TY - JOUR
AU1 - Strykowski,, Jill
AU2 - Hadsall,, Ron
AU3 - Sawchyn,, Bethany
AU4 - VanSickle,, Stacey
AU5 - Niznick,, Dan
AB - Abstract Purpose Results of a study at two hospitals to validate and test systems for bar-code-assisted medication administration (BCMA) are reported, including data on bar-code scanning failures and BCMA-related staff resource needs. Methods To prepare for BCMA implementation, pharmacy inventories at the two study sites were characterized by scanning product bar codes with a handheld device. The custom database built to house BCMA data was programmed to match National Drug Code (NDC) information on package labels with NDC information in the BCMA database. The data collected during inventory scanning were used (1) to quantify and address failed bar-code scans, (2) to predict the number of products that would require repackaging and relabeling to ensure accurate bedside scans, and (3) to estimate full-time equivalent (FTE) staff resources for BCMA-related repackaging work. Results During inventory assessment, scanning failures occurred with about 12.5% of products at the two pharmacy sites, mainly due to the absence of a bar-code label (49–53% of failed scans) or the inability to identify NDCs within the package bar code (38–46% of failed scans). It was determined that 8–10% of products inventoried at the two sites would require repackaging before dispensing, with associated technician resource needs estimated at 0.3–0.5 unit of FTE labor. Conclusion The results of inventory scanning revealed that most failed scans were attributable to the lack of a bar code on some products or problems with NDC recognition by the BCMA database. After those problems were addressed, a three-month pilot test on one patient care unit indicated an overall scanning success rate of >96%. Am J Health-Syst Pharm. 2013; 70:154-62 Medication administration is a complex, error-prone step in the medication-use process, accounting for an estimated 34% of all adverse drug events in hospital environments. 1 Bar-code-assisted medication administration (BCMA) is safety technology that has been shown to reduce medication administration errors by as much as 54– 86%. 2,3 When BCMA is coupled with computerized prescriber order entry and an electronic medication administration record, a technological loop extending from the transmission of the order to the administration of the medication at the bedside is closed. In a recently published pre–post study, the use of these technologies was shown to produce a 50.8% reduction in potential adverse drug events (from a rate of 3.1% to a rate of 1.6%). 4 Preventable adverse drug events are estimated to cost hospitals over $2 billion annually due to associated increases in morbidity and mortality. 5 Properly implemented and managed medication-use technologies have the potential to moderate these costs. In addition to the prospect of major patient safety and cost-control benefits, there are direct financial incentives for health systems to adopt these technologies as a result of the 2009 American Recovery and Reinvestment Act (ARRA), which established “meaningful use” criteria to guide health systems in optimizing the use of electronic health records and related technologies; BCMA is one focus of the ARRA incentives. 6 Adoption of BCMA remains a challenging, multidisciplinary implementation process. An important part of project planning for BCMA implementation is estimating the resources necessary to repackage and relabel products to ensure accurate bar-code scans. Equally important is estimating the numbers of pharmacists and pharmacy technicians needed to execute and sustain the new work. We therefore sought to develop a database to help approximate these resources as part of an implementation project at two hospitals within a large health system. Although the electronic medical record (EMR) was the foundation for accurate bar-code scanning, it did not have the ability to store scanned data that could be analyzed for resource predictions. A standalone product was therefore devised. To guide the development of the product, we set out to understand, at a fundamental level, the history of the technology and associated implementation barriers. Background The foundation for adoption of BCMA technology was laid when the Food and Drug Administration (FDA) promulgated rules in 2004 that require all drug manufacturers to place a usable bar code on prescription medications, insulin, and over-the-counter products commonly used in hospitals. 7 The products dispensed must be source-coded with a linear bar code that contains a National Drug Code (NDC). The NDC is a unique three-segment, 10-digit number that identifies the labeler, the product, and the trade package code. 8 Medication product packages in the United States must contain an NDC, which serves as a universal product identifier for human drugs. Medication bar codes typically contain a single “check digit” after the NDC string; the digit ensures data integrity and is calculated from a mathematical formula. When a scan is conducted, an autocalculation to validate the code against the check digit is completed. Some bar codes contain a prefix digit or number string (e.g., a GS1 code) plus a check digit, for a total of 12 or more numbers. A 10-digit NDC can appear in one of three segmental configurations: 5-4-1, 4-4-2, and 5-3-2. The first segment is determined by FDA and identifies the manufacturer, repackager, or relabeler. The