Patient Misidentification in Laboratory Medicine. A Qualitative Analysis of 227 Root Cause Analysis Reports in the Veterans Health Administration

Transcription

1 Patient Misidentification in Laboratory Medicine A Qualitative Analysis of 227 Root Cause Analysis Reports in the Veterans Health Administration Edward J. Dunn, MD, ScD; Paul J. Moga, DO, PhD N Context. Mislabeled laboratory specimens are a common source of harm to patients, such as repeat phlebotomy; repeat diagnostic procedure, including tissue biopsy; delay in a necessary surgical procedure; and the execution of an unnecessary surgical procedure. Mislabeling has been estimated to occur at a rate of 0.1% of all laboratory and anatomic pathology specimens submitted. Objective. To identify system vulnerabilities in specimen collection, processing, analysis, and reporting associated with patient misidentification involving the clinical laboratory, anatomic pathology, and blood transfusion services. Design. A qualitative analysis was performed on 227 root cause analysis reports from the Veterans Health Administration. Content analysis of case reports from March 9, 2000, to March 1, 2008, was facilitated by a Natural Language Processing program. Data were categorized by the 3 stages of the laboratory test cycle. Results. Patient misidentification accounted for 182 of 253 adverse events, which occurred in all 3 stages of the test cycle. Of 132 misidentification events occurring in the preanalytic phase, events included wrist bands labeled for the wrong patient were applied on admission (n = 8), and laboratory tests were ordered for the wrong patient by selecting the wrong electronic medical record from a menu of similar names and Social Security numbers (n = 31). Specimen mislabeling during collection was associated with batching of specimens and printed labels (n = 35), misinformation from manual entry on laboratory forms (n = 14), failure of 2-source patient identification for clinical laboratory specimens (n = 24), and failure of 2-person verification of patient identity for blood bank specimens (n = 20). Of 37 events in the analytic phase, relabeling all specimens with accession numbers was associated with mislabeled specimen containers, tissue cassettes, and microscopic slides (n = 27). Misidentified microscopic slides were associated with a failure of 2-pathologist verification for cancer diagnosis (n = 4), and wrong patient transfusions were associated with mislabeled blood products (n = 3) and a failure of 2-person verification for blood products before release by the blood bank (n = 3). There were 13 events in the postanalytic phase in which results were reported into the wrong patient medical record (n = 8), and incompatible blood transfusions were associated with failed 2-person verification of blood products (n = 5). Conclusions. Patient misidentification in the clinical laboratory, anatomic pathology, and blood transfusion processes were due to a limited set of causal factors in all 3 phases of the test cycle. A focus on these factors will inform systemic mitigation and prevention strategies. (Arch Pathol Lab Med. 2010;134: ) Specimen misidentification in laboratory medicine can have significant consequences for patients. Mislabeled tissue specimens will often result in harm to patients, such as undergoing unnecessary surgical or diagnostic procedures. In other cases, patients will be subjected to unnecessary diagnostic studies or experience significant delays in the treatment of medical conditions they never knew they had. Mislabeling specimens can lead to unnecessary hospitalizations or failure to treat unreported conditions. If a patient suffers no physical harm from Accepted for publication April 30, From the Lexington VA Medical Center, and the Department of Surgery, University of Kentucky, Lexington (Dr Dunn); and the Veterans Health Administration, Division of Ambulatory Care and Integrative Medicine, VA Ann Arbor Healthcare System, Ann Arbor, Michigan (Dr Moga). The authors have no relevant financial interest in the products or companies described in this article. Reprints: Edward J. Dunn, MD, ScD, Lexington VA Medical Center, 1101 Veterans Dr, Lexington, KY ( edward.dunn@va.gov). specimen misidentification, the prospect of being subjected to additional procedures is not without risk. Such additional procedures, such as prostate biopsy, colonoscopy, fine-needle aspiration of the lung, phlebotomy, or other diagnostic testing, are not without risks that have associated opportunity costs to patients, families, the health care system, and society. We set 3 objectives for this study. The first was to develop an algorithm for the categorization of adverse events in laboratory medicine mapped to the 3 phases of the testing cycle. The second was to better understand how and why patient specimens can be misidentified in the laboratory. In the third objective, we considered potential solutions to reduce the probability of future recurrence of these adverse events. Background Q-PROBES was established in 1989 as the first major interinstitutional quality improvement program sponsored by the College of American Pathologists; it has 244 Arch Pathol Lab Med Vol 134, February 2010 Qualitative Analysis of 227 RCA Reports in the VHA Dunn & Moga

2 yielded a multitude of studies evaluating indicators of laboratory quality and led to Q-TRACKS in 1998, which has developed measures for tracking laboratory quality. In a Q-PROBES study of cases from 417 health care institutions, Nakhleh and Zarbo 1 identified laboratory specimen accessioning and identification deficiencies in cases (6%). The most common deficiency was no clinical history or diagnosis present on the requisition slip. 1 Valenstein et al 2 surveyed 127 clinical laboratories for information about specimen misidentification and found 345 adverse events reported during a 5-week period, or a rate of 55 events per 1 million billable tests. Most of the events caused material inconvenience to patients without causing harm. Extrapolating these data to all clinical laboratories in the United States, the authors estimated that more than adverse events per year result from patient misidentification of patients laboratory specimens. Wagar et al 3 reported a longitudinal study of laboratory specimen misidentification. Of 4.29 million specimens and 2.3 million phlebotomy requests during a 26-week period, misidentified specimens were found, for a rate of approximately 1 in 1000 opportunities. The 3 most common categories of misidentification events were mislabeled specimens (1%), requisition/specimen mismatches (6.3%), and unlabeled specimens (4.6%). The authors implemented 3 patient safety projects: (1) expanding phlebotomy service to 24/7; (2) introduction of an electronic event-reporting system; and (3) activation of an automated processing system. Trend analysis showed a decrease in all misidentification events during a 26-month period. Valenstein and Sirota 4 concluded that specimen misidentification occurs most frequently in the preanalytic phase of the laboratory test cycle, with specimen collection, labeling, and transport to the laboratory. They estimated the misidentification rate in transfusion medicine with blood products to occur in 0.05% of specimens, with a much higher rate of approximately 1% of specimens for the clinical laboratory. However, they acknowledged that the multiple specimen transfers and handoffs with anatomic pathology specimens may confer the highest misidentification rate of all laboratory specimens. 