Patient Identification Errors Are Common in a Simulated Setting

Article Outline

• Abstract
• Introduction
• Materials and Methods
  • Study Design
  • Setting
  • Selection of Participants
  • Interventions
  • Methods of Measurement
  • Primary Data Analysis
• Results
• Limitations
• Discussion
• References
• Copyright

Study objective

We evaluate the frequency and accuracy of health care workers verifying patient identity before performing common tasks.

Methods

The study included prospective, simulated patient scenarios with an eye-tracking device that showed where the health care workers looked. Simulations involved nurses administering an intravenous medication, technicians labeling a blood specimen, and clerks applying an identity band. Participants were asked to perform their assigned task on 3 simulated patients, and the third patient had a different date of birth and medical record number than the identity information on the artifact label specific to the health care workers' task. Health care workers were unaware that the focus of the study was patient identity.

Results

Sixty-one emergency health care workers participated—28 nurses, 16 technicians, and 17 emergency service associates—in 183 patient scenarios. Sixty-one percent of health care workers (37/61) caught the identity error (61% nurses, 94% technicians, 29% emergency service associates). Thirty-nine percent of health care workers (24/61) performed their assigned task on the wrong patient (39% nurses, 6% technicians, 71% emergency service associates). Eye-tracking data were available for 73% of the patient scenarios (133/183). Seventy-four percent of health care workers (74/100) failed to match the patient to the identity band (87% nurses, 49% technicians). Twenty-seven percent of health care workers (36/133) failed to match the artifact to the patient or the identity band before performing their task (33% nurses, 9% technicians, 33% emergency service associates). Fifteen percent (5/33) of health care workers who completed the steps to verify patient identity on the patient with the identification error still failed to recognize the error.

Conclusion

Wide variation exists among health care workers verifying patient identity before performing everyday tasks. Education, process changes, and technology are needed to improve the frequency and accuracy of patient identification.

SEE EDITORIAL, P. 511.

Back to Article Outline

Introduction 

A patient's identity should be verified before medical interventions are performed to ensure that the intervention is done on the correct patient. Whether taking a history, selecting a medical record, performing a procedure, administering a medication, or taking the patient to another place in the hospital, the verification process should be performed.

It is unknown how many errors are caused because a caregiver did not accurately identify a patient. They are probably infrequent. However, patient identification errors are particularly worrisome because they are preventable and they affect 2 patients instead of 1. The first patient receives an unintended intervention, whereas the second patient does not receive the intended intervention, and this action may harm one or both patients.

The Joint Commission, which accredits most hospitals, established the National Patient Safety Goals in 2005 to help reduce medical errors. One of the goals has been to “improve the accuracy of identifying patients” to reduce or eliminate patient identification errors. The Joint Commission states that the appropriate patient identifiers are name, date of birth, and medical record number.¹ A description of the process of verifying patient identity is shown in Figure 1.

• 
  • View Large Image
  • Download to PowerPoint

• Figure 1. 

  The process of verifying patient identity. MRN, Medical record number.

In our previous study of emergency providers ordering tests on a simulated patient in a computer order entry system, we found that most (92%) did not verify patient identity and ordered tests on an incorrect patient when presented with an unexpected patient identity error.²

The purpose of this study is to investigate the frequency and accuracy of other health care workers performing the process of verifying patient identity during common, everyday tasks on simulated patients.

Back to Article Outline

Materials and Methods 

Study Design 

This was a prospective, observational study of emergency health care workers simulating the performance of common patient-related tasks while wearing an eye-tracking device. The study had institutional review board approval, and all participants gave their signed informed consent before their participation.

Setting 

The study was conducted in a simulated emergency department (ED) patient care space in a 600-bed, urban, Level I trauma, pediatric, and tertiary referral center in western Massachusetts, with an annual ED census greater than 100,000.

