Chapter 43. Prevention of Misidentifications
Heidi Wald, M.D.
University of Pennsylvania School of Medicine
Kaveh G. Shojania, M.D.
University of California, San Francisco School of Medicine
Machine-readable automatic identification (ID) systems, including bar codes, magnetic stripes, optical character recognition and radiofrequency labeling, have improved productivity and quality in diverse industries. Bar codes represent the oldest and most common of these machine-readable ID systems,1,2 and are widely used in industrial manufacturing, shipping and inventory tracking operations. Prior to bar coding, these processes would have involved keystroke entry of identification numbers, producing approximately one error in every 300 entered characters. In contrast, bar coding produces misidentification errors at rates ranging from one character in 15,000 to one character in 36 trillion.3
The use of bar coding in healthcare was first described over 30 years ago in clinical laboratories and blood banks.1,4 In 1984, Rappoport identified 3 areas for the use of automatic ID technology in lab medicine: patient identification, document identification, and specimen identification.1 However, in a 1987 survey by the American Hospital Association, the use of bar codes was most widespread in materials management departments, rather than in clinical application.5 Other areas in which hospitals employed bar codes at that time included the clinical laboratory, pharmacy, radiology, medical records and asset management. Despite the Health Industry Bar Code Council's call for standardization in the mid-1980s, the implementation of bar code technology has been stymied by lack of industry standards and failed cooperation among all stakeholders.2,6
As bar coding has the potential to substantially increase productivity and accuracy, one would expect it to be applied to important patient safety practices. Unfortunately, the published literature contains very little evidence regarding healthcare applications. In this chapter we focus on 4 areas in which bar coding shows promise for improving patient safety: patient identification, medication dispensing and administration, specimen handling, and medical record keeping.2,7-10
Machine-readable patient ID systems could replace conventional wrist-banding, and might reduce patient ID errors ranging from specimen collection to medication and blood product administration. Attempts to create such machine-readable ID systems were reported in the blood banking literature as early as 1977.11 Transfusion Medicine is particularly attuned to issues of patient identification, as specimen collection and blood product administration account for the majority of preventable transfusion errors.12,13
Most scenarios for patient identification involve the substitution or supplementation of the traditional wristband for one with a unique bar code patient identifier. All patient specimens, medications and released blood products then receive the patient's unique bar code ID. No procedure or treatment can occur unless the patient's ID is scanned with a portable scanner and matched with a bar code generated by the doctor's order. For example, a phlebotomist would carry the scanner, check the patient's ID against a bar coded specimen label or collection list, and draw blood only in the event of a match. Similarly, for administration or treatment, the patient's ID and the intended therapeutic would be scanned at the bedside with a portable reader. If a match exists, the transfusion or medication is allowed and the time and date are recorded and even transmitted directly to the hospital computer system. The nurse's bar code ID can also be scanned and a timed administration record can be created. If there is no match, an alarm is sounded, and the administration delayed until the problem is resolved. 9,14
Other technological means to reduce error in patient identification have been examined in the transfusion literature. Several researchers explored the use of a system providing a mechanical barrier to transfusion through a series of locks.15 This system appears to be cumbersome and easily circumvented. In addition, electronic blood banking, with point-of-care crossmatching and computer-controlled release of blood, has been examined for high volume transfusion areas such as the operating room or intensive care unit.16 Due to major barriers to large-scale implementation, these practices will not be discussed further.