second segment is the product code and identifies a strength, dosage form, and formulation; this segment is determined by the manufacturer. The third segment denotes the package size and type and is also set by the manufacturer. The potential for confusion arises when commercial drug databases used by most hospital pharmacies add extra digits to the second and third segments because of a conflict with the standard of an 11-digit NDC specified by the Health Insurance Portability and Accountability Act of 1996. Most problematic are additions of zeros to the number string, which can, in effect, “reconfigure” the 10-digit NDC as read by a bar-code scanner. For example, 12345-0678-09 can be read as 12345-678-09 (the 5-3-2 configuration) or 12345-0678-9 (the 5-4-1 configuration). The NDC is directly linked to bar-code symbology via encoding and then printed on the unit-dose and bulk medication inventory. The issues surrounding the use of NDCs as the basis for bar coding multiply with the sheer number of manufacturers, repackagers, and relabelers, as well as the speed of bar-code generation necessary to maintain the supply chain. Adding to the issues is a lack of standardization of product coding; for example, an acetaminophen 325-mg tablet has multiple product codes, as there are many manufacturers of that dosage form. Therefore, using the NDC as the foundation for medication bar-code scanning produces significant BCMA implementation challenges. An unsuccessful bedside scan due to system inadequacies can lead to end-user workarounds, resulting in failure to realize the full benefits of the safety technology. Koppel et al. 9 characterized workarounds associated with implementation of BCMA over a seven-year time frame at five hospi-tals. 9,10 The team found that nurses overrode BCMA alerts for 4.2% of patients and for 10.3% of medications; the reasons for these actions were captured for further analysis. The most prevalent causes were that medications did not have accurate bar codes and that bar codes were missing or damaged in some fashion. A Dutch study validated these findings. 11 In that study, only 55.3% of medication administrations were verified by BCMA; one reason given for workarounds was the lack of bar coding on some medications. Clearly, the use of safety technology in preventing medication mishaps has potential limitations. A readable bar code on each and every medication will contribute to the success of the technology. High scanning success rates facilitate optimal nursing processes at the bedside by obviating workarounds associated with a poor scan rate. The purpose of the study described here was twofold: (1) to understand the causes of product-scan failures at two hospitals using the same BCMA technology and (2) to estimate the staff resources needed to ensure accurate scans before BCMA implementation and successful ongoing bedside scanning. Knowing the inconsistencies prevalent in NDC-based medication bar coding and the workarounds that can lead to suboptimal scan rates, we set out to understand variances between expected and actual scanning results. The primary aim was to capture the detail of variances in a custom-built database. The database was programmed to match NDCs within package bar codes with NDCs stored in the database through an iterative digit-checking algorithm. With the information captured during the NDC-matching process, the pharmacy inventory was described; ascertaining which products would likely be involved in scanning failures enabled the development of a priori corrections before dispensing. The scanned data in the custom database were then tested against the medication data in the hospitals’ vendor EMR to determine if the information provided by the relatively static custom database was as predictive of scanning success or failure as the more dynamic EMR data. Testing was conducted at a midsize community hospital and a smaller hospital within the same health system. Our secondary aim was an assessment of end-user success or failure in bedside scanning. Rates of failed bedside scans after BCMA implementation at the two hospitals were assessed; in addition, scanning failure rates during designated periods over a roughly three-month time frame were compared as a means of testing the accuracy of the predicted failure rates while ensuring that the technology was embedded in the nursing workflow. Methods Allina Hospitals & Clinics (AHC) is a 23,000-employee health system serving patients in the Minneapolis– St. Paul area and comprises 11 hospitals (4 metropolitan and 7 regional), over 50 clinics, and 15 community pharmacies. Mercy Hospital is a 271-bed licensed facility and 1 of the 4 metropolitan hospitals in the AHC system; it has a broad range of community offerings, including a comprehensive cardiology program, oncology services, a family care program, and a large pediatrics group and community emergency department. St. Francis Regional Medical Center (St. Francis RMC) is a 93-bed facility in the AHC system. The focus of the St. Francis RMC service line is family care, and the hospital has a large cancer center with a high proportion of oncology outpatients. The 11 hospitals of AHC have a standardized formulary and purchase products from one wholesaler. Not all products on the formulary are stocked at both sites. Decisions regarding the choice of a particular product manufacturer and the package size of a given product (e.g., bulk versus unit dose packaging) are made