Selection of Participants 

Emergency nurses, emergency service associates (clerks), and technical associates volunteered to participate in the study during one of their day or evening shifts. Participants were told that the purpose of the study was to evaluate how expert health care workers use visual cues to perform common tasks. Study participants were told that they would wear an eye-tracking device that would videotape the field of vision in front of them and place crosshairs on the recorded video, showing exactly what the health care worker was looking at each moment. Health care workers were not aware that the purpose of the study was to investigate the process of verifying patient identity.

Interventions 

Figure 2 shows the study description. Each health care worker performed a specific task on 3 simulated patients (paid volunteers): nurses administered an intravenous medication, technicians labeled a blood specimen, and emergency service associates (clerks) placed an identity band on their wrist. The health care worker was asked to perform the task the same way he or she does every day in the ED, except for the simulated aspects (ie, giving an actual medication or drawing blood).

• 
  • View Large Image
  • Download to PowerPoint

• Figure 2. 

  Study description. RN, Nurse; ESA, emergency service associate; HCW, health care worker; TA, technical associate.

For the nurses and technical associates, the third patient had the same name but different date of birth and medical record number on the identity band on the wrist than on the task artifacts (eg, labeled medication, specimen label) the nurses and technical associates brought into the room. For the emergency services associates, the patient had the same name but a different date of birth than was on the identity band (ie, artifact) the emergency service associate brought into the room.

After placing an eye-tracking device on the health care worker and calibrating it (ASL Mobile Eye; Applied Science Laboratories, Bedford, MA; Figure 3), the health care worker was taken to a series of numbered rooms and provided with a list of patients and room numbers. For each simulated patient, the health care worker was given a clipboard with patient-specific materials. The nurses were given a labeled order/documentation page and a labeled intravenous medication. The emergency service associates were given an identity band and an unlabeled documentation page, and the technical associates were given a blood specimen label and an unlabeled documentation page.

• 
  • View Large Image
  • Download to PowerPoint

• Figure 3. 

  Eye-tracking device.

(Photo courtesy of Tracy Zafian.)

The labels used in the simulations contained patient identity information similar to that of regular hospital labels but with the information more spread out. The patient name, date of birth, and medical record number were placed on a different vertical or horizontal axis on the label to make it easier to differentiate which specific identity information the health care worker was looking at. The same labels were used for all materials, including the identity bands. Figure 4 shows an identity label used in the simulation.

• 
  • View Large Image
  • Download to PowerPoint

• Figure 4. 

  Label for patient with identification error.

(Photo courtesy of Tracy Zafian.)

Methods of Measurement 

An observer (T.A.P.) followed the health care workers to each simulated patient and completed a standardized data sheet. The observer also recorded any additional observations. A second observer (E.A.H.) scored an identical data sheet on 6 participants (18 patient scenarios) to determine interrater reliability.

Identifying the patient identity error was defined as not completing the assigned task on that patient with or without voicing that the patient had a different identification than the labels or identity band brought into the room.

The ASL Mobile Eye (Applied Science Laboratories) is a tetherless eye-tracking system that allows freedom of movement and can be worn by an active participant. The eye tracker includes a scene camera, optics, and reflecting mirror, all mounted on safety glasses (Figure 3). Pupil-corneal reflections are used to measure the position of the eye.

The eye-tracking device is first calibrated to each user. The calibration process has subjects look at 12 specific reference points in their field of view. A mark will appear on the video near each reference point when it is fixed. The marks are then adjusted to each of the specific reference points. The output is stored on a tape. The tape is analyzed with the Mobile Eye software program, which after calibration is able to overlay crosshairs (plus signs) at the approximate location in a scene where the individual is gazing at each point.