4 Zarbo and D Angelo 5 decried the risks for misidentification in surgical pathology as directly related to transfer points in sequential processes with identification defects occurring in mostly manual work processes. They found that 78% of defects were related to the slide-labeling processes, citing pencil writing on slides, affixing labels to slides after staining, and mislabeled cassettes as contributing factors. 5 In June 2008, Paxton 6 described the College of American Pathologists 2007 Q-Probes report from the College of American Pathologists Quality Practices Committee, titled Specimen mislabeling errors. In this study, Wagar et al reviewed 3.4 million clinical laboratory specimens collected at 147 institutions and identified 3043 mislabeled or unlabeled specimens. They calculated a median specimen misidentification rate for US laboratories to be 1.31 per 1000 specimens. The chief recommendations from this study were to institute ongoing quality monitors for specimen identification and to have 24/7 phlebotomy services for centralized specimen collection for inpatients. 6 With regard to transfusion medicine, in a Japanese study based on a survey of 777 hospitals during a 4-year period, 115 hospitals (20%) had experienced ABO blood type mismatched transfusion reactions. These cases were due to misidentification of blood bags (43%), incorrect blood typing (15%), and failure to identify the patient (11.5%). 7 In a prospective cohort study of surgical specimens, Makary et al 8 found 91 surgical specimen misidentification cases from a total of specimens, for a rate of 4.3 per These misidentification events were due to 18 unlabeled specimens, 16 empty containers, 16 laterality incorrect, 14 incorrect tissue site, 11 incorrect patient, 9 no patient name, and 7 no tissue site. These events were more prevalent when specimens were collected from outpatient clinics (53 of ; 0.512%) than from an operating room (38 of ; 0.346%). The authors estimated from their data an annualized rate of surgical specimen misidentification in their health system of 182 mislabeled specimens per year. 8 The Joint Commission for the Accreditation of Healthcare Organizations has issued national patient safety goals that apply to laboratory medicine services, including surgical pathology. The first 2008 Laboratory Services National Patient Safety Goal is: Improve the accuracy of patient identification. 9 The College of American Pathologists has issued its own national goals for patient safety in the laboratory, the first of which is to [i]mprove patient and sample identification. 10 Improvement Efforts for Specimen Identification To improve the accuracy of patient identification in the laboratory, the underlying contributing factors that have led to misidentification must be understood. Nakhleh 11 identified many of the important contributing factors leading to patient adverse events in surgical pathology: variable input, complexity, human intervention, time constraints, handoffs, and inflexible hierarchic culture. Compared with the clinical laboratory and transfusion service, surgical pathology was acknowledged as more vulnerable to specimen misidentification because of the number of handoffs with tissue specimens: collection, labeling, transport, accessioning, dissection, transfer to cassette, transfer to block, transfer to slide, transcription of findings, and reporting. There has been considerable interest in the laboratory medicine community for improving specimen identification. Most agree that automation of the specimen processes will enhance patient safety, citing the aphorism that humans do poorly at routine repetitive tasks [m]achines on the other hand are best for these tasks. 11 Many of the processes in the clinical laboratory and some in surgical pathology have been successfully automated. Wagar et al 3 demonstrated a significant decrease in mislabeled specimens during a 26-month period after implementing 24/7 laboratory medicine phlebotomy service, electronic event reporting, and an automated processing system. Manual labeling of tissue containers, requisition slips, and microscopic slides remains a common practice today. The batching of specimens and printed labels for multiple patients is also a prevalent practice. One solution to address this problem that has been applied in several health systems is to use wireless bar code technology to match the identification of patient and specimen at the episode of specimen collection and affix the specimen label generated at the point of care. 3 Hayden et al 12 Arch Pathol Lab Med Vol 134, February 2010 Qualitative Analysis of 227 RCA Reports in the VHA Dunn & Moga 245

3 Table 1. Laboratory Medicine Root Cause Analysis (RCA) Events, (N = 253 Adverse Events From 227 RCA Reports) No. (%) Preanalytic phase Wrong patient medical record entry 39 Mislabeled specimens: batching labels and specimens 35 Failure: 2-source identification at specimen collection or transport 24 Failure: 2-person verification of blood bank specimen 20 Lost specimen from collection to laboratory 15 Laboratory form misinformation with manual data entry 14 No laboratory order or no label on specimen 3 Total preanalytic 150 (59) Analytic phase Wrong accession number applied to new label 27 Technical failure: process, equipment, technology 21 Wrong interpretation: clinical laboratory or anatomic pathology 7 Lost specimen in laboratory 7 Failure: 2-person verification of blood product in blood bank 3 Mislabeled blood product in blood bank 3 Total analytic 68 (27) Postanalytic phase Delay in reporting clinical laboratory or anatomic pathology results 16 Reporting into wrong patient medical record 8 Failure: 2-person verification of blood product before transfusion 5 Reporting inaccurate results: clinical laboratory, anatomic pathology 5 Reporting clinical laboratory result to wrong provider 1 Total postanalytic 35 (14) Total 253 reported their experience in using bar code identifiers and handheld personal digital assistants supporting real-time order verification. They estimated a decreased rate of misidentification during the 3-year course of their study, and they estimated that 62 mislabeling events were prevented in the first year of operation. 