by each AHC site, so equivalent products and dosage forms purchased by different sites can have different manufacturers. Database development Using Microsoft Office Access 2003 (Microsoft Corporation, Redmond, WA), a proprietary database was developed. The database was programmed to work in tandem with a recently updated NDC database (First Databank Inc., South San Francisco, CA) and gather drug-specific information from the EMR (Epic Systems Corporation, Verona, WI). A wireless scanner with 2-D symbology capability (DS6878 Cordless 2D Imager, Motorola Solutions, Inc., Schaumburg, IL) was used to scan products in the pharmacy inventory, with the information captured in the database and housed on a laptop computer. All scans were performed using the product that most closely resembled the dosage unit typically administered at the bedside. With each bar-code scan, the First Databank database was searched for an exact NDC match with the NDC portion of the product bar code. Code was written for Access to allow for prefix and suffix truncation as the database checked for the NDC in any of the three valid configurations. This was a simple iterative process ( Figure 1 ). When a successful match occurred, the program produced a unique sound. In addition, each entry made for a scanned product captured the following data on the scan log: site (hospital), location of scan (pharmacy shelf, new inventory delivery), scanned NDC, corrected NDC, scan problem, internal unique identifier, name of drug, package description, package quantity, unit (e.g. tablet, capsule, volume of liquid), and user. If there was no match after several programmed iterations, the scanner produced a different sound, and the user manually entered the product information along with the reason for the failed scan in a popup box on the system display. The primary types of scanning failures were categorized as “no matching NDC found”; “no scannable label on dosage unit”; “needs relabeling, as product in unit-dose form but no bar code”; and “scanned outer package only.” Scanning schema Inventory scanning was conducted in two phases. The phase 1 scanning scheme relied on the metric of “inventory turns,” defined as the total cost of inventory on hand divided by the total purchase cost per year. At Mercy Hospital, the average inventory turn rate is 17; therefore, it was determined that daily scanning of all inventory deliveries over a one-month period would be necessary to capture information on most unique products. In fact, at the end of one month of scanning, data analysis demonstrated that data on only 61% of products in the inventory (including all new products arriving from the wholesaler during the scanning period) had been captured. The initial phase 1 scanning (January 2010) was conducted at the metropolitan hospital. In October 2010, the process was repeated at the regional site to capture comparative data. Products that were already repackaged by the pharmacy using the Medi-Dose System (The Medi-Dose Group, Ivyland, PA) were not included in the analysis of scanning rates described here, as they did not require further repackaging. However, the existing stock of such products was relabeled with appropriate custom labels containing a custom “internal NDC” for purposes of the study only; the corresponding Medi-Dose records were altered accordingly. During this custom-labeling process, 428 new records for quarter- and half-tablet doses, compounded products, and other items requiring unique relabeling were created in the Medi-Dose database. Although this work was a critical step toward BCMA implementation readiness, the associated scan results were not relevant to our investigation, as those data did not pertain to manufacturer-related NDC formatting variances. Phase 2 of the research project included the scanning of all pharmacy stock (i.e., not just daily inventory deliveries) at both Mercy Hospital (March 2011) and St. Francis RMC (April 2011). Full-inventory scanning was repeated six months later (September 2011) at Mercy Hospital to determine if any improvement in the accuracy of manufacturer-generated NDCs had occurred over time. During the phase 2 scanning period, product-code NDCs that could not be matched to NDCs in the Access database were evaluated against information in the EMR system as a means of validating scan failures. This approach was taken for two reasons. First, although the First Databank NDC database is similar to the NDC database linked to the EMR, the truncation schemes used to locate a match are quite different, as shown in Figure 1 . The Epic EMR system uses pattern-matching logic to parse numerical “regular expressions” that effectively mask the NDC within the entire string of product-code characters. The Access-based bar-code tracking tool uses a one-character-shift iteration scheme to find an exact NDC match. Second, while Epic system data at the two study sites are updated on a monthly basis, the custom Access database was updated only at the beginning of the first full-inventory scans (i.e., March 2011 at Mercy Hospital and April 2011 at St. Francis RMC); however, even though it contained less-current NDC data, the Access-based system was chosen for our investigation, as it was deemed that the Epic system would not allow the storage of real-time scanning data in a readily retrievable manner. Comparing scanning success rates with the Access- and Epic-based methods allowed for validation of the former and modeling