After the experiment, all videos were reviewed by 2 independent observers (T.A.P., M.M.C.), who recorded whether the health care workers looked or did not look at each of the patient identifiers on the labels and identity bands. Videos were reviewed in 0.4-second intervals, specific to the software program, to determine what patient identifiers the participant looked at. Disagreements between the 2 observers were resolved by a third observer (P.L.H.). A health care worker was assumed to have looked at a specific patient identifier if 2 of the independent reviewers determined that the specific identity information was within an imaginary 1-cm² box outlined by the crosshairs on the calibrated video during a 0.4-second interval.

Eye-tracking failures were documented. Eye-tracking failures occurred when there was no video image or the crosshairs did not appear on the video image or when aspects of patient identification could not be differentiated on the labels because the image was washed out from excessive glare.

Verifying patient identity was defined in 2 ways: matching the patient to the identity band and matching the task artifact(s) to the patient or the patient's identity band (Figure 2). Verifying patient identity required the use of 2 available patient identifiers (ie, name, date of birth, or medical record number). Matching the patient to the identity band could only be done by asking the patient his or her name and date of birth and confirming an exact match to the name and date of birth on the identity band attached to the patient's wrist. The process of matching the task artifact(s) to the patient or the identity band required the health care workers to match 2 identifiers from the artifact(s) to the patient or the patient's identity band. The task artifact was defined as the object(s) to be used in the planned task: for the nurse, the task artifacts were the identity label on the medication bag and the identity label on the order/documentation page; the task artifact for the technical associates was the specimen label to be applied to a blood specimen before sending it to the laboratory; the task artifact for the emergency service associates was the identity band to be applied to the patient.

Primary Data Analysis 

Windows Media (Microsoft, Redmond, WA) was used to review the calibrated videos in the program-specific 0.4-second intervals. For statistical analyses, we used Stata/SE 10.1 (StataCorp, College Station, TX). We calculated 95% binomial exact confidence intervals (CIs) for all percentages. χ² Tests were used for all categorical inferences, except when the cell counts were less than 3, in which case Fisher's exact test was used. P<.05 (ie, α=0.05) was considered significant. Percentage agreement and κ statistics were calculated for interrater agreement.

Back to Article Outline

Results 

Sixty-one health care workers participated in the study: 28 nurses, 16 technical associates, and 17 emergency service associates. The above participants represent 22% of all full- and part-time nurses, 48% of all technical associates, and 33% of all full- and part-time emergency service associates of the ED.

Table 1 includes observational data. The interrater agreement between the independent reviewers was 100% for the observational data (κ=1.00). In half of the scenarios, the health care worker introduced himself or herself to the patient at the start of the scenario (52%, or 95/183; 95% CI 44% to 59%). There was substantial variation among health care worker types in the frequency of self-introductions: 56% among nurses, 79% among technicians, and 20% among emergency service associates.

Table 1. Main study results.^⁎

Variable	RN	TA	ESA
Total participants	28	16	17
Simulated patients	3	3	3
Patient scenarios (TPxSP)†	84	48	51
Ask name, %	90(76/84)	81(39/48)	96(49/51)
Ask DOB, %	39(33/84)	75(36/48)	63(32/51)
Catch error, %	61(17/28)	94(15/16)	29(5/17)
Perform task wrong patient, %	39(11/28)	6(1/16)	71(12/17)
% Total scenarios with eye tracking, %	80(67/84)	69(33/48)	65(33/51)
Percentage agreement eye tracking	90	94	88

In most scenarios (91%, or 167/183; 95% CI 86% to 95%), the health care workers asked the patient his or her name. Of those asking about the name, 59% (99/167) asked “What is your name?” and 44% (74/167) asked whether the patient's name was “X”; 5% (4/74) asked only about the last name and 5% (4/74) asked only about the patient's first name. In 102 of 183 of the patient scenarios (56%; 95% CI 48% to 63%), the health care workers asked about the patient's date of birth. Of those asking about the date of birth, 92% (94/102) asked “What is your date of birth?” and 12% (12/102) asked whether the patient's date of birth was “X.”