12 Davies et al 13 demonstrated that a bar code patient identification system with handheld computers applied in cardiac surgery improved blood sample collection, the collection of blood from blood refrigerators, and the documentation of transfusion. MATERIALS AND METHODS Research Design A qualitative data analysis was conducted on a data set of 227 root cause analysis (RCA) reports extracted from the database of the Veterans Affairs (VA) National Center for Patient Safety (NCPS). These data were self-reported narratives of RCA cases that were submitted by VA medical centers to the NCPS from March 9, 2000, to March 1, All RCA case reports were organized in a format that followed a specific protocol as outlined in the Veterans Health Administration (VHA) National Patient Safety Handbook, available on the NCPS Web site. 14 From a total of individual and aggregate RCA case reports in our database, we built a data set of 227 RCA reports (Table 1) of 253 adverse events in laboratory medicine through an iterative process of data mining and data cleaning. We identified 26 RCA reports with multiple adverse events. Of those reports, 23 cases included 2 events, and 3 cases had 3 events. We identified 182 RCA reports (72%) involving patient misidentification in laboratory medicine (Table 2). Our search of the RCA database was an iterative process that generated 2 summary reports 15,16 in 2006 and culminated in a comprehensive search employing a Natural Language Processing (NLP) program, PolyAnalyst, a data and text mining software program (Megaputer Intelligence Inc, Bloomington, Indiana). Primary and secondary search terms that were employed are listed in Table 3. These search terms were used for text mining of the narrative data using the NLP program. A combination of data-driven discovery and automated categorization of reports in the RCA database facilitated significant improvements in the quality and speed of analyses compared with traditional key word searching and manual analysis of text. The NLP program facilitated our detection of common patterns and emerging trends of causal contributing factors specific to close calls and adverse events in laboratory medicine. Study Population The unit of analysis for this study was the VA medical center providing laboratory medicine services for veterans. Each case report represented an individual event or, in only 4 instances, aggregated events, that involved laboratory medicine services occurring in VA medical centers. During the 8-year timeframe for this study, the VHA consolidated from more than 170 health systems into 153 VA health systems providing these services for inpatient and ambulatory care by medical centers and affiliated community-based outpatient clinics. The community-based outpatient clinics have typically collected patient specimens Table 2. Laboratory Medicine Misidentification Events, (n = 182) Laboratory Medicine Events No. (%) Laboratory medicine RCA events Total laboratory medicine RCA events 253 (100) Laboratory medicine misidentification events 182 (72) Laboratory medicine misidentification events Preanalytic misidentification 132 (73) Analytic misidentification 37 (20) Postanalytic misidentification 13 (7) Total misidentification 182 (100) Abbreviation: RCA, root cause analysis. 246 Arch Pathol Lab Med Vol 134, February 2010 Qualitative Analysis of 227 RCA Reports in the VHA Dunn & Moga

4 Primary Search Term Identification Laboratory Misidentification Specimen Table 3. Search Terms for Mining Root Cause Analysis Database a Secondary Search Terms patient, patient bar code, patient identifier anatomic pathology, transfusion service, chemistry, hematology, microbiology, processing, transfusion patient, specimen accession, biopsy, blood, diagnosis, handle/handling, lab/laboratory, label, mislabel, missing, notification, patient, pathology, process, reporting, result, surgical, tissue, transport a Primary and secondary search terms were developed from text mining of narrative data facilitated by the Natural Language Processing program. from their patients and forwarded them to their affiliated medical centers for laboratory processing, analysis, and reporting of the results. The RCA case reports for this study involved veterans receiving care in these health systems. Data Mining and Search Progression in NLP Our data search was augmented by the NLP software program. This application facilitated the creation of text fields associated with RCA case reports relevant to patient identification in laboratory medicine. This program enabled data cleansing through text analysis. After the RCA database was loaded into the program ( Dataset Link icon; Figure 1), general search terms with embedded event categories were selected for the initial search ( Dimension Matrix icon; Figure 1). Primary search terms, such as laboratory, misidentification, delay in diagnosis/treatment, and correct surgery, were used, and these terms were expanded to include specimen and identification. Text fields incorporating the primary and secondary search terms listed in Table 1 were created and used to refine the search. Next, specific cases were identified and added to the data set, and other cases were removed in the process of building a relevant data set within the boundaries of the study ( Create Text Field and Filter Column icons; Figure 1). Data mining was further refined as to location of event and major event categories, or root nodes (Figure 2). Finally, the identified RCA case reports were compiled for export into a data file suitable for individual case reviews. As of this writing, the NCPS RCA database included RCA reports, of which were individual cases and 4477 were aggregate reviews involving multiple similar cases from the same medical center (NCPS RCA database, accessed June 16, 2008). Of the 227 RCA reports identified in our data set, 223 were individual reports and 4 were aggregate reviews, which involved 126 individual events. In every case report, at least 1 patient was harmed at some level. The level of patient harm ranged from the inconvenience of a second phlebotomy to the risk of undergoing an unnecessary medical diagnostic procedure or unnecessary surgical procedure, or significant delays in medical treatment. Therefore, the minimum number of patients harmed in our data was In each of the anatomic pathology cases, at least 2 patients were harmed one who had undergone an unnecessary diagnostic or therapeutic procedure and another who experienced delay in diagnosis or definitive treatment for a medical condition. Forty-two variables were coded for each RCA case report. Of these, 8 were selected for review. These included case identification, event date, station number, case description or final understanding (a summary of the event), root cause contributing factors (why the event happened), action (recommendations for reducing the probability of future recurrence), and the short phrase categories event ( selected event 1 ) and subcategory event ( selected event 2 ). Each RCA report was reviewed and coded by the authors independently. A unique coding algorithm was developed Figure 1. Natural Language Processing map for root cause analysis data. CSV, comma separated values. Arch Pathol Lab Med Vol 134, February 2010 Qualitative Analysis of 227 RCA Reports in the VHA Dunn & Moga 247

