, nurses were interrupted a average of 43 times in a ten-hour period during medication rounding (Tucker & Spear, 2010). Other studies showed that 17% of all medication administrations were interrupted in some way (Young et al., 2015). It is unreasonable to think that this type of interruption can be completely eliminated, but it can be reduced.
Another type of interruption occurs when well-meaning families insist on persistently asking the nurse about the patient when the nurse is trying to calculate or titrate the dose of a drug or intravenous medication. This problem may be especially prevalent in critical care units with an open visitation policy and where powerful vasoactive medications are being titrated. Family members may not even be aware that this type of interruption can cause a nurse to commit a potentially dangerous medication error. Some hospitals have implemented a system where if a nurse is at the medication dispensing machine, calculating a medication, titrating or adjusting an intravenous drip, or is in the patient/medication scanning phase of the medication administration process, they are not to be interrupted (Bower et al., 2015;Bravo, Cochran, & Barrett, 2016;Cloete, 2015). While this sounds workable in theory, operationally it might be difficult to implement effectively and consistently. Interventions which purport to reduce interruptions during the medication administration process have not been shown to be very effective (Raban & Westbrook, 2014). Additionally, hospitals might be slow to implement any actions which seem to discourage people or families from asking questions.
The third type of interruption occurs when a physician, nurse, other provider or staff member interrupts the nurse to relay insignificant or inconsequential information that may even be unrelated to the patient that is being care for. Tucker and Spear (2010) found that when physician's interrupted nurses during medication administration to relay redundant messages such as "I have written new orders" often had the effect of making the nurse feel frustrated and demeaned as they are well-aware of the need to check for new orders and the procedure for checking new orders is hardwired into their daily practice. This interruption will also have the same effect as other interruptions since the nurse will have to re-attend to the original task and creating the possibility of an error.
Since medication errors can happen at any time, it is prudent to examine any potential differences between hospital work shifts.
The nurses in this health system continue to work 12-hr non-rotating shifts. The perception of this organization was that nursing care and nursing quality on the night shift and possibly on weekends might be compromised because of the increased use of per diem and external agency staff.

| E THIC AL CONS IDER ATIONS
The health system's Institutional Review Board (IRB) approved the research prior to data collection. This body also determined that the study met all ethical requirements of the National Institutes of Health. Since the examination of medication errors could have potentially uncovered sensitive information with a possible impact of career about the nurses involved in these misadventures, there was no information collected which could reasonably be used to identify the nurses involved.

| ME THODS
This research focused primarily on the nurse-patient interface and the errors that occur at this point in the process, the actual administration of the medication. While the author fully recognizes that medication errors can (and do) occur at every part of the medication administration cycle, the nurse is the final barrier between the medication and the patient and arguably represents the most critical element of the process. This study is best described as a retrospective exploratory study involving two urban medical centres in north Texas. One of the hos- tient. An error that is coded as Severity Level 3 means that the patient not only required an increase in monitoring, but also required additional time in the hospital. This level also includes an upgrade in the level of care. Severity Level 4 means that there was temporary harm to the patient while Level 5 event, while relatively rare, means that the patient will have some degree of permanent harm as a result of the error. Severity Level 6 is reserved for deceased people who, after extensive chart review, have died with the event in question as the primary or major contributing factor which leads to the death.
Quality department abstractors at each facility, via each medical centre's event reporting system, identified all pertinent medication events for the research time frame. The data were then scrubbed of any patient identifying information as well as the names of individual nurses involved which helped ensure little to no patient, staff risk or that individual harm would result from the study. To preserve data integrity and ensure uniformity, only one abstractor at each facility was tasked with identifying and pulling medication error events from the event reporting system. The scrubbed data, redacted of identifiable protected health information, were delivered electronically to the author for database creation and subsequent analysis.
The final data set comprises 605 medication events that reached the patient. Near misses, although important, were not used in this analysis since they did not represent an actual drug administration error. Results were reported by category using simple percentages.
Point-Biserial correlation was used to test for significant differences between event severity and the a.m. and p.m. shifts.

