Several reports have shown that medication errors can have a serious and sometimes fatal impact on patient outcomes.1–3 Bar-coded medication administration (BCMA), a technology aimed at decreasing medication administration errors, has been successfully implemented at a number of hospitals. The Leapfrog Group and the Veterans Administration’s National Center for Patient Safety have identified BCMA as an initiative that helps hospitals improve patient safety4(p24) and aid in automation of medication inventory and billing.4(p152) According to a study conducted at an academic medical center, BCMA was found to reduce medication errors by 41.4%.5 Computerized physician order entry (CPOE) aided the process by allowing pharmacists to verify medication orders in a timely fashion and deliver them to the unit so nursing can administer them with the help of BCMA, closing the loop on medication safety. Anecdotal evidence at the hospital studied here has been bearing this out: nurses have stated, for example, that “BCMA does catch errors,” citing instances where the technology has prevented a medication error.
Emergency department (ED) patients may be at a higher risk of medication errors than inpatients.6 However, BCMA does not seem to be widely used in the ED. Blank et al7 have recommended a “back to basics” approach to reducing ED medication errors, including checking the five rights three times, but BCMA was not among the solutions they recommended for the ED. A search of the literature revealed little mention of BCMA use in the emergency room except to note that BCMA was not used in the ED.8,9
Bar-coded medication administration was implemented throughout the hospital of this study in 2010. The technology has had consistently higher utilization on some floors than on others, suggesting that it is easier to use BCMA on some floors than on others. Bar-coded medication administration was used to chart medications 95.6% of the time on two medical-surgical units during the 6-month period from January 1 through June 30, 2011. The ED was more than 20% lower (74.7%). Anecdotal comments revealed staffs’ varying opinions. For example, one ED nurse said “I love the concept, but I hate the implementation.” The purpose of this study was to explore the factors that contribute to lower BCMA utilization in the ED.
Bar-coded medication administration was implemented in 2010 at a 637-bed tertiary-care hospital with medical-surgical, oncology, cardiac, maternal/child, pediatric, psychiatric, and special care inpatient units and a 64-bed emergency room. Nurses and respiratory therapists used wireless handheld devices, also known as PDAs, for medication and patient wristband bar-code scanning. The PDAs sent messages to communication servers via a wireless network. The communication servers processed messages from the PDAs and communicated with the BCMA database, which held patient and medication information. The BCMA system was connected to other systems, including the hospital information system (HIS) and the pharmacy information system (PIS). The system architecture is depicted in Figure 1.
Patient and medication data were maintained in multiple databases, including the HIS, BCMA, and PIS. Hospital information system functions included CPOE, an electronic medication administration record (eMAR) and some clinical documentation such as flow sheets and structured notes. Other pieces of clinical documentation were done on paper. Pharmacists verified patient medications in the PIS. Medication verifications were transmitted from the PIS to the HIS and BCMA systems.
Rollout of the Bar-Coded Medication Administration System
The project began with an upgrade of the wireless network in 2008. Information Services developed and tested the application in 2009. Hospital-wide rollout began in January 2010 with a short-stay unit. Three more units went live in March. Starting in May 2010, two units went live every 2 weeks, ending with the ED in November 2010. Users were trained in 2-hour, nonmandatory classes that covered operation of the wireless handheld devices and bar-code scanning techniques. There was a dedicated technical team, including nursing and pharmacy analysts, a medication safety nurse and pharmacist, training specialists, project managers, helpdesk staff, and hardware technicians. In addition, two RNs, five upper-level undergraduate, and two graduate-level nursing students provided 24/7 on-site support and training in the use of the PDAs during the 2 weeks following each unit’s go-live. The support team played a critical role in the success of the rollout, more specifically, the training of users who were not able to attend classes, reinforcing learning for other users, entering problem reports, escalating issues to the technical team, and answering technical questions. The on-site technical team did not advise nurses regarding clinical matters; they served as experts on the use of the system, assisting users in getting logged onto the system and showing them how to use the PDAs to scan the small bar codes on many medications. Bar-coded medication administration utilization and bar-code override reports enabled the support team to identify and assist users who were challenged by the new technology.