5 Figure 2. Taxonomy for Natural Language Processing search of laboratory medicine root cause analysis reports in the National Center for Patient Safety database. during the process of case reviews (Table 4). During the course of our case reviews, we built our coding algorithm through a process of iterative improvements by refining our categorization of the RCA events. As a result of this process, we established major categories and subcategories for coding our data. We also added a variable for patient harm, coded as 0 or 1. After independent case reviews and coding, the authors reconciled their differences in coding by a second collaborative review of every case report. We found minor differences in our coding in 10% of cases. Initial coding was done according to major laboratory test cycle categories as described in the Department of Health and Human Services Clinical Laboratory Improvement Amendments of This law was the basis for federal regulation of all clinical laboratories testing specimens from humans. The major test cycle categories included preanalytic (from specimen collection until laboratory transport), analytic (from specimen accession into the laboratory until completion of laboratory analysis with result), and postanalytic (from determination of laboratory result until reporting of result or, in the case of the transfusion service, distribution of a blood product). The laboratory result could be from the hematology, chemistry, or microbiology laboratory, from anatomic pathology, or from the transfusion service testing for future blood product administration. Each RCA case report was coded for major laboratory category and subcategory. In each case, the root cause contributing factors were identified, and action step recommendations to address those factors were noted. In addition, representative case examples were selected and summarized for each laboratory category and subcategory to enhance understanding. The focus of our analysis was on patient misidentification in laboratory medicine, which accounted for 88% of preanalytic cases, 47% of analytic cases, and 37% of postanalytic cases. Our goal in this study was to increase understanding of how and why these events occurred through qualitative analysis of emerging themes in the data. We were not attempting to estimate the prevalence of these events. 248 Qualitative Analysis of 227 RCA Reports in the VHA Dunn & Moga Arch Pathol Lab Med Vol 134, February 2010 RESULTS All 227 RCA case reports were examined by year for trends in reporting (Figure 3). Each RCA reported at least 1 adverse event related to laboratory medicine; 23 RCAs included 2 events, and 3 RCAs included 3 events. A total of 253 events were reported in 227 RCAs. There was a general increase in the number of laboratory medicine RCA reports from 2000 to The data were segregated according to phase in the laboratory test cycle (Figure 4). From a total of 253 laboratory medicine adverse events, 150 occurred in the preanalytic phase, 68 in the analytic phase, and 35 in the postanalytic phase. Patient misidentification events accounted for 73% of the 253 adverse events. The breakdown of the misidentification events included 73% preanalytic, 20% analytic, and 7% postanalytic phases. Seven of the RCA cases in our data set

6 Table 4. Root Cause Analysis (RCA) Coding Algorithm by Phase in Laboratory Test Cycle a Preanalytic A. Mislabeled specimen a. Printed labels (.1 patient in a location) b. Batched specimens (switched labels or same label for 2 or more patients) c. Wrong patient CPRS (medical record) entry d. Wrong patient wrist band applied on admission e. Failure of 2-person verification for blood draw for type and crossmatch f. Failure of 2-source identification at collection (full SSN, full name, birth date) B. Lost specimen (point of collection to laboratory) C. Laboratory form failures in access or execution (SF 515 and SF 518) a. CPRS access b. CPRS execution c. Handwritten execution d. Human factors design impact on usability D. No order for laboratory test E. No label on specimen container (blood, fluids, tissue) Analytic A. Mislabeled specimen a. Wrong accession number b. Handwritten labels (accession number, name, case number) c. Mislabeled tissue cassettes or blood specimen i. Batching tissue specimens ii. Batching blood specimens in laboratory d. Mislabeled pathology slides i. Batching pathology slides in the same slide folder e. Misidentified blood product i. Cautionary tag discrepant with SF 518 form B. Technical laboratory failure a. Contamination of tissue specimen b. Process failure leading to inaccurate result c. Equipment failure leading to inaccurate result d. Lost specimen in laboratory C. Misinterpretation of clinical laboratory and anatomic pathology data a. Missed diagnosis b. Failure of 2-pathologist verification of cancer diagnosis D. Wrong patient release of blood product from blood bank a. Failure of 2-person identification of blood product E. Wrong blood product release from blood bank for correct patient a. Failure of 2-person identification of blood product Postanalytic A. Reporting into wrong patient medical record (CPRS) a. Microscopic slide review of multiple patient cases in same folder b. Poor patient identification labeling for microscopic slides (accession number only) c. Cut and paste pathology report into wrong patient medical record d. Failure of 2-pathologist review of cancer diagnosis B. Wrong patient blood product transfusion a. Failure of 2-person identification of patient and blood product C. Delay in reporting clinical laboratory or anatomic pathology results D. Reporting results to wrong provider (physician or nurse) E. Failure to recognize and report critical laboratory values in a timely fashion F. Failure to report blood product availability G. Reporting inaccurate results a. Clinical laboratory b. Anatomic pathology Abbreviations: CPRS, computerized patient record system; SSN, Social Security number. a Taxonomy of events as described by RCA teams. This taxonomy identifies what happened in the event investigated by the team. It does not identify why or how this event occurred. For example, mislabeling from a failure of 2-source identification or batching of printed labels describes 2 observed facts but does not describe the underlying factors that allowed these events to occur. The purpose of the RCA is to uncover these factors. were coded as overlapping 2 phases in the laboratory test cycle. The RCA case reports were divided according to phase in the laboratory test cycle, event category, and subcategory. These events were examined by year for trends in reporting (Figure 5). Since 2001, the number of preanalytic laboratory medicine events has increased steadily, with an accelerated rate of increase observed in A similar accelerated rate of increase was also observed in analytic and postanalytic case reports for Onepotentialrationaleforthesteadyincreaseinthe rate of reported events could be an enhanced awareness of patient misidentification related to laboratory medicine through patient safety programs in VA facilities. In addition, development of the VA Bar Code Expansion project started in This project has focused on specimen labeling with application of wireless bar code technology at the episode of specimen collection. The Bar Code Expansion project has brought national attention to the risks of patient misidentification Arch Pathol Lab Med Vol 134, February 2010 Qualitative Analysis of 227 RCA Reports in the VHA Dunn & Moga 249