Twenty-nine percent (5/17) of the emergency service associates detected the patient identification error in patient 3 and did not place an identity band on the patient. Seventy-one percent did not detect the patient identification error and placed an identity band on the wrong patient. Sixty-one percent (17/28) of nurses detected the identity error, whereas 39% “administered” an intravenous medication to the wrong patient. Ninety-four percent (15/16) of the technical associates detected the identity error; 6% “drew and sent” a laboratory specimen with the wrong patient's label. The difference in identity error detection among the 3 groups was statistically significant (P<.001), with the technical associates performing substantially better than the nurses or emergency service associates. All health care workers (61% [37/61]; 95% CI 47% to 73%) who did not complete the task on the third patient voiced that there was a patient identification error.

Table 2 lists the eye-tracking results. There were data on 133 scenarios because 50 of the 183 patient scenarios (27%) did not have eye-tracking data. The reasons for the eye-tracking failures included no crosshairs on the video image (32/50), glare or poor focus obscuring the identity information (15/50), and inability of participants to wear the eye-tracking device over their own glasses (3/50). There was a 90% agreement between the 2 independent observers reviewing the eye-tracking data (κ=0.77).

Table 2. Eye-tracking results^⁎: verifying patient identity.

	RN, Pt to ID†	RN, Artifact to Pt/ID‡	TA, Pt to ID	TA, Artifact to Pt/ID	ESA, Artifact to Pt/ID
Patient 1	3/23	15/23	4/11	10/11	7/11
Patient 2	3/23	15/23	7/11	10/11	8/11
Patient 3	3/21	15/21	6/11	10/11	7/11
Total	9/67,13%	45/67,67%	17/33,51%	30/33,91%	22/33,67%

Pt, Patient; ID, identity.

Overall, in 26% of the scenarios (26/100; 95% CI 18% to 36%), the health care workers matched the patients to their identity band, and 73% (97/133; 95% CI 65% to 80%) matched the task artifact(s) to the patient or the patient's identity band before performing their assigned task.

Of the 43 health care workers with eye-tracking data for the patient with the identification error (ie, patient 3), 33 “matched” the artifact to the patient or the identity band, but only 28 of the 33 “caught” the error. Five health care workers (2 nurses, 3 emergency service associates; 15%; 95% CI 5% to 32%) asked for and looked at the appropriate identity information to match the artifact to the patient or the identity band but did not recognize the patient identification error.

Table 3 shows the frequency and type of self-reported patient identification errors during a 2-year period at our academic urban medical center.

Table 3. Reported patient identification errors (events) in a medical center's voluntary safety reporting system.^⁎

	2007	2008
Total number patient ID events	465	435
Total number SRS events	10,127	9,099
Total inpatient days	191,496	191,767
Total discharged inpatients	39,107	38,510
Total ED visits	110,022	111,433
Total clinic visits	324,377	318,535
Type of patient ID events
Medication events	48	51
Surgical events	2	1
Laboratory events	330	322
Radiology events	7	1
Missing or incorrect ID	78	60
Total number of patient ID events (% total SRS events)	465(4.6)	435(4.8)

SRS, Safety reporting system.

Back to Article Outline

Limitations 

The main limitation of our study is that it was performed in a simulated task environment without acclimation and without any of the stressors commonly found in the ED (eg, time pressures, noise, interruptions and distractions). However, simulation was the ideal way to study patient identification errors because medical ethics precluded us from embedding identification errors in a clinical setting. The study cannot determine whether using simulated patients made it more or less likely for staff to perform the process of verifying patient identity. The proportion verifying patient identity was similar for each of the 3 simulated patients (Table 2), implying that participant behavior was similar for each patient. In other contexts, it has been reported that hazard-anticipation behaviors in the field are similar to those observed in a driving simulator, and the controlled aspects of simulation make it an ideal way to study the process of verifying patient identity.³ It is also unknown whether participants discussed the study with subsequent participants during the 4-day study period (they were asked not to), although the percentage noticing the patient identity error did not vary significantly from day 1 to day 4.