Data Collection and Analysis
This article focused on collecting and analyzing data from the HIS regarding the use of the BCMA covering the period from January 1, 2011, to June 30, 2011. Only medications that were given to a patient were included in the retrieved data; doses charted as “not given” were not included. No identifiable patient data were collected; permission was requested from patients and nurses whenever medication administrations were observed. The proposal of this project was reviewed and approved by the institutional review board.
The medical-surgical units selected for comparison had roughly the same number of beds as the ED. The ED has a total of 64 beds, but only the ED’s acute care, pediatric, psychiatric, and clinical decision units (58 beds) were using BCMA at the time of this study and were included in the data presented here. There were 73 beds in the two medical-surgical units used as the comparison floors: a 29-bed geriatric unit and a 38-bed general medical-surgical unit, and its 6-bed step-down unit.
Utilization reports were developed during the time of the rollout to identify trends and assist users who were challenged with the new technology. Data for this study were collected using SQL (Structured Query Language). The utilization reports were extended for the purpose of this study to include medication frequencies, administration routes, the number of medications in each medication administration occurrence, and the percentage of occurrences where only one medication was given.
The fragmented system architecture was a challenge when collecting and analyzing the data. Utilization data, including the data reported in Table 1, were obtained from the HIS. There was a BCMA system limitation that rendered utilization statistics obtained from the BCMA system inaccurate: medications given with BCMA but later edited in the HIS were marked in the BCMA database as charted in the HIS. For example, a pain medication and the initial pain score could be documented with BCMA, but nurses had to go back into the HIS to note pain ratings given by patients after the medication had had time to take effect. When viewed in the BCMA database, these tasks appeared to be charted in the HIS, making utilization statistics obtained from the BCMA database appear lower than they actually were.
Other data, such as bar-code overrides and wrong medication alerts, were stored exclusively in the BCMA database. Many wrong medication alerts were valid, but some were due to a lack of synchronization between systems. Observations and interviews with nurses revealed that nurses were forced to override bar-code scans when invalid medication alerts occurred, weakening the safety of the system. These factors made it difficult to assess the accuracy of the data, and ultimately the efficacy of BCMA.
The study looked at the number of medications given in each medication administration occurrence, “average number of medications per occurrence” and “occurrences with only one medication”; data are shown in Table 1. Multiple medications could be charted in one transaction (ie, one medication administration occurrence) when BCMA was used, and this made it possible to retrieve the number of medications being administered at the time of that medication administration occurrence. Hospital information system charting on the eMAR does not give an accurate time of the administration of a medication as BCMA does, so this method could not be extended to include medications charted in the HIS.
The data from the collected queries (ie, utilization, medication frequencies and routes, and the number of medications administered in each occurrence) were processed using descriptive statistics. Audits of the data, observations of medication administrations, and interviews with nurses were used to validate the data and to organize the results describing the barriers to BCMA use.
The findings revealed limitations and medication order patterns that made BCMA more difficult to use in the ED (medications charted without BCMA: 25.3%) than on the medical-surgical floors (medications charted without BCMA: 4.4%).
An order had to be in the system before BCMA could be used. This limited the use of BCMA in the ED because verbal or protocol orders were often needed for medical emergencies.
Only nurses and respiratory therapists could administer medications with BCMA. However, there was no provision for BCMA use when medications were dispensed by physicians so that patients would have enough medication to last until their prescriptions could be filled. These medications had to be documented in the HIS.
Most of the ED’s medication orders (72.6%) were “stat” (ie, medications given within 30 minutes of ordering) or “now” (ie, medications given within 1 hour of ordering). The medical-surgical units’ orders, on the other hand, were scheduled medication occurrences, with only 2.8% of medications administered “stat” or “now.” The differences were statistically significant (P < .001; Table 1). Orders entered in the HIS were sent to the BCMA system’s communication servers, processed, transmitted to the BCMA database, and then uploaded over the wireless network to the PDAs. The observations and interviews revealed that slow system synchronization occasionally resulted in invalid wrong medication alerts when nurses attempted to give medications before the order had been transmitted to the PDA. This occurred more often when “stat” or “now” medications were administered, and thus this was a more common challenge in the ED than on the medical-surgical floors.