7 Figure 3. Laboratory medicine root cause analysis reports to the Veterans Affairs National Center for Patient Safety by year. through mislabeled clinical laboratory and anatomic pathology specimens. The RCA case reports were segmented by laboratory phases in the test cycle, as well as categories and subcategories of laboratory medicine events according to our algorithm (Table 4). Each reported event was coded for what happened. We did not code events for why the event happened (causal contributing factors) or for recommendations to reduce the probability of the event recurrence in the future. In our data set, patient misidentification was associated with 88% of preanalytic events, 47% of analytic events, and 37% of postanalytic events. The RCA case examples in abbreviated form appear in Table 5. These cases are organized by phase in laboratory test cycle, a brief description of what happened, causal factors attributed by the RCA teams, and associated patient harm. COMMENT Since 2002, there has been a trend of increasing overall RCA case reports (Table 6). Quarterly reports of aggregate reviews for multiple similar events from the same facility became required during the period, which may explain the high rate of reporting during that time period. The reporting requirement of aggregate reviews was reduced to biannual frequency in 2007, which may explain the relative decrease in reporting these events in The frequency of RCA events in our database does not represent their prevalence, because we cannot assume 100% reporting of all events. However, the VHA has developed a much more reporting culture since the inception of the patient safety program in 1999; the larger number of RCA case reports that have accrued over this time period compared with the private sector is evidence of that fact. Therefore, the trends in reported laboratory medicine events are relevant to comparisons between VA medical centers within the context of the VA health system. The rate of increasing annual laboratory medicine related RCA reports followed the same trend of overall RCA case reporting. This trend was true for laboratory medicine reported events in all 3 phases of the test cycle, although case reports from the preanalytic phase had a much sharper increase (Figure 5). However, an accelerated rate of case reporting was observed for laboratory medicine events in 2007 that was also consistent with the sharp increase in individual RCA case reports (Figure 6). One explanation for accelerated reporting in 2007 could be the new minimum requirement, which began in 2007, for VA medical centers to submit 4 RCA reports per year. An alternate explanation could be an increased awareness of patient safety in the organizational culture of VA facilities. Finally, patient misidentification events related to laboratory medicine have had a higher profile because of the attention brought to VA facilities by the national program development of the Bar Code Expansion project, which began in Figure 4. Laboratory medicine adverse events by test cycle phase and patient misidentification reported to the Veterans Affairs National Center for Patient Safety, A MisID (misidentification) event is an adverse event in 1 of the 3 phases of the laboratory medicine testing cycle that results in patient misidentification. 250 Arch Pathol Lab Med Vol 134, February 2010 Qualitative Analysis of 227 RCA Reports in the VHA Dunn & Moga

8 Figure 5. Laboratory medicine adverse events by phase in test cycle, frequency, and year. Figure 6. Laboratory medicine events by test cycle, frequency, and year. Agg Rvws, aggregated reviews; RCA, root cause analysis. Of 227 RCA case reports, 72% were associated with patient misidentification from mislabeled specimens; failure to use 2 sources of patient identification, or failure of 2-person patient identification when drawing blood for a type and crossmatch; misinformation on laboratory forms; mislabeling specimens, tissue cassettes, and microscopic slides with the wrong accession number; failure of 2-pathologist reviews of microscopic sides with a cancer diagnosis; and wrong blood product transfusion associated with failure of 2-person identification prior to initiating transfusion (Figure 7). The major finding from our study was the considerable risk for patient harm due to laboratory specimen mislabeling. Several common themes were identified by RCA teams in our analysis of the narrative data. Patient misidentification in the preanalytic phase of the test cycle was largely due to mislabeling during the process of specimen collection. Several contributing factors identified by RCA teams were the batching of unlabeled specimens and the presence of printed labels from multiple patients in common areas of the emergency department, operating room suites, and nursing units. These labels were often collocated in proximity to a common printer in a clinical unit. Another important factor was the collocation of patients in the same clinical units, such as the operating room, clinic, med-surg unit, or emergency department, with similar names, Social Security numbers, and birthdays. Surgical specimens from different patients were submitted in the same container to the laboratory. Laboratory tests were ordered for the wrong patient because of selection from a menu of patients with similar names and last 4 digits of their Social Security number from Vista, the VA electronic medical information system. Patients had the wrong patient wrist band affixed on admission, which was due to the selection from a menu of patients with similar names, last 4 Social Security number digits, and birthdays. Manual entry of identifying information on outdated and user-unfriendly laboratory forms, such as the SF 515 or SF 518, contributed to patient misidentification. Several RCA teams cited these forms as user friendly, with detailed analysis from a usability perspective. In addition, the elimination of the embossed addressograph patient identification cards requiring manual entry of identifying patient information was attributed by RCA teams as a vulnerability for patient misidentification on these forms. In the analytic stage of the test cycle, manual entry of accession numbers on laboratory specimen labels after they had been received into the laboratory contributed to specimen misidentification. Mislabeling of microscopic slides was also associated with manual entry and limiting the patient identifier to accession number only. Batching multiple slides from different patients in the same folder for microscopic review was associated with pathologists reporting findings into the wrong patient s medical record. The postanalytic stage comprised reporting laboratory results into the wrong medical record, delays in reporting critical results, and reporting inaccurate results. Arch Pathol Lab Med Vol 134, February 2010 Qualitative Analysis of 227 RCA Reports in the VHA Dunn & Moga 251