of actual BCMA conditions via a t test and calculation of 95% confidence intervals. Figure 1. Open in new tabDownload slide Comparison of two National Drug Code (NDC)-matching processes as applied to Hospira 1000 mL 0.9% sodium chloride with 20 meq potassium chloride (matching portions of NDC string in bold). In this example, the package NDC is 010030409711509 and the database NDC is 0409711509. Figure 1. Open in new tabDownload slide Comparison of two National Drug Code (NDC)-matching processes as applied to Hospira 1000 mL 0.9% sodium chloride with 20 meq potassium chloride (matching portions of NDC string in bold). In this example, the package NDC is 010030409711509 and the database NDC is 0409711509. The results of Access-based scanning were tabulated for the time intervals shown in Table 1. The data collected included unique products, total scan problems, products requiring new repackaging, existing repackaged products, prerepackaging and postrepackaging needs, and scan discrepancies that could not be resolved through either the Access- or Epic-based code-matching method. Products purchased before the study in bulk, typically oral tablets and capsules, were excluded from scanning because they were repackaged with the Medi-Dose system and were therefore assigned institution-specific bar codes. Products requiring new repackaging were segregated into the following subcategories: “NDC not found” (i.e., bar code present but NDC not readable), “no bar code on package,” and “poor readability of the bar code.” Given that the true test of a successful scan is at the bedside, the final test of the Access-based scanning approach involved evaluating how often an accurate scan was achieved by end users during medication administration. The percentage of bedside scans resulting in scanning failures was measured three times at four-week intervals on a cardiology step-down unit at Mercy Hospital designated as a pilot-testing unit. The Access database predicted a scanning failure rate of 9.8% for products scanned during the month of March 2011; those products were either repackaged or relabeled before administration to patients on the pilot unit. Bedside scanning success rates were evaluated during weeks 4–8, 8–12, and 17–20 after implementation, and monthly scanning success rates were calculated in terms of both absolute values and 95% confidence intervals via an unpaired t test in order to determine if addressing predicted scanning problems by repackaging translated to successful scans at the bedside. Results and discussion Phase 1 results During the phase 1 partial-inventory scanning, 922 unique products were scanned at Mercy Hospital, with 510 unique products scanned at St. Francis RMC; scan failures captured by the Access database were 115 and 64, respectively ( Table 1 ). The scanning failure rate for these products was 12.5% at both hospitals. Packaging-related scanning failures during phase 1 were characterized as shown in Figure 2 , with the highest number of failures involving injectable products. Of note, while there were 115 failed scans at Mercy Hospital, only 83 products required new repackaging; the explanation for this variance was that all products involved in scanning failures were not immediately dispensed: 32 unique products remained in the pharmacy for final product preparation. Actual drug-use data on the 83 repackaged unique products for the same one-month time frame indicated that a total of 6831 doses, or approximately 340 unit doses per day, were repackaged or relabeled. A high-level time–motion study conducted during phase 1 demonstrated that repackaging 100 tablets required 15 minutes of technician time, or about 0.1 unit of full-time equivalent (FTE) labor. Phase 2 results During the full-inventory scanning phase of the study, unique scanning failures at Mercy Hospital numbered 148 in March 2011 and 149 in September 2011, corresponding to failure rates of 9.8% and 9.4%, respectively. There were 93 unique scanning failures at St. Francis RMC, for a failure rate of 8.1%. Forty-three of the products involved in phase 2 scanning failures at Mercy Hospital (March 2011) and St. Francis RMC (April 2011) were identical. These scanning failures were categorized by product manufacturer. Nine manufacturers accounted for 4 or more unique products involved in failed scans; seven manufacturers accounted for 30 unique products associated with scanning failure rates of ≤3%, demonstrating consistently accurate NDCs. To assess repackaging resource demands, pharmacy technicians at Mercy Hospital logged the number of repackaged doses and associated time requirements for each medication “batch.” During March 2011, 59 unique repackaging tasks and a total of 3641 repackaged doses were logged, with a total repackaging time of 649.2 minutes. The repackaging time per dose was estimated at 10.7 seconds (0.1783 minute), corresponding to 17.83 minutes per 100 doses, which was roughly in line with the phase 1 repackaging estimate of 15 minutes per 100 doses. Table 1. Results of Bar-Code Scanning of Pharmacy Inventory, by Study Phase, Hospital, and Scanning Perioda Variable Phase 1 Phase 2 Mercy Hospital (January 2010) St. Francis RMC (October 2010) Mercy Hospital (March 2011) St. Francis RMC (April 2011) Mercy Hospital (September 2011) Total no. unique products 922 510 1507 1146 1581 No. (%) unique-product scanning failures 115 (12.5) 64 (12.5) 148 (9.8) 93 (8.1) 149 (9.4) Variable Phase 1 Phase 2 Mercy Hospital (January 2010) St. Francis RMC (October 2010) Mercy Hospital (March 2011) St. Francis RMC (April 2011) Mercy Hospital (September 2011) Total no. unique products 922 510 1507 1146 1581 No. (%) unique-product scanning failures 115 (12.5) 64 (12.5) 148 (9.8) 93 (8.1) 149 (9.4) a RMC = Regional Medical Center. Open in new tab Table 1. Results of Bar-Code Scanning of Pharmacy Inventory, by Study Phase, Hospital, and Scanning Perioda Variable Phase 1 Phase 2 Mercy Hospital (January 2010) St. Francis RMC (October 2010) Mercy Hospital (March 2011) St. Francis RMC (April 2011) Mercy Hospital (September 2011) Total no. unique products 922 510 1507 1146 1581 No. (%) unique-product scanning failures 115 (12.5) 64 (12.5) 148 (9.8) 93 (8.1) 149 (9.4) Variable Phase 1 Phase 2 Mercy Hospital (January 2010) St. Francis RMC (October 2010) Mercy Hospital (March 2011) St. Francis RMC (April 2011) Mercy Hospital (September 2011) Total no. unique products 922 510 1507 1146 1581 No. (%) unique-product scanning failures 115 (12.5) 64 (12.5) 148 (9.8) 93 (8.1) 149 (9.4) a RMC = Regional Medical Center. Open in new tab A detailed stratification of reasons for unique-product scan failures is shown in Table 2. New repackaging needs captured by the database at Mercy Hospital were 134 and 127 products in March and September 2011, respectively; at St. Francis RMC, 85 products scanned in April 2011 required repackaging. Those values did not correspond precisely to failed-scan counts, as not every scanning failure recorded in the database resulted in product repackaging or relabeling. As in phase 1 of the study, the phase 2 full-inventory scanning included sterile and nonsterile product preparations not intended for administration in the scanned form. For example, among the products involved in the 148 failed scans at Mercy Hospital in March 2011, only 134, or 8.9% of the total of 1507 scanned unique products in the inventory, needed to be repackaged before dispensing to ensure a successful bedside scan. Reasons for failed scans. Further analysis of the scanning failures revealed that the primary reason was the lack of a bar code on the package. During phase 2, that problem accounted for 49–58% of unique-product scanning failures. The second most common reason for failed scans was the lack of a corresponding NDC in the Access database; during phase 2 of the project, this accounted for 33.3–46.3% of total unique-product scanning failures at the two hospitals. Scanning success rates. As previously mentioned, when the Access-based method resulted in scanning failures, the products involved were scanned using the Epic system’s code-matching method. The difference in success rates with the two scanning methods was tested for significance using a t test of data collected in March and September 2011 ( Table 2 ). In March 2011, before full-inventory scanning was conducted at St. Francis RMC and Mercy Hospital, the custom Access database was updated; there was no significant difference in scanning success rates with the Access- and Epic-based methods during that time period. In the months preceding phase 2 scanning at Mercy Hospital in September 2011, the Epic system continued to receive updates while the Access database was not updated. Consequently, when the phase 2 scanning was conducted, the scanning success rates documented by the Access and Epic databases differed significantly ( p < 0.001). Figure 2. Open in new tabDownload slide Phase 1 bar-code scanning failures categorized by packaging type. Figure 2. Open in new tabDownload slide Phase 1 bar-code scanning failures categorized by packaging type. Table 2. Failed Bar-Code Scans During Phase 2 Full-Inventory Scanning, by Type and Scanning Period a Variable Mercy Hospital (March 2011) St. Francis RMC (April 2011) Mercy Hospital (September 2011) Fraction (%) of scanned products requiring new repackaging 134/1507 (8.9) 85/1146 (7.4) 127/1581 (8.0) Reasons for failed scans, as fraction (%) of scanned products NDC not found in database 57/148 (38.5) 31/93 (33.3) 69/149 (46.3) No bar code on package 78/148 (52.7) 54/93 (58.0) 73/149 (49.0) Poor readability of bar code 13/148 (8.8) 8/93 (8.6) 7/149 (4.7) Variable Mercy Hospital (March 2011) St. Francis RMC (April 2011) Mercy Hospital (September 2011) Fraction (%) of scanned products requiring new repackaging 134/1507 (8.9) 85/1146 (7.4) 127/1581 (8.0) Reasons for failed scans, as fraction (%) of scanned products NDC not found in database 57/148 (38.5) 31/93 (33.3) 69/149 (46.3) No bar code on package 78/148 (52.7) 54/93 (58.0) 73/149 (49.0) Poor readability of bar code 13/148 (8.8) 8/93 (8.6) 7/149 (4.7) a RMC = Regional Medical Center, NDC = National Drug Code. Open in new tab Table 2. Failed Bar-Code Scans During Phase 2 Full-Inventory Scanning, by Type and Scanning Period a Variable Mercy Hospital (March 2011) St. Francis RMC (April 2011) Mercy Hospital (September 2011) Fraction (%) of scanned products requiring new repackaging 134/1507 (8.9) 85/1146 (7.4) 127/1581 (8.0) Reasons for failed scans, as fraction (%) of scanned products NDC not found in database 57/148 (38.5) 31/93 (33.3) 69/149 (46.3) No bar code on package 78/148 (52.7) 54/93 (58.0) 73/149 (49.0) Poor readability of bar code 13/148 (8.8) 8/93 (8.6) 7/149 (4.7) Variable Mercy Hospital (March 2011) St. Francis RMC (April 2011) Mercy Hospital (September 2011) Fraction (%) of scanned products requiring new repackaging 134/1507 (8.9) 85/1146 (7.4) 127/1581 (8.0) Reasons for failed