| RE SULTS
See of the total with significantly lesser percentages logged in the L&D, perioperative, emergency and neonatal intensive care areas.
The researcher also captured the time that the ordered medication was due to be administered (Graph 1). The histogram displays a diffuse bimodal distribution of the data points with peaks roughly corresponding to 09:00 and 20:00.
Finally, a Point-Biserial correlation was conducted using shift as the dichotomous variable (Table 2). There were no statistically significant findings in Event Severity between the two shifts. The mean magnitude was 2.6 for the a.m. shift versus 2.5 for the p.m. shift.
There were notable differences between the a.m. and p.m. shifts that were revealed during this research (

| D ISCUSS I ON
Error magnitude is of particular importance since any error has the potential to result in permanent sequelae or death. There are also financial implications with regard to medication errors. Extended lengths of stay (LOS), changes in the level of patient care (acute care to intensive care) and the payment of compensatory damages can adversely affect a hospital's operational budget and public reputation.
Somewhat surprising are that errors involving analgesics (which includes Schedule II opioids) accounted for only 13.2% of the total errors. The researchers surmised prior to beginning the research that this would be considerably higher. This less than expected finding may be a result of the increased scrutiny being placed on the availability and suitability of this class of drugs for analgesia during and following routine procedures and for short-term pain control as well as increased institutional and regulatory oversight.
Interestingly, drug administration to the wrong patient accounted for only 1.0% of the total. This is an encouraging finding as earlier studies exploring medication errors put this figure at closer to 6% (Hughes & Blegen, 2008;Mohammed, Human, Esmaeil, & Syyedeh, 2013). It may be that the Joint Commission's (TJC) continued emphasis on accurate patient identification is finally becoming hardwired into a routine culture of safety in the medication administration process instead of being regarded by the bedside nurse as another cumbersome step which had to be completed and which was frequently seen as expendable.
An examination of the errors when compared with the day on which they occurred shows no meaningful difference on the weekends. This was a somewhat unexpected finding since the perception among the administrative staff and the hospital's coordinating councils was that more errors occurred on weekends as there is less administrative oversight. In this study, that perception was not supported.
The relatively low percentage of errors reported in the emer- The graph that represents when the medications involved in the errors were due to be give provides some clarity about when the errors are occurring. These times also happen to be when many routine medications are timed in the hospital setting. Of additional interest is that on the p.m. shift, the 19:00 time also shows a peak in administration time errors. This is also the time that most hospitals undergo a change of shift. This may represent an artefact of the data, or it may be reflective of the evidence that errors seem to occur more frequently during the changeover period from day shift to night shift (Mardis et al., 2016). A similar spike is not seen at the time of the 0700 shift change.
The lack of statistically important differences with regard to severity when comparing shifts was not anticipated. There were, however, important differences between the shifts when examin-