9 Table 5. Root Cause Analysis (RCA) Case Examples of Patient Misidentification Test Cycle Phase (Year) What Happened? Causal Factors Patient Harm Preanalytic (2003) Mislabeled FNA specimens with printed labels from previous patient Printed labels remaining from previous patient in same OR suite Preanalytic (2005) Seventy-five patients with mislabeled blood specimens: ward, clinic, ED Batching blood specimens and printed labels; failure 2-source identification Preanalytic (2007) Wrong patient type and crossmatch Dialysis tech drew blood from wrong and transfusion patient: failure 2-person verification of patient identification Preanalytic (2004) Mislabeled prostate Bx specimens Batching specimens and printed labels in OR from previous patient Preanalytic (2002) Preanalytic (2006) Preanalytic (2004) Preanalytic (2005) Preanalytic (2007) Mislabeled surgical Bx specimen by Circ. RN Mislabeled blood specimens from ER Mislabeled colon polyp from GI laboratory Patients in same clinic switched on procedure schedule Patient identification wrist band with manual data entry. Date of birth applied in lieu of last 4 digits of the SSN (copied from SF 518) Two patients with same name but different middle initial in CPRS; failure 2-source patient identification Batching blood specimens and printed labels in ER; failure 2-person patient identification Printed labels from previous patient in same procedure room Two patients in clinic with similar last name, SSN; failure 2-source identification Manual data entry of wrist band; SF 518 usability Patient A: wrong patient pulmonary resection; patient B: delay in Dx, Rx Blood specimen redraws Hemolytic transfusion reaction Patient A: unnecessary radical prostatectomy; patient B: delay in Dx, Rx Malignant breast lymphoma Dx entered into wrong patient medical record Wrong patient blood transfusion with transfusion reaction Repeat colonoscopy Wrong patient bronchoscopy (supposed to have colonoscopy) Blood specimen redraw Preanalytic (2005) Errant data entry into SF 518 form Manual data entry into SF 518 form Wrong blood product transfusion Preanalytic (2007) Sixteen patients mislabeled tissue and blood specimens Manual entry SF 515 and SF 518 forms; printed labels from previous case Tissue Bx redone; blood redraws Analytic (2003) Mislabeled specimens with wrong accession numbers Manual entry of accession numbers; laboratory environment: human factors Wrong patient laboratory data in medical record; Analytic (2001) Analytic (2001) Mislabeled cassettes leading to mislabeled pathology slides Mislabeled cassettes leading to mislabeled pathology slides Manual labeling; slides from multiple cases in same folder with accession numbers only Manual numbering of cassettes blood redraws Patient A: unnecessary pulmonary lobectomy; patient B: delay in Dx, Rx Patient A: unnecessary hysterectormy; patient B: delay in Dx, Rx Analytic (2007) False-positive diagnosis Failure 2-pathologist review of Bx slides Patient A: unnecessary radical prostatectomy; patient B: delay in Dx, Rx Analytic (2007) False-negative diagnosis (switched chest/facial specimens) Failure 2-pathologist review of Bx slides Analytic (2002) False-negative Dx of Papanicolaou Failure 2-pathologist review of Bx slides smear Analytic (2004) Mislabeled pathology slides Batching slides with printed labels for all surgery cases of day; accession number only identification Postanalytic (2001) Postanalytic (2003) Postanalytic (2003) Postanalytic (2007) Postanalytic (2002) Postanalytic (2005) Postanalytic (2006) Pathology report entry into wrong medical record Wrong patient medical record entry of lung small cell carcinoma Dx Wrong patient medical record entry of clinical laboratory test results Pathology report into medical record without clinician response Anesthesia provider selected wrong blood product for patient in OR Dermatology Bx report to wrong patient medical record Three patients received treatments that were reported into the wrong medical records Manual slide labeling; slides from multiple cases in same folder with accession numbers only Two cases of transbronchial Bx slides in same folder with accession number labels only Failure of interface with Vista preventing automatic download to CPRS; failure of 2- source identification View alert of pathology Dx did not require provider signature confirming receipt Blood products for multiple patients comingled in OR central refrigerator Two patients with same last name and last 4 digits of SSN; manual entry into SF 515 form Three patients received wrong wrist bands because of similar names and last 4 digits of SSNs of other patients in system Two-month delay in Dx, Rx of malignant melanoma Nine-month delay in Dx of cervical carcinoma Patient A: unnecessary radical prostatectomy; patient B: delay in Dx, Rx Patient A: unnecessary radical prostatectomy; patient B: delay in Dx, Rx Patient A: unnecessary chemotherapy; Patient B: delay in Dx, Rx Wrong patient medical Dx and Rx Three-month delay in Dx and Rx of malignant melanoma Wrong blood product transfusion to surgical patient in OR Wrong patient diagnosis Patients received correct treatments but recorded in the wrong medical records Abbreviations: Bx, biopsy; Circ. RN, circulating registered nurse; CPRS, computerized patient record system; Dx, diagnosis; ED, emergency department; ER, emergency room; FNA, fine-needle aspiration; GI, gastrointestinal; OR, operating room; Rx, treatment; SSN, Social Security number. 252 Arch Pathol Lab Med Vol 134, February 2010 Qualitative Analysis of 227 RCA Reports in the VHA Dunn & Moga

10 Table 6. Root Cause Analysis (RCA) Case Reports and Aggregated Reviews by Year Year Aggregated Review RCA Total Total NA NA NA diagnosis and necessary medical treatment, blood specimen redraw, and repeat tissue biopsy, as well as wrong patient diagnosis, reporting, and treatment (eg, medications and blood products). Abbreviation: NA indicates no aggregated reviews from 1999 through Blood transfusion events were attributable to blood drawn from the wrong patient for type and crossmatch because of failure of confirming 2-source patient identifiers, misidentified blood products from discrepant attached forms, and failure of 2-person identification of blood product and patient prior to initiating the transfusion. Each RCA case was evaluated for harm to the patient, which was described in the analysis. Each event subcategory was analyzed from the perspective of why and how these events occurred. The authors reviewed recommendations offered by RCA teams to prevent recurrence of these events in the RCA reports. Patient harm was identified in virtually all adverse events and close calls from our study sample. Each episode of patient misidentification was found to affect more than 1 patient. Mislabeled biopsy specimens were associated with wrong patient surgical procedures (eg, hysterectomy, radical prostatectomy, pulmonary lobectomy), delay in necessary surgical procedures, delay in Usability of Laboratory Forms There were 14 reported cases involving errant manual entry of patient identifying data into standard forms that accompanied laboratory specimens and blood products. Misinformation on the SF 515 form for clinical laboratory and anatomic pathology and the SF 518 for blood bank specimens and products contributed to patient misidentification. Although these forms have been used successfully in the VA and the Department of Defense for decades, RCA teams cited similar usability challenges in both forms, such as small font size, limited white space for manual data entry, and spaces for extraneous data elements. These forms had been originally designed for addressograph stamping of embossed patient identification cards. In 2003, because of