A smaller percentage of oral medications were given in the ED than on medical-surgical floors. Medications with routes of “PO,” “chew,” or “sublingual” comprised 56.7% (112 719 of 198 673 doses) of the medical-surgical unit’s medications, but only 45.7% (28 736 of 62 938 doses) of the ED’s medication routes. The differences were statistically significant (P < .001; Table 1). Most oral medications can be charted with fewer steps than nonoral medications: unit-dosed packaging used for tablets or capsules lends itself to BCMA much more easily than do medications drawn from bulk containers where amounts being given must be entered. Injectable medications often come in vials with small, curved bar codes that frequently could not be read with the PDA’s embedded bar-code scanner. Mandatory observations such as route had to be entered for most nonoral medications, resulting in more steps to administer them.
Emergency department nurses tended to administer fewer medications at a time (average number of medications per occurrence with BCMA: 1.39) and did not enjoy the same efficiency of scale as did the medical-surgical nurses (average number of medications per occurrence with BCMA: 2.24). Almost three quarters (73.2%) of the medication administrations in the ED involved just one medication, whereas slightly more than half (55.3%) of the medication administrations on the medical-surgical floors involved one medication. Smaller numbers of medications per administration occurrence were less efficient since the overhead of cleaning the PDA, logging on, and viewing patient information screens took the same amount of time regardless of the number of medications being given at one time (Table 1).
One of the limitations of this study is that the data were not initially collected for research purposes. Existing databases were used, and the accuracy of the data was validated using audits of the data, medication administration observations, and interviews with nurses.
The data for comparison were compiled based on the number of beds in the units. It is possible that this is a limitation, because the unit cultures and nursing staff were different, and it is possible to assume that these differences had an impact when medical-surgical units were compared with the ED.
It is also important to emphasize that this study did not explore the impact of bar-code overrides on medication safety. Providers who were using the BCMA system but overrode bar-code scans were counted as utilizing the system.
The impact of alert fatigue was not possible to quantify precisely or identify during the observations. Emergency department nurses were more likely to receive at least one type of invalid, wrong medication alerts due to lack of system synchronization when “stat” or “now” medications were administered. The precise impact of these additional alerts would have required further investigation using more in-depth interviews.
Data analysis, observations of medication administrations, and interviews with nurses showed that numerous factors impacted BCMA utilization in the ED, including verbal and protocol orders, medications administered by others, unpredictable workflow, increased workload, alert fatigue, and hardware limitations. Some of these factors, such as workflow, more steps to chart medications with BCMA, and alert fatigue, have affected medical-surgical nurses as well as ED nurses. Hardware problems have impacted all users.
Verbal and Protocol Orders
Many BCMA systems, including the one discussed here, cannot be used until an order is entered into the HIS, and this feature limited BCMA’s usefulness during codes and other emergent situations. Nursing management recognized this limitation and excluded the use of BCMA in two of the studied hospital’s ED areas: (1) the critical care area, where patients require immediate life saving medications and orders are given verbally, and (2) triage, where nurses utilize protocol orders to administer a limited number of medications. However, this change of the practice introduced utilization problems when nurses floated from triage or critical care to ED areas where BCMA was used. Because nurses were not using BCMA in triage or critical care, they often forgot to use it when they floated to other areas. Furthermore, there were often patients in the ED’s acute care areas who needed immediate medications such as nitroglycerin or aspirin where verbal orders were given, and in these cases, it was not possible to utilize BCMA.
The Department of Veterans Affairs (VA) modified their software after this shortcoming was recognized. Wideman et al10 have mentioned a “Nursing Medication Order Button” on the BCMA screen that could be used to enter verbal orders in emergencies. This feature became a VA requirement for BCMA use in the intensive care unit. The medication “hot button” was implemented in 2002 as part of version 2 of the VA’s BCMA software.11
Medications Administered by Others
Not all medications at the hospital studied here were given by nurses or respiratory therapists. “Starter packs” were packages of medication given to patients by physicians to ensure adequate doses of medication to take before prescriptions were filled. These medications could not be documented with BCMA. This shortcoming affected the use of BCMA in the ED much more than other units in the hospital.