Our study was performed on staff from a single institution. It is unknown how much of our results can be generalized to other institutions, though obviously patient identification errors occur across the nation. In simulated settings at our hospital, we have observed a wide range in the frequency and accuracy in the verification process, from 8% by medical providers to 96% for technicians.² We believe that it is likely that a range in the frequency and accuracy in the verification process will be found in other institutions among other health care workers.

Finally, the eye-tracking device failed in 27% of the participant scenarios. This high failure rate is probably due to the increased mobility of our participants versus the more stationary scenarios previously studied with this device. None of the failures were related to patient identity or study protocol. Reducing the failure rate of the eye tracker should be possible by better controlling the light in the simulated setting and by expanding the reference points in the calibration process to include a wider area of the scene.

Back to Article Outline

Discussion 

Our study showed that patient identification errors can occur easily in a simulated setting. The majority of the nurses and technical associates (74% [74/100]; 95% CI 64% to 82%) did not match the patients to their identity band, and one quarter of all health care workers (27% [36/133]; 95% CI 20% to 35%) did not match the task artifact(s) to the patients or their identity band before performing their task. Almost two thirds of health care workers were able to catch a patient identification error, but more than a third performed their assigned task on the wrong patient. Fifteen percent of those who performed their task on the wrong patient actually completed the steps to verify patient identity but still missed the identification error.

Technical associates detected the patient with the identification error (94%) more often than nurses (61%) and emergency service associates (29%; P<.001). Mislabeled laboratory specimens have been a commonly reported medical error in our ED and hospital (Table 3).⁴, ⁵ In response to the high incidence of mislabeled laboratory specimens, the technical associates have undergone additional training.⁶ Mislabeled specimens that involve the blood bank are also included in the performance evaluations of the technical associates. Our results suggest that the verification of patient identity may improve with supplemental education and tying reported errors to performance evaluations.

It is disturbing that so many health care workers did not match the patients to their identity band and yet still used the identity band to verify patient identity. Also, when emergency services associates are presented with an unexpected patient identity error in clinical simulations, the majority will place an identity band on the wrong patient. In a study of 2.5 million patient wrist bands at 712 hospitals, phlebotomist found that 9% of bands had erroneous data, 8% of patients had multiple bands with different information, and 0.5% of patients were wearing wrist bands with another patient's identity information.⁷ Ideally, identity bands should always accurately identify a patient regardless of the patient's ability to confirm its accuracy.

Before performing a health task, the health care workers should match the artifact(s) to be used in the task to the patient. In each of the studied tasks, the process of matching the artifact to the patient was potentially different. Emergency service associates could only ask the patients for their name and date of birth and confirm that they were an exact match to the artifact (ie, the identity band to be placed on the patient). The technical associates had a single artifact (the specimen label) to match to the stated name and date of birth of the patient or the name, date of birth, or medical record number on the identity band the patient was wearing. Nurses had 2 artifacts they could use (ie, the chart and medication labels) to match to the patient or the patient's identity band. Perhaps the simplest, most efficient way to verify an awake patient's identity before performing a task is to ask the patient his or her name and date of birth and compare the stated name and date of birth to the name and date of birth on the identity band the patient is wearing and to the task artifact; this procedure confirms the accuracy of the identity band and matches the task artifact to the patient.

Patient identification errors are infrequently reported (less than 5% in our institution). They are most often reported for mislabeled laboratory specimens and are usually reported by the laboratory personnel. A less commonly reported error is giving a medication to the wrong patient, whereas placing an identity band on the wrong patient is rarely reported.⁴, ⁵

The patient identification error we included in the study has been documented to occur but is probably a rare phenomenon in a typical ED in the United States (ie, same first and last name and different date of birth and medical record number).⁷, ⁸, ⁹ We know that 11% of the time in our ED, there are at least 2 patients with the same last name. However, the presence of 2 patients in our ED with the same first and last name and different date of birth occurs less than 0.035% of the time.² The relatively rare type of patient identification error we studied will likely contribute to the health care workers being less vigilant in detecting it.