privacy concerns, the embossed cards were replaced with the Veterans Identification Card, which contains the patient identifying data in a magnetic strip. However, the laboratory and blood bank forms had not been redesigned since that change occurred. These problems will be addressed with the implementation of the VA Laboratory Reengineering Project. This program will automate the SF 515 and SF 518 forms with digital templates, redundant patient identification screens, and electronic signatures to improve the accuracy of information transfer. Limitations Root cause analysis case reporting is required for all VA medical centers. Nevertheless, we cannot assume 100% reporting of all adverse events in the VA. As source data in this study, RCA reports do not represent the prevalence of Figure 7. Laboratory medicine event frequency by phase in test cycle and contributing factors. ID, identification; bl, blood; Bl Bank, blood bank; clin lab, clinical laboratory; AP, anatomic pathology; Clin Lab, clinical laboratory; CL, clinical laboratory; man, manual; med, medical. Arch Pathol Lab Med Vol 134, February 2010 Qualitative Analysis of 227 RCA Reports in the VHA Dunn & Moga 253

11 Table 7. Recommendations From Root Cause Analysis Teams to Reduce Laboratory Medicine Adverse Events Intervention 1. Apply wireless bar code technology to real-time specimen labeling during specimen collection (validated in other health systems by Hayden et al 12 ) 2. Require full SSN as a unique identifier for patient identification. Use other assigned unique identification number if not the SSN. 3. View alert in Vista for patients with the same last name and last 4 SSN digits in the same health system 4. Veterans Affairs Laboratory Software Reengineering Project for laboratory automation 5. Preprocedure briefings and postprocedure debriefings in the OR and procedure rooms 6. 24/7 centralized phlebotomy service for specimen collection by laboratory medicine 7. Eliminate specimen label printing to common printer in clinical area 8. Eliminate all paper labels from the OR or procedure rooms as a necessary part of the room turnover process 9. Require electronic data entry for 2 of 3 patient identifiers on all specimen labels, including slides and microbiology specimens (name, full SSN, birth date) 10. Require data entry for the names of 2 health care professionals verifying patient identification as required fields on the automated SF 518 form necessary before blood is drawn for type and crossmatch and before blood product transfusion is initiated 11. Automate SF 515 and SF 518 forms accessible from CPRS/Vista. Electronic data entry required. No manual data entry. 12. Eliminate relabeling clinical laboratory and anatomic pathology specimens with accession numbers after they are received in the laboratory and move the accession numbering process forward to the original label generated immediately after specimen collection. Eliminate manual entry of accession numbers. 13. Forcing function of 2 pathologists to sign final pathology report with cancer diagnosis as required fields before report is accepted as complete in CPRS 14. View alert on computer to provider with required signature receipt acknowledging critical patient diagnosis laboratory medicine adverse events and close calls. Rather than focus on the prevalence of events, we were interested in how and why these events occurred and what could be done to reduce their probability of future recurrence. With that purpose in mind, we observed emerging themes from text analysis of these narrative reports. It is important to understand the limitations of RCA case reports. Although these reports are replete with rich narrative data, they do not include all of the patient clinical information that can be found in a medical record. In conducting an RCA, team members have access to the patient s medical record. Technical details, such as laboratory values, operative notes, and radiograph reports, cannot be found in an RCA report. However, the medical record is not part of an RCA narrative report. Another limitation of this study is linked to limitations inherent in RCA data. The strength of an RCA is directly related to rigor of inquiry by an interdisciplinary team. It is entirely possible that RCA teams in our data set missed important causal contributing factors associated with the events they were examining. Similarly, the recommendations of these teams are dependent on the rigor of their analysis and limited by hindsight bias, as suggested by Hofer and Hayward. 18 However, the authors admitted Effect 1. Standardizes automated labeling process; eliminates manual entry labeling; eliminates user-unfriendly laboratory forms with manual entry 2. Reduce patient misidentification due to patients with similar names and last 4 SSN digits. If the SSN is not used, an assigned unique identifier should be used. 3. Call attention to patient misidentification hazard 4. Push forward the accession number application during specimen collection to eliminate relabeling in the laboratory; links automated laboratory processes to laboratory information system 5. Confirm anticipated specimens: type, transport medium, laboratory tests, labeling, and postoperative disposition 6. Reduce patient misidentification from specimen mislabeling 7. Reduces probability of batching printed labels from multiple patients in a common area of a clinical unit 8. Reduces probability of affixing printed label from previous patient 9. Forcing function for reviewing 2 patient identifiers for every laboratory specimen collected and examined; reduces vulnerability associated with manual entry 10. Forcing function for 2-person verification of patient identity before blood is drawn for type and crossmatch and before blood product transfusion is initiated 11. Reduces probability of misinformation by eliminating manual entry of anatomic pathology and blood bank forms and increases form accessibility 12. Reduces probability of applying the wrong accession number introduced by the added vulnerability of relabeling specimens in the laboratory 13. Two-pathologist review with both reconciling patient identification will reduce incidence of patient misidentification and inaccurate cancer diagnosis 14. Forcing function for provider acknowledgment and response to critical patient diagnosis Abbreviations: CPRS, computerized patient record system; OR, operating room; SSN, Social Security number. their case study was largely one of clinical judgment and technical competence rather than one of systems issues, which an RCA is designed to address. Nevertheless, the value of RCA is its ability to explain behavior rather than make judgments of what should have been done. 19 Recommendations Many of the important RCA team recommendations from this study are displayed in Table 7. The focus of RCA team recommendations in the VHA is to reduce the likelihood of future adverse events in their respective medical centers. Some of the strongest recommendations included the following:. Application of wireless bar code technology at the bedside to confirm patient identity and affix a bar code label to a specimen immediately after collection.. Application of bar code technology to the blood transfusion process.. Use of unique patient identifier for selecting a patient medical record and for labels on all specimens and blood products (full Social Security number is currently the unique patient identifier in the VA). 254 Arch Pathol Lab Med Vol 134, February 2010 Qualitative Analysis of 227 RCA Reports in the VHA Dunn & Moga