| RECOMMENDATIONS
Three proposed recommendations for improvement will be discussed in this section. The first is that unnecessary interruptions during medication administration rounds must be significantly curtailed. A variety of measures have been attempted to ensure that the nurse preparing and administering medications can do so with as few interruptions as possible. Some include lighted lanyards which alert staff, visitors and people that the nurses are on medication rounds and should not be unnecessarily interrupted, the "sterile cockpit" approach has been tried by some organizations with limited success (Kapur, Parand, Soukup, Reader, & Sevdalis, 2015), medication preparation areas can be marked or cordoned off and only permit one nurse at a time to enter and procure medications (Hayes, Jackson, Davidson, Daly, & Power, 2017), and some have adopted a strategy where the nurse who is giving medication dons a brightly coloured vest to denote that medication rounding is occurring (Johnson et al., 2017).
Unfortunately, all of these methods may promote a sense, at least from people and families that the nurses should never be interrupted.
Impairing communication with the healthcare team or people may be an unintended consequence of these implementations and impeding communication with any member of the healthcare team would not promote a desirable outcome. Other approaches include the silencing all intrahospital communication devices during the medication preparation and administration processes and routing incoming personal calls through the unit clerk for vetting prior to transferring any call to the nurse. In many organizations, overhead paging has been curtailed or even eliminated altogether in an effort to create a quieter, more restful and less frenetic clinical environment.
The second recommendation is that barcoding of people and medications be enacted in every patient care area where medications are procured and administered. While by no means a panacea, implementation of barcoding is becoming standard in most hospitals (Mekonnen, Abebe, McLachlan, & Brien, 2016). Initially, there had been considerable concern that the process of barcoding would take up an inordinate amount of time and could possibly harm people if the nurse or physician was not able to access emergency drugs in a rapid and expeditious manner. In fact, these problems did occur with the 1st generation of barcoding technology in that it would sometimes take several minutes to reconcile and "load" into the automatic medication dispensing machine or convey the medication to the nurse (Poon et al., 2008). In response, nurses would develop workarounds to ensure that emergency medications were always available. There were also concerns that people might view barcoding as unnecessary or too personally invasive or that the nurse, already strapped for time, would now spend even more time away from their people to scan and reconcile medications. Yet, when barcode scanning is consistently practiced and the supporting technology platform is robust and accurate, the result is that fewer medication errors occur (Seibert, Maddox, Flynn, & Williams, 2014;Truitt, Thompson, Blazey-Martin, NiSai, & Salem, 2016).

G R A P H 1 Ordered medication times
The third recommendation is somewhat controversial, but is important to broach and discuss. The notion of a completely non-punitive culture of safety, while of the best intent, may be indirectly contributing to medication errors. When the non-punitive approach to reporting medication errors became more widespread during the early 21st century, it was of great importance since it allowed realtime data to be obtained and made it possible to root out process obstacles which had been contributing to medication errors. It was im-

| S TRENG TH S AND WE AK NE SS E S
Studies of this nature are difficult to analyse using statistically robust techniques such as ANOVA and logistic regression. This limits the ability to infer the results to other organizations. The information presented, however, is important in that it shows how errors are distributed in this hospital system and also provides future direction for research. It also catalysed the nursing leaders in this organization with impetus to continue to proactively root out causes for medication errors. Additionally, it garnered important information on how errors trended as the work week progressed and quelled suspicions that the level of medication errors was markedly different on weekends and on the p.m. shift. Future research will help refine and focus on the continuing problem of medication administration errors.

| CON CLUS ION
As administrators and leaders, it is incumbent on us to act in a way that ensures an increased level of patient safety. The discussion of drug administration errors has assumed a very important place in contemporary healthcare dialogue. Systems that are designed to reduce errors have not proven to be completely successful and errors are still occurring. This "first look" at a hospital systems effort to more carefully study drug administration errors of omission and commission has revealed interesting trends as well as actionable data which have been brought to the bedside. There is no existing universal solution, and there is no reason to think that any system can be developed that is able to exclude the intrinsic human factors in medication administration. The mere act of seeking solutions to these problems will certainly unearth more focused opportunities for improvement. Nurses on the front lines of medication administration must be open to learning new strategies and adapting new ways to practice more safely. The overarching goal is for there to be no reason for any patient or people' loved one to be fearful about a medication mistake being made during a routine hospital admission.

ACK N OWLED G EM ENTS
While there was no monetary support allocated for this study, the Executive team at the medical centre(s) were very interested in the study results and were generous in allowing time for us to build the database, analyse and interpret relevant statistics and to prepare the manuscript.

CO N FLI C T O F I NTE R E S T
There was no conflict of interest in any phase of this research.

AUTH O R CO NTR I B UTI O N
Dr. Morelock was responsible for designing the study and the preparation of the manuscript. Mr. Kirk assisted in the statistical analysis and in the interpretation of the results.