Clinical laboratories have integrated bar codes in specimen handling with a great deal of success.3,8 Several authors have described the development of central laboratory information systems (LIS) that employ bar code technology. Collection list and label software can be modified to integrate and produce bar coded information. Confirmation of labeling and patient ID occurs at the bedside. Specimen sorting and aliquoting in the laboratory can be shortened or eliminated by various setups. At Rush-Presbyterian Hospital in Chicago, a central receiving station rapidly sorts and aliquots bar coded samples as they move on a conveyor belt.3 At the University of Kansas Medical Center, the collection tubes are also used for analysis, thus eliminating the need for aliquoting samples. Additionally, the computer sends the bar code-coordinated orders to each of the 2 chemistry analyzers. Because a sample can then be appropriately processed at either analyzer, the need for sorting samples has also been eliminated.8 Clinical labs also employ bar code technology in harder-to-automate processes. For instance, the University of Utah uses bar codes to replace common keystrokes for text reporting of microbiology results—e.g., a technician might use a bar code "pick list" to scan the single bar code that means "no growth for 24 hours" and eliminate the need to type this phrase.17,18
Medication dispensing and administration
The use of bar codes is uniquely suited to the dispensing and administration stages of the medication process.19 Bar coding may be used to simplify the patient cassette (the medicine tray for bedside delivery) filling and verification process.20 For instance, a technician fills a cassette according to a computerized, bar coded medication schedule and completes a quick verification by scanning the label of each unit-dose that has been placed in it with a handheld scanner. The computer can generate an error message if an incorrect medication is entered. The administration of bar coded medications can also be tracked at the point-of-care using a portable scanner and compared against the hospital computer's medication orders.9,21 Bar coded medication administration has the added capability of creating a record of the administration (i.e., RN, date, time) and a bill. This type of system could be integrated with a patient identification system in an attempt to eliminate errors resulting in administration of medication to the wrong patient.
Medical record keeping
Radiology and medical records departments use bar code technology to track the location and status of studies and charts.22-24 Even more creative use of bar coded information has been reported in the emergency medicine and pharmacy literatures. As with the applications in the microbiology lab, bar codes can be used to replace frequently used text for the creation of medical records. "Pick lists" of bar codes with their text equivalents can be employed in circumstances requiring speed and accuracy. Several uses of bar coded scripts have been examined in resuscitation events, and in mass casualty recording.10,25-27 The use of bar code "pick lists" for the documentation of pharmacists' clinical activities has also been explored.28-30
Prevalence and Severity of the Target Safety Problem
Bar code technology may be used to address any number of patient safety issues in medicine. For this discussion, we will define the target safety problem as patient identification in general, using transfusion medicine as a specific example.
Patient identification remains a challenge in hospitals because of the number of complex interventions that occur to patients ranging from meals to surgeries. These interventions occur in a variety of locations and are provided by large teams of staff who work in shifts. In addition, sick patients, or those who have a language barrier, are not always capable of responding to questions about their identity or treatment plans. Hospitals generally rely on standardized wristbands containing the patient's name and other identifying information such as medical record number or date of birth. Unfortunately, conventional wristbands are not reliable sources of patient identification. A 1991 national sample of 712 hospitals estimated error rates for conventional patient identification wristbands to be 5.5%.31 In half of the errors, the patient's wristband was absent altogether. The error rates were significantly lower in hospitals where phlebotomists had responsibility for monitoring wristband accuracy, as the phlebotomy staff would not perform routine lab work unless the band was corrected. Other errors included more than one wristband with conflicting data (18.3%); wristbands with incomplete (17.5%), erroneous (8.6%), or illegible data (5.7%); and rarely, patients wearing wristbands with another patient's data (0.5%). As patient identification data are only as good as the information entered at registration, the use of bar coded ID data could not be expected to correct certain types of errors such as a wristband with incorrect data entered at admission, although it is potentially beneficial in eliminating other types of errors such as illegible data.
Even when wristbands are free of errors, protocols for patient identification (such as dual witness verification of identification for blood transfusion) are easily circumvented or performed incorrectly.32 In an analysis of major transfusion errors reported to the FDA over a 10-year period from 1976-1985, Sazama found 10 patient deaths where the actual and intended patients shared the same last name, and 5 deaths where the 2 shared the same hospital room.13 Consequently, automatic patient identification systems have been proposed as a technological solution to remove human factors (Subchapter 41.1) from the patient identification process.