scans, as fraction (%) of scanned products NDC not found in database 57/148 (38.5) 31/93 (33.3) 69/149 (46.3) No bar code on package 78/148 (52.7) 54/93 (58.0) 73/149 (49.0) Poor readability of bar code 13/148 (8.8) 8/93 (8.6) 7/149 (4.7) a RMC = Regional Medical Center, NDC = National Drug Code. Open in new tab To determine whether any improvement in the scannability–readability of manufacturer-generated NDC data occurred over time, during phase 2 of the research all inventory was scanned at Mercy Hospital in March and September 2011; during those periods, 138 (9.2%) and 96 (6.1%) of unique-product scans resulted in scanning failures documented in the Access and Epic databases, respectively. There was a significant decrease in the rate of failed scans from March to September ( p < 0.001), suggesting that the accuracy of NDC data in the Epic database had improved over time. Another primary reason for failed scans was poor bar-code readability. Information in the Access database showed that 4.7–8.8% of documented unique-product scanning failures were due to poor bar-code readability. The majority of these failures involved the products of one manufacturer of suppository and topical products. In general, our data demonstrated that the majority of failed scans were related to the data within the bar code rather than the physical properties of the dosage form or the scanner technology. Pilot study results A successful scan at the bedside is the ultimate patient safety goal of BCMA; all other assessments up to that point are surrogate measures of success. BCMA was implemented at Mercy Hospital during the study period, allowing us to capture data on end-user scanning success ( Table 3 ). Our BCMA system is programmed to capture four reasons for scanning failures: “bar code unreadable,” “emergency,” “patient’s own [medication] supply,” and “scanner not available.” The majority of failures are documented by end users as “bar code unreadable,” suggesting that perhaps that categorization is used as a “catchall” for any reason a scan is not completed. A 95% confidence interval was determined for the calculated percentage of scanning failures categorized as being due to an unreadable bar code. Given the large sample size (>8000 scanned products during all three scanning periods), the confidence interval is narrow. An overall scanning success rate of >96% was maintained for six months after BCMA implementation. In preparation for implementation of BCMA within our health system, we successfully developed a database to quantify the proportion of inventory needing repackaging or relabeling. To our knowledge, this is the only published evaluation of a method for assessing key BCMA readiness metrics (repackaging–relabeling needs and associated labor requirements) that has included end-user validation. Our results demonstrated that regular vendor NDC data downloads and the ability of the BCMA system to generate a customized bar-code label in the event of a nonrecognizable package bar code are core components of predicting the rate of accurate scans. At our institution, the Epic system is updated on a monthly basis, whereas the Access database was updated only at the beginning of the March 2011 scanning session. During the March 2011 inventory scans (i.e., soon after the Access system received an NDC data update), we did not observe a significant difference in the scanning results obtained with the Epic tool and with the Access tools (failure rates of 9.2% and 9.9%, respectively). Six months later, however, we observed a significant difference between the Epic tool and the Access tool with regard to documented scanning failure rates (9.4% versus 6.1%, p < 0.001), likely due to ongoing Epic system downloads of updated NDCs that were not entered into the static Access database. Table 3. Results of Bedside Scanning on Cardiology Step-down Unit at Mercy Hospital After BCMA Implementation a Variable Postimplementation Period June 20–July 17 (Weeks 4–8) July 18–August 14 (Weeks 9–12) September 12–October 9 (Weeks 17–20) No. medication administrations 8589 8258 9067 No. (%) scanned products with unreadable bar codes 263 (3.06) 277 (3.35) 355 (3.915) % (95% confidence interval) failed scans for which “bar code unreadable” was listed as reason for failure 96.94 (96.944–96.936) 96.65 (96.646 –96.654) 96.08 (96.076–96.084) Variable Postimplementation Period June 20–July 17 (Weeks 4–8) July 18–August 14 (Weeks 9–12) September 12–October 9 (Weeks 17–20) No. medication administrations 8589 8258 9067 No. (%) scanned products with unreadable bar codes 263 (3.06) 277 (3.35) 355 (3.915) % (95% confidence interval) failed scans for which “bar code unreadable” was listed as reason for failure 96.94 (96.944–96.936) 96.65 (96.646 –96.654) 96.08 (96.076–96.084) aBCMA = bar-code-assisted medication administration. Open in new tab Table 3. Results of Bedside Scanning on Cardiology Step-down Unit at Mercy Hospital After BCMA Implementation a Variable Postimplementation Period June 20–July 17 (Weeks 4–8) July 18–August 14 (Weeks 9–12) September 12–October 9 (Weeks 17–20) No. medication administrations 8589 8258 9067 No. (%) scanned products with unreadable bar codes 263 (3.06) 277 (3.35) 355 (3.915) % (95% confidence interval) failed scans for which “bar code unreadable” was listed as reason for failure 96.94 (96.944–96.936) 96.65 (96.646 –96.654) 96.08 (96.076–96.084) Variable Postimplementation Period June 20–July 17 (Weeks 4–8) July 18–August 