The ED environment is markedly different from medical-surgical inpatient units, where patients have longer stays, and most medications are delivered during scheduled medication administrations. Medication administrations on the medical-surgical units at the hospital studied were generally a much smoother process than medication administrations in the ED: the medical-surgical nurses went to the medication rooms at scheduled times such as 9:00 AM. They had received report from the night nurses and had some familiarity with most of the patients they were taking care of. Medications were prepared, PDAs were obtained, and medications were administered. There was more “flow” to the workflow on medical-surgical floors than there was in the ED.
Observations at the ED studied here revealed significant variations in the amount of time needed to administer medications. Oral medications in unit-dosed packaging, such as ciprofloxacin tablets, could be passed in 3 to 4 minutes. One observed administration that took 36 minutes involved two medications, including one that had to be drawn from a bulk container. The container was taken to the patient’s room so the bar-code could be scanned when the medication was given. Only one of the two medications was actually administered: the patient refused the medication drawn from the bulk container. However, both medications were charted as given with BCMA, and the nurse had to go into the HIS to correct the error. She also spent time in the room trying to fix a malfunctioning intravenous infusion pump, explaining procedures to the patient and the patient’s family and trying to coax the patient to accept the medication that was refused.
The work of all nurses is unpredictable to some degree, and all nurses would benefit from a BCMA system designed to recognize and compensate for the chaotic nature of nursing work. Cornell et al12 have talked about the chaotic nature of nursing workflow, where nurses are constantly switching from one task to another, often switching to a new task before completing an old one. Elganzouri et al13 performed a medication administration time study and concluded that a typical medication administration took at least 15 minutes, and nurses could be distracted or interrupted whenever they passed medications. Bar-coded medication administration systems must be robust enough to compensate for the task switching, distractions, and interruptions that are part of a nurse’s job.
Technology designed to support medication administration must be flexible enough to deviate from set sequences. Cain and Haque14 have noted that many information technology systems assume a linear 4-step workflow model and process where medication is ordered, verified, delivered, and administered. It is important to understand what happens when this process is disrupted, as it often is in critical care situations where there is no time to enter an order into the system or when the pharmacy cannot verify the order in time. Nurses seem to have a tendency to adapt their workflow to work around shortcomings in the system, unintentionally bypassing BCMA safeguards.
Improved patient safety comes at a price: there were more steps in medication administration workflow when BCMA was used. Table 2 shows that there were twice as many steps at the hospital studied here when charting a simple medication (ie, a medication with no mandatory observations or overrides) with BCMA compared with charting medications without using BCMA. Bargren and Lu15 noted a twofold increase in workflow when BCMA was used.
The small size of the PDA screen (240 × 320 pixels) severely limited the amount of information that could be displayed by the application studied here, forcing providers to view multiple screens for each medication administered. There may have been as many as 18 steps in the process when overrides and mandatory observations were entered. Figure 2 exemplifies the number of steps to administer more complicated medications at the hospital in this study. (Some steps may have involved multiple screens, so the administration of some medications could have involved more than 18 screens.)
In this study, approximately 50 000 medications were administered each week, and thus providers received hundreds of alerts. Most of the alerts were valid and prevented medication errors. However, some alerts were due to incorrectly configured medications, system latency, and system limitations. There is anecdotal evidence that users at the hospital studied have ignored valid alerts, thinking that the warning message presented by the BCMA system was not meaningful.
Incorrectly configured medications were a cause of invalid wrong medication alerts. An audit of 1111 alerts received during the week of February 24 to March 3, 2011, revealed that many of these alerts were due to medications that were not correctly configured in the BCMA system. There was a vigorous effort to adjust medication information in the BCMA, and the average number of weekly alerts dropped to less than 900.