Despite technical improvements in testing for blood group identification, fatal ABO-incompatible transfusions in the United States continue to occur at a rate ranging from approximately 1:600,000 to 1:800,000, with as many as two dozen fatalities in the US annually.12,13 Thus, the chance of a patient suffering a fatal transfusion reaction due to ABO-incompatibility is roughly equivalent to the risk of acquiring HIV infection from a blood transfusion.33,34 Patient misidentification represents the most common cause of ABO-incompatible transfusion, accounting for 46-57% of these errors.12,13 Since the rate of patient and donor having blood group compatibility by chance is approximately 60%, it is estimated that the total number of ABO-incompatible transfusions is much higher than the rate of fatal errors. A study from New York State estimated that as many as one in 12,000 transfusions involve administration of a blood product intended for another patient or release of blood of an incorrect group.12
Opportunities for Impact
The implementation of automatic patient identification may present a large opportunity to bring transfusion medicine and other hospital interventions closer to the goal of zero risk. According to a survey conducted by the American Society of Hospital-System Pharmacists, 1.1% of responding hospitals use bar coding of drug products in conjunction with bar coding on the patient's identification tag.35
Multiple reports of the use of bar codes appear in the medical literature, but most of these relate to inventory management. Few authors examine bar codes for patient identification. Only one of these studies was a prospective evaluation, and in the study bar coded patient identification comprised only one small part of the intervention.10 One observational study examined a bar code patient ID system for medication administration.9 In the study, however, routine patient ID scanning was easily circumvented and the actual error rate was not provided. The remainder of the reports are descriptive in nature.14,36,37 Therefore, error rates in automated patient identification could not be readily compared to usual practice.
The use of bar coding in other clinical care applications (aside from inventory control) has been examined prospectively in trauma recording27 and in documenting pharmacists' interventions.28 One additional observational study examined bar coding in pharmacy dispensing.20
Some of the studies report error rates in transfusion or medication errors (Level 2), while others report related outcomes such as speed of data entry (Level 3), transcription errors (Level 2) in a variety of experimental and clinical settings, and user satisfaction (Level 3).
Evidence for Effectiveness of the Practice
A point-of-care information system for medication management was implemented at a tertiary care center in Colorado.9 The system provided online patient and medication data that verified medication administration at bedside using hand-held scanners to record patient ID, nurse ID, and medication unit-dose bar codes. When intervention data were compared with historical controls, the pre-intervention medication error rate of 0.17% dropped to 0.05%, sustained over 3 years for an overall decrease of 71% (p value not reported). There was a 33% decrease in "wrong drug" errors, a 43% decrease in "wrong time" errors, and a 52% decrease in "omitted dose" errors. There was a 47% decrease in "transcription/order-entry" errors. There was no change in "wrong patient" errors or "wrong dosage" errors, perhaps because the component of the multifaceted intervention most likely to mitigate these errors, the use of the scanners for patient ID, was easily and frequently circumvented. It is unclear if bedside scanning added an unwanted layer of work for the nurses, or if they were uncomfortable performing this task in front of patients and families. Computerized pharmacy and bar code tracking of medications led to qualitative improvements in documentation time, scheduling of administration, nursing-pharmacy communications, and pharmacist drug monitoring. However, the contribution of bar coding to the decreased error rate is not distinguishable from that of the entire intervention. It also appears that the point-of-care patient ID portion of the intervention was easily bypassed, and was therefore not adequately evaluated.9
On a large medical ward of a university hospital, the satellite pharmacy was reconfigured to implement bar code technology for drug dispensing.20 The hospital bar coded all medications, patient medication cassettes, patient wristbands, and employees. Standard dispensing time was estimated at 8.24 seconds, while dispensing time for bar coded medications was 6.72 seconds (p value not reported). Accuracy of the standard cassette fill system was 99.6% (equivalent to one error in 250 doses), while the accuracy of bar coded cassette fill was reported to be 100% (based on 0 errors in about 20,000 doses). The pharmacists were freed to do other work as the burden of dispensing was shifted to pharmacy technicians.
As pharmacists have begun to use bar code technology for medication distribution, 2 pharmacy groups described the use of bar codes in recording their clinical interventions.28, 29 One group found bar code documentation to have lower overall error rates compared to manual documentation, while the marginal cost of implementing bar coding was less than $100 (factoring in savings in labor costs related to manual entry).28 These data were limited by the small number of operators who differed on their preferences for manual versus bar code recording.
A prospective trial reviewed bar code technology in trauma recording. Experienced emergency room nurses found bar coded pick lists to be easy to use and produced fewer errors per record compared with handwriting (2.63 ± 0.24 vs. 4.48 ± 0.3, p<0.0001) for videotaped trauma resuscitations.27 In a prospective study of simulated mass casualty incidents in The Netherlands, bar coded computer registration produced 25% fewer inaccuracies than handwritten medical charts.10
The limitations of these studies are numerous, including their small sample sizes and lack of generalizability. However they demonstrate that bar code technology is generally easy for operators to use, can be applied in a variety of creative ways, and produces easy-to-demonstrate gains in accuracy and efficiency.