Additional training is just one approach to improving the frequency and accuracy of health care workers verifying patient identity. Bar code readers that compare the bar codes on the patient's identity band and a medication can be used to verify patient identity. Similarly, bar code readers can print specimen labels that match the patient's identity band.¹⁰ Unfortunately, workarounds have been documented by nursing staff using these devices.¹¹ Some combination of technology and training may be required to overcome the intrinsic limitations of the human mind during these types of verification procedures. In light of the critical requirement that the patient's identity band be accurate, it may be prudent to develop specific procedures for verifying all identity bands, immediately after they are applied, by an individual other than the one who applied the band.

The finding of inattention (ie, looking at disparate identity information from 2 sources but not noting the discrepancy) in 3 emergency service associates and 2 nurses supports the use of technology to aid the health care workers when they are verifying patient identity. Humans will make mistakes. Systems must plan for human error and have processes that adapt to these errors.

In summary, we found that many health care workers in clinical simulations do not verify patient identity before performing a task, which resulted in more than a third of health care workers performing a task on the wrong patient in a simulated setting when there was an unexpected patient identity error. Although patient identification errors are infrequent, they may result in serious adverse events and are preventable. Improved training and a better use of technology may improve the way health care workers verify patient identity, and additional research on these methods is warranted.

Back to Article Outline

References 

1. Joint Commission. National patient safety goals. http://www.jointcommission.org/PatientSafety/NationalPatientSafetyGoalsAccessed December 12, 2008
   • View In Article
2. Henneman PL, Donald L, Fisher DL, et al. Providers do not verify patient identification during computer order entry. Acad Emerg Med. 2008;15:641–648
   • View In Article
   • CrossRef
3. Fisher DL, Pradhan AK, Pollatsek A, et al. Empirical evaluation of hazard anticipation behaviors in the field and on a driving simulator using an eye tracker. Transp Res Rec. 2008;2018:80–86
   • View In Article
4. Fordyce J, Blank D, Pekow P, et al. Errors in a busy emergency department. Ann Emerg Med. 2003;42:324–336
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (111 KB)
   • CrossRef
5. Henneman PL, Blank FS, Smithline H, et al. Voluntarily reported emergency department errors. J Patient Saf. 2005;1:126–132
   • View In Article
   • CrossRef
6. Blank SJ, Henneman PL, Maynard AM, et al. Development and implementation of a patient safety program in an academic, urban emergency department. J Emerg Nurs. 2006;32:491–496
   • View In Article
   • Full Text
   • Full-Text PDF (85 KB)
   • CrossRef
7. Renner SW, Howanitz PJ, Bachner P. Wrist band identification error reporting in 712 hospitals. Arch Pathol Lab Med. 1993;117:573–577
   • View In Article
   • MEDLINE
8. Hakimzada AF, Green RA, Sayan OR, et al. The nature and occurrence of registration errors in the emergency department. Int J Med Inform. 2008;77:169–175
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (262 KB)
   • CrossRef
9. Bittle MJ, Charache P, Wassilchalk DM. Registration-associated patient misidentification in an academic medical center: causes and corrections. J Qual Patient Saf. 2007;33:25–33
   • View In Article
10. Mulder DR. Care fusion: positive patient identification from Cardinal Health (Pharmacy Purchases & Products May 2007). P 16 http://www.pppmag.com/documents/V5N5/p16.pdfAccessed December 3, 2009
    • View In Article
11. Koppel R, Wetterneck T, Telles JL, et al. Workarounds to barcode medication administration systems: their occurrence, causes, and threats to patient safety. J Am Med Inform Assoc. 2008;15:408–423
    • View In Article
    • CrossRef