Some invalid alerts were due to the system architecture. Medication orders were entered and processed in the HIS. Medication charting tasks were sent the BCMA system and then over the wireless network to the PDAs. This introduced system latency that sometimes caused the generation of invalid medication errors. The week of July 11 to 17, 2011, was arbitrarily selected for an audit of wrong medication alerts. Most of the 772 wrong medication alerts that week were valid, but some were due to lack of synchronization between systems: there were 29 invalid wrong medication alerts generated by the BCMA when nurses were attempting to administer medications within 15 minutes of order entry on the HIS. More than half of these medications were “stat” or “now” (eight “stat” medications and eight “now” medications). This type of invalid wrong medication alert was more likely to affect ED nurses than medical-surgical nurses because they administered more “stat” and “now” medications than did medical-surgical nurses. Two other randomly selected weekly audits (September 12 to 18, 2011, and September 26 to October 2, 2011) demonstrated a similar pattern.
Another type of invalid wrong medication alert was due to inflexible “medication mapping,” a limitation of the BCMA system. A wrong medication alert resulted when the nurse administered the correct medication in a dose or concentration that did not exactly match the original order. This often occurred when the pharmacy ran out of the prescribed dose and had to substitute an equivalent medication. Wrong medication errors would occur if there was, for example, a shortage of 15 mg ketorolac vials and 30 mg vials had to be substituted.
Bar-coded medication administration worked well for unit-dosed medications, but bulk containers were problematic. The BCMA system studied here contributed to nurses’ alert fatigue by generating three alerts when a multidose container was scanned. For example, a nurse administering liquid acetaminophen scanned the bottle and saw a warning to “Review Order—Please refer to the order text to determine the correct dose to type in the ‘You are charting’ field.” A second warning appeared after the patient’s wristband was scanned, “WARNING! One of the med(s) you are administering have doses that differ from the amount you scanned. Take extra care to administer the correct doses.” After dismissing the second warning, the nurse was faced with a screen where the medication was marked with two asterisks and this third warning in red at the bottom of the screen, “Verify the dose being administered is correct.” The three warnings tended to unnerve nurses when they were indeed administering the correct dose.
Bar-coded medication administration systems must be designed to limit alerts such that only clinically meaningful alerts are presented to users. Phalsalkar et al,16 in their research on human factors, have emphasized the importance of limiting these alerts. Barton17 has noted that alerts have become irritations that are often ignored rather than aiding in patient care. Emergency Care Research Institute (ECRI), a nonprofit healthcare research organization, has put alarm hazards at the top of their list of technology hazards for 2012.18
Wireless handheld devices were used exclusively for BCMA at the hospital studied here. The devices were very portable, allowing nurses to perform BCMA at the bedside in crowded semiprivate rooms. There were, however, drawbacks associated with this choice of hardware, including frequent loss of wireless connectivity, dead batteries, small screen size, limited bar-code scanning capability, and a high price (approximately $1800 per device). Infection control was an ongoing challenge, as nurses struggled to clean the devices and wash their hands before and after administering medications to each patient. Workstations with attached bar-code scanners would have been a more robust, cost-effective solution for newer rooms with adequate space and a physical layout such that nurses could have easily observed the display while scanning medications and patient wristbands. Early et al9 evaluated BCMA hardware and recommended workstations with bar-code scanners over wireless handheld scanners. Additional display space available on a workstation would (1) allow for the addition of the “Nursing Medication Order Button” recommended by Wideman et al10 for verbal orders and (2) enable more functionality of a given screen, reducing the number of “clicks” (taps on a PDA) needed to complete a medication administration. This would certainly be a superior solution when there are workstations in every patient room.
Verbal orders, medications administered by others, unpredictable workflow, increased workload, and alert fatigue were factors that affect ED nurses more than medical-surgical nurses. Hardware limitations affected medical-surgical nurses as well. The hospital studied here is planning the implementation of a new HIS with integrated BCMA and pharmacy functionality, which should eliminate some problems, particularly the invalid wrong medication alerts arising from lack of synchronization between databases. In the future, workstations with scanners will replace the handheld wireless devices where space permits, allowing more display space on full-sized screens, better reliability, improved bar-code scanning capability, and solving the infection control problems that arise when devices are brought in and out of rooms.
© 2013 Lippincott Williams & Wilkins, Inc.