Potential for Harm
There is no clear detriment to patient identification with a bar code system. However, as with the addition of any new technology, the possibility exists that the complexity of the information system, especially if it grows, could create more routes for potential failure. For instance, an error during patient registration might be perpetuated throughout the hospitalization by the information systems, and be more difficult to correct than with conventional systems. The system's data are only as accurate as that entered by fallible humans.
Costs and Implementation
Significant barriers need to be overcome before the full potential of bar coding can be exploited in the clinical setting. The process of medication dispensing and administration highlight several examples of these barriers. First, pharmaceutical manufacturers have yet to adopt a universal bar code standard like the UPC system used in grocery store inventory.38 As of yet, there has been no regulatory mandate by the FDA to serve as an incentive, although the American Society of Hospital-System Pharmacists recently urged the FDA to take action.39 Second, placing bar codes on unit-doses of medications often requires major changes in packaging (such as an increase of the package size to accommodate the bar coded label). Bar code tracking systems would have difficulty with unusual doses such as halved tablets. IV doses would still require bar code labeling in the pharmacy when prepared.
At this point, implementation of bar coding requires a commitment on the part of the hospital to relabel every unit-dose of medication using a hospital standard. The hospital pharmacies that have implemented these systems have repackaged and relabeled many unit-doses at considerable cost.9,20 One health system estimated costs at $119,516 annually, with the per dose costs of bar code labeling estimated at 2.73 cents.20 The bottom line is that, at present, the costs of implementing bar coding in an entire pharmacy inventory are significant and the logistics complex.
The use of bar coding in simpler clinical scenarios (i.e., using a blood transfusion wristband for patient identification) may be implemented with very modest outlay of resources (estimated to be less than 5 cents per wristband).36 The costs of the scanners themselves are moderate. A new model scanner, software and recharger were priced at about $1100 in 1997.28
Bar coding is a fast and accurate method of automated data capture that in experimental settings provides qualitative improvements in speed and accuracy of data entry. Bar code technology might be creatively applied to any number of data-driven processes in medicine. As the rapid and accurate transfer of data is paramount in healthcare, the thoughtful application of appropriately piloted and evaluated bar code technology is likely to be well received and deserves further investigation.
1. Rappoport A. A hospital patient and laboratory machine-readable identification system (MRIS) revisited. J Med Syst 1984;8:133-156.
2. Weilert M, Tilzer LL. Putting bar codes to work for improved patient care. Clin Lab Med 1991;11:227-238.
3. Maffetone MA, Watt SW, Whisler KE. Automated sample handling in a clinical laboratory. Comput Healthc 1988;9:48-50.
4. Brodheim E, Ying W, Hirsch R. An evaluation of the codabar symbol in blood bank automation. Vox Sang 1981;40:3.
5. Longe K. The status of bar codes in hospitals: a survey report. Hospital Technology Series Chicago: American Hospital Association; 1989:8.
6. Garza D, Murdock S, Garcia L, Trujillo JM. Bar codes in the clinical laboratory. Clin Lab Sci 1991;4:23-25.
7. Zarowitz BJ, Petitta A, Rudis MI, Horst HM, Hyzy R. Bar code documentation of pharmacotherapy services in intensive care units. Pharmacotherapy 1996;16:261-266.
8. Tilzer LL, Jones RW. Use of bar code labels on collection tubes for specimen management in the clinical laboratory. Arch Pathol Lab Med 1988;112:1200-1202.
9. Puckett F. Medication-management component of a point-of-care information system. Am J Health-Syst Pharm 1995;52:1305-1309.
10. Noordergraff G, Bouman J, Van Den Brink E, Van De Pompe C, Savelkoul T. Development of computer-assisted patient control for use in the hospital setting during mass casualty incidents. Am J Emerg Med 1996;14:257-261.
11. Sherer P, Chambers R, Taswell H, Altshuler C, Aster R, Covino K, et al. Automated donor-recipient identification systems as a means of reducing human error in blood transfusion. Transfusion 1977;17:586-597.