12 . Automate laboratory forms limited to electronic data entry and eliminate all manual entry for specimen labeling.. Eliminate relabeling clinical laboratory and anatomic pathology specimens with accession numbers after they are received in the laboratory, and move the accession numbering process forward to the original label generated immediately after specimen collection.. Continuously available centralized phlebotomy service for hospital inpatients.. Eliminate all paper labels in the operating room with all room turnovers before admitting the next patient to the room.. Forcing function for 2-pathologist review as required documentation in final pathology report of all anatomic pathology slides with a cancer diagnosis. There is evidence that automating laboratory processes through all 3 phases of the laboratory test cycle has been demonstrated to reduce patient misidentification, increase efficiency, and increase the productivity of operations. When these improved processes were linked to an updated laboratory information system, the turnaround time for reporting laboratory results to clinical decision makers was reduced, and the quality of patient care was enhanced. 20 Conclusions The goal of this study was to enhance our understanding of system vulnerabilities that lead to adverse patient events related to laboratory medicine. Our focus was on patient misidentification from mislabeled laboratory specimens, failure of 2-source patient identification during specimen collection, and failure of 2-person verification of patient identity before collecting blood specimens for type and crossmatch or prior to blood product transfusions. We wanted to understand why and how these events occurred. Our purpose was also to stimulate ideas of what could be done to prevent future recurrence of these events. Coding of laboratory medicine events by 3 phases in the laboratory testing cycle, with event categories and subcategories, facilitated our data analysis and identification of emerging themes. Subsequent review of RCA team recommendations stimulated our thinking about further recommendations for consideration by health systems. If any of these interventions are implemented, future studies will be necessary to evaluate their effectiveness in reducing the likelihood of patient misidentification in laboratory medicine. We wish to acknowledge the efforts of Dea Mannos Hughes (New York Harbor VA Health System, New York, New York) and Carol Samples (VA National Center for Patient Safety, Ann Arbor, Michigan) for their work in data collection and summary reporting in the early phases of this study. We also wish to extend our appreciation to Scott McKnight, Aartee Ignaczak, and Jim Turner (VA National Center for Patient Safety) for their assistance in recent data collection and application of the NLP software program to facilitate our analysis. We thank the VA Ann Arbor Healthcare System Department of Pathology and Laboratory Medicine: Stephen Chensue, MD, service chief; Hedwig Murphy, MD, staff pathologist; and Aron Pollock, Lyn St. Dennis (surgical pathology), Joseph J. Mraz (blood bank), Bruce Haugen (hematology), Parul Shah (microbiology), and Cheryl Rollins (blood bank) for providing their insight into laboratory processes. Finally, we wish to thank Michael Brophy, Quynh Vantu, and Valerie Miller from the VA Central Office Laboratory Medicine, Washington, DC, for their insightful comments, which informed this study. References 1. Nakhleh R, Zarbo R. Surgical pathology specimen identification and accessioning: a College of American Pathologists Q-Probes Study of 1,004,115 cases from 417 institutions. Arch Pathol Lab Med. 1996;120(3): Valenstein P, Raab S, Walsh M. Identification errors involving clinical laboratories: a College of American Pathologist Q-Probes study of patient and specimen identification errors at 120 institutions. Arch Pathol Lab Med. 2006; 130(8): Wagar E, Tamashiro L, Yasin B, Hilborne L, Bruckner D. Patient safety in the clinical laboratory: a longitudinal analysis of specimen identification errors. Arch Pathol Lab Med. 2006;130: Valenstein PN, Sirota RL. Identification errors in pathology and laboratory medicine. Clin Lab Med. 2004;24(4): Zarbo R, D Angelo R. The Henry Ford production system: effective reduction of process defects and waste in surgical pathology. Am J Clin Pathol. 2007;128: Paxton A. Punching a hole in specimen ID errors. CAP Today. June 2008: %2Fportlets%2FcontentViewer%2Fshow&_windowLabel5cntvwrPtlt&cntvwrPtlt% 7BactionForm.contentReference%7D5cap_today%2Fcover_stories%2F0608_Punching_ a_hole.html&_state5maximized&_pagelabel5cn. Accessed July 24, Murakami J. Present state of transfusion errors. Rinsho Byori. 2003;51(1): Makary M, Epstein J, Pronovost P, Millman E, Hartmann E, Freischlag J. Surgical specimen identification errors: a new measure of quality in surgical care. Surgery. 2007;141(4): Joint Commission Laboratory Services National Patient Safety Goals. Joint Commission Web site. NationalPatientSafetyGoals/08_lab_npsgs.htm. Accessed June 6, CAP Laboratory Patient Safety Plan. CAP Web site. apps/cap.portal?_nfpb5true&cntvwrptlt_actionoverride5%2fportlets%2fcontent Viewer%2Fshow&_windowLabel5cntvwrPtlt&cntvwrPtlt%7BactionForm.content Reference%7D5patient_safety%2Flaboratory_patient_safety_plan.html&_state5 maximized&_pagelabel5cntvwr. Created April 17, Accessed June 6, Nakhleh R. Patient safety and error reduction in surgical pathology. Arch Pathol Lab Med. 2008;132: Hayden R, Patterson D, Jay D, et al. Computer-assisted bar-coding system significantly reduces clinical laboratory specimen identification errors in a pediatric oncology hospital. J Pediatr. 2008;152(2): Davies A, Staves J, Kay J, Casbard A, Murphy M. End-to-end electronic control of the hospital transfusion process to increase the safety of blood transfusion: strengths and weaknesses. Transfusion. 2006;46(3): VHA Handbook VHA National Patient Safety Handbook. NCPS Web site. Accessed June 13, Samples C, Dunn E. Patient harm from anatomic surgical specimen management in the operating room. Topics in Patient Safety. 2006;6(4): Dunn E, Samples C. Specimen management in the laboratory opportunities for improvement. Topics in Patient Safety. 2006;6(5): Department of Health and Human Services. Clinical Laboratory Improvement Amendments of Public Law Department of Health and Human Services Web site. Published October 31, Accessed June 16, Hofer T, Hayward R. Are bad outcomes from questionable clinical decisions preventable medical errors?: a case of cascade iatrogenesis. Ann Intern Med. 2002;137(5): Dekker S. The Field Guide to Human Error Investigations. Burlington, VT: Ashgate Publishing Co; 2002: Blick K. No more STAT testing. MLO Med Lab Obs. 2005;37(8):22, 24, 26. Arch Pathol Lab Med Vol 134, February 2010 Qualitative Analysis of 227 RCA Reports in the VHA Dunn & Moga 255