14 (Weeks 9–12) September 12–October 9 (Weeks 17–20) No. medication administrations 8589 8258 9067 No. (%) scanned products with unreadable bar codes 263 (3.06) 277 (3.35) 355 (3.915) % (95% confidence interval) failed scans for which “bar code unreadable” was listed as reason for failure 96.94 (96.944–96.936) 96.65 (96.646 –96.654) 96.08 (96.076–96.084) aBCMA = bar-code-assisted medication administration. Open in new tab We also tested the iterative NDC-matching methods used by the Access and Epic tools. Although the logic behind the two digit-checking methods is different, scanning success rates with the Access and Epic tools were not significantly different as long as both systems’ NDC information was current. We tested the Access database on fast-moving inventory over a one-month period and documented a 12.5% scanning failure rate at both a small regional hospital and a larger metropolitan hospital within the same health system. One may predict that with a systemwide formulary, the results reported here could be anticipated. In fact, however, the inventories stocked at the two hospitals are different, the hospitals’ clinical service lines are prioritized differently, and the average patient acuity level varies between the two hospitals. The general usability of our method is not dependent on the medication administration record, NDC database, or scanner used, and the digit-checking methodology is not vendor specific. When searching for an exact NDC match, our tool uses a simple serial one-character-shift process, and the effectiveness of the method is independent of other technology variables. Our study demonstrated that physical problems were a minor reason for scanning failures; only suppository dosage forms and some topical products were unscannable for this reason, with associated failure rates in the range of 4.7–8.8%. Instead, problems with the NDC not being found by the database (33.3–46.3% of failed scans in phase 2) and no bar code found on the package (49–58% of failed scans in phase 2) were primary reasons for new repackaging needs. Before the transition to BCMA, the Mercy Hospital pharmacy repackaged 525 unique products and St. Francis RMC repackaged 391 unique products, or 24.9% and 25.4% of their total inventories, respectively. After the BCMA readiness assessment, inventory scanning at both hospitals identified an additional 261 products needing repackaging at Mercy Hospital (134 products in March 2011 and 127 products in September 2011) and an additional 85 products to be repackaged at St. Francis RMC. These products accounted for an overall increase in repackaging needs of 22–31%. We calculated repackaging labor needs in both phase 1 and phase 2 of the study, and the results were remarkably similar (approximately 0.2–0.3 unit of FTE technician labor). Phase 1 results were estimated by direct observation that 100 tablets could be repackaged using the Medi-Dose system within 15 minutes. Our database captured the products that would need to be repackaged, and these products were matched to actual utilization data within Epic. At 6831 doses expected to be manipulated from an estimated 115 unique product failures during Phase 1 at Mercy Hospital, 0.1 unit of FTE technician time would be needed for this additional function. During phase 2, our pharmacy technicians at Mercy Hospital were asked to document the time spent on each repackaged batch in order to estimate repackaging time. A total of 59 entries were logged for a total of 3641 doses over 649.2 minutes, yielding a repackaging time estimate of 0.1783 minute per dose or 17.83 minutes per 100 doses repackaged. Although the two estimates of BCMA-related technician resource needs were remarkably congruent, it is important to consider that the work involved in repackaging does not depend solely on the number of doses but also on the number of product “batches”; therefore, the FTE value is an underestimate of total additional resource needs. A substantial amount of total repackaging time is spent making new labels, and making one label for one bottle potentially takes considerably more time than repackaging 100 tablets of the same product. Many of our products are one-half and one-quarter tablets that need to be handled first and then repackaged. Additionally, in practice, pharmacy buyer time must be taken into account. The buyer completes a validation scan on all product received from the wholesaler and corrects product designations in our EMR. This work accounts for an additional 0.1–0.2 unit of FTE staff time, or an estimated 30 minutes per wholesaler delivery twice a day during 10 days in a two-week pay period. Our database did not directly capture pharmacy buyer work performed outside the study time frame, as time constraints did not allow for this work to be captured during the study period. Finally, although the terms relabeling and r epackaging are used interchangeably (in our study and generally), operationally there is a difference. Relabeling refers to labels that need to be printed from the software and attached to products that are in individual units. Depending on the capabilities of the institution, it may be unreasonably time-consuming to relabel unit-dose medications, and it may be more cost-effective to repackage bulk bottles. Although we substituted one term for another, other institutions may have different time requirements for each function. Implementation of BCMA at the bedside was successful, with the percentage