12. Linden J, Paul B, Dressler K. A report of 104 transfusion errors in New York State. Transfusion 1992;32:601-606.
13. Sazama K. Reports of 355 transfusion-associated deaths: 1976 through 1985. Transfusion 1990;30:583-590.
14. Jensen N, Crosson J. An automated system for bedside verification of the match between patient identification and blood unit identification. Transfusion 1996;36:216-221.
15. Wenz B, Burns E. Improvement in transfusion safety using a new blood unit and patient identification system as part of safe transfusion practice. Transfusion 1991;31:401-403.
16. Cox C, Enno A, Deveridge S, Seldon M, Richards R, Martens V, et al. Remote electronic blood release system. Transfusion. 1997;37:960-964.
17. Aller R, Freidman W. Rapid accurate entry of microbiology results. Arch Pathol Lab Med 1996;120:57-61.
18. Willard K, Stanholtzer C. User interface reengineering: Innovative applications of bar coding in a clinical microbiology laboratory. Arch Pathol Lab Med 1995;119:706-712.
19. Bates D. Using information technology to reduce rates of medication errors in hospitals. BMJ 2000;320:788-791.
20. Meyer G, Brandell R, Smith J, Milewski F, Brucker J, P, Coniglio M. Use of bar codes in inpatient drug distribution. Am J Hosp Pharm 1991;48:953-966.
21. Gebhart F. VA slashed drug errors via bar-coding. Drug Topics 1999;143:44.
22. Mudie D. Bar coding in a medical records department: a case study. Hosp Technol Ser 1988;7:1-15.
23. Arenson RL, London JW. Comprehensive analysis of a Radiology Operations Management computer system. Radiology 1979;133:355-362.
24. Connelly A. Bar codes help control incomplete medical records. J Am Med Rec Assoc 1987;58:48-49.
25. Burt TW, Bock HC. Computerized mega code recording. Ann Emerg Med 1988;17:339-342.
26. Bock HC. Field verification methodology using bar coding to record data. Ann Emerg Med 1993;22:75-8; discussion 78-79.
27. Chua R, Cordell W, Ernsting K, Bock H, Nyhuis A. Accuracy of bar codes versus handwriting for recording trauma resuscitation events. Ann Emerg Med 1993;22:1545-1550.
28. Kanmaz T, Haupt B, Peterson A. Comparison of manual and bar-code systems for documenting pharmacists' interventions. Am J Health-Syst Pharm 1997;54:1623-1626.
29. Zarowitz B, Petitta A, Mlynarek M, Touchette M, Peters M, Long P, et al. Bar-code technology applied to drug use evaluation. Am J Hosp Pharm 1993;50:953-959.
30. Scott M, McElnay J, Burnett K. Using bar-code technology to capture clinical intervention data in a hospital with a stand-alone pharmacy computer system. Am J Health-Syst Pharm 1996;53:651-654.
31. Renner S, Howanitz P, Bachner P. Wristband identification error reporting in 712 hospitals. Arch Pathol Lab Med 1993;117:573-577.
32. Feldman S, Roblin D. Medical accidents in hospital care: applications of failure analysis to hospital quality appraisal. Jt Comm J Qual Improv 1997;23:567-580.
33. Schreiber G, Busch M, Kleinman S, Korelitz J. The risk of transfusion-transmitted viral infections. The Retrovirus Epidemiology Donor Study. N Engl J Med 1996;334:1685-1690.
34. Lackritz J, Satten G, Alberle-Grasse J, al e. Estimated risk of transmission of the human immunodeficiency virus by screened blood in the United States. N Engl J Med 1995;333:1721-1725.
35. American Society of Hospital-System Pharmacists. National survey of pharmacy practice in acute care settings: dispensing and administration—1999. Am J Health-Syst Pharm 2000;57:1759-1775.
36. Lau F, Wong R, Chui C, Ng E, Cheng G. Improvement in transfusion safety using a specially designed transfusion wristband. Transfusion Medicine 2000;10:121-124.
37. Amber R, Everett V. Emergency department patient tracking: A cost-effective system using bar code technology. J Emerg Nurs 1996;22:190-195.
38. Davis NM. Initiative for reducing medication errors: the time is now. Am J Hosp Pharm 2000;57:1488.
39. American Society of Hospital-System Pharmacists. ASHP policy recommendation. Am J Health Syst Pharm 2001;58:603-606.