of medications scanned at the bedside found to be >96% (with a very narrow confidence interval) over the entire study period at Mercy Hospital. St. Francis RMC implemented BCMA in 2012. Our database predicted that 8–10% of the hospital’s inventory will need to be repackaged, with an overall 22–31% increase in repackaging needs and an estimated additional requirement of 0.3–0.5 unit of FTE technician labor. As BCMA implementation within the 11-hospital system continues, the resource estimates resulting from our study will help with budgeting and planning. Our organization is fortunate to have a depth of information technology support that may not be available in a smaller health system. We were able to build the database and update the First Databank data download using internal resources. We recognize that not all inventory is exactly the same temporally, and the results support this assertion. At Mercy Hospital, 148 and 149 scanning failures were documented in March 2011 and September 2011, respectively, yet 81 failed scans during both periods involved identical products; moreover, eight products scanned successfully in September had not scanned successfully in March. A limitation of the study is that drug stock is dynamic, in particular in an era of extreme drug shortages. The results of the evaluation confirmed that the NDC identifier continues to be fraught with standardization issues. FDA recognizes the challenge and recently called for field comments on improving barcode technologies in its review of the agency’s 2004 “Bar Code Final Rule.” 12 As we work through the challenges of NDCs, we may encounter new struggles when we employ technology such as radiofrequency identification and pursue other opportunities to improve the medication-use system. 5 Conclusion The results of inventory scanning revealed that most failed scans were attributable to the lack of a bar code on some products or problems with NDC recognition by the BCMA database. After those problems were addressed, a three-month pilot test on one patient care unit indicated an overall scanning success rate of >96%. Footnotes Margaret Schmidt, Pharm.D., M.B.A., is acknowledged for her assistance in shaping the concept of the study described in this article and her leadership in implementing bar-code-assisted medication administration at Allina Health; and Heather Britt, Ph.D., is acknowledged for her support in drafting the associated research grant request. The authors have declared no potential conflicts of interest. Supported by a 2009 Medication Safety Team grant from the ASHP Research and Education Foundation. References 1 Bates DW Cullen DJ Laird N et al. Incidence of adverse drug events and potential adverse drug events. Implications for prevention . JAMA . 1995 ; 274 : 29 – 34 . Google Scholar Crossref Search ADS PubMed WorldCat 2 Neuenschwander M Cohen MR Vaida AJ et al. Practical guide to bar coding for patient medication safety . Am J Health-Syst Pharm . 2003 ; 60 : 768 – 79 . Google Scholar Crossref Search ADS PubMed WorldCat 3 Paoletti RD Suess TM Lesko MG et al. Using bar-code technology and medication observation methodology for safer medication administration . Am J Health-Syst Pharm . 2007 ; 64 : 536 – 43 . Google Scholar Crossref Search ADS PubMed WorldCat 4 Poon EG Keohane CA Yoon CS et al. Effect of bar-code technology on the safety of medication administration . N Engl J Med. 2010 ; 362 : 1698 – 707 . Google Scholar Crossref Search ADS PubMed WorldCat 5 Cescon DW Etchells E . Barcoded medication administration: a last line of defense . JAMA . 2008 ; 299 : 2200 – 2 . Google Scholar Crossref Search ADS PubMed WorldCat 6 Centers for Medicare and Medi-caid Services . www.cms.gov/EHR IncentivePrograms (accessed 2011 Nov 19). 7 Bar code label requirement for human drug products and biological products . Final rule. Fed Regist. 2004 ; 69 : 9119 – 71 . PubMed WorldCat 8 Food and Drug Administration . National Drug Code database background information . www.fda.gov/drugs/development approvalprocess (accessed 2011 Aug 19). 9 Koppel R Wetterneck T Telles JL et al. Workarounds to bar code medication administration systems: their occurrences, causes, and threats to patient safety . J Am Med Inform Assoc. 2008 ; 15 : 408 – 23 . Google Scholar Crossref Search ADS PubMed WorldCat 10 Diconsiglio J . Creative ‘work-arounds’ defeat bar coding safeguard for meds. Study finds technology often doesn’t meet the needs of nurses . Mater Manage Health Care . 2008 ; 17 : 26 – 9 . WorldCat 11 Van Onzenoort HA van de Plas A Kessels AG et al. Factors influencing bar code verification by nurses during medication administration in a Dutch hospital . Am J Health-Syst Pharm. 2008 ; 65 : 644 – 8 . Google Scholar Crossref Search ADS PubMed WorldCat 12 Health Industry Washington Watch . FDA announces review of bar code final rule . www.healthindustrywashingtonwatch.com/2011/11/articles/regulatory-developments/hhs-developments/fda-announces-review-of-bar-code-final-rule/ (accessed 2012 Jan 2). Copyright © 2013 by the American Society of Health-System Pharmacists, Inc. All rights reserved.
TI - Bar-code-assisted medication administration: A method for predicting repackaging resource needs
JF - American Journal of Health-System Pharmacy
DO - 10.2146/ajhp120200
DA - 2013-01-15
UR - https://www.deepdyve.com/lp/oxford-university-press/bar-code-assisted-medication-administration-a-method-for-predicting-z0EgKnD6BH
SP - 154
VL - 70
IS - 2
DP - DeepDyve
ER -