Classifying Laboratory Incident Reports to Identify Problems That Jeopardize Patient Safety

(1)

Classifying Laboratory Incident Reports to Identify

Problems That Jeopardize Patient Safety

Michael L. Astion, MD, PhD,

1

_{Kaveh G. Shojania, MD,}

2

_{Tim R. Hamill, MD,}

3

_{Sara Kim, PhD,}

4

and Valerie L. Ng, MD

3

Key Words: Incident reports; Patient safety; Adverse events; Laboratory tests; Laboratory error DOI: 10.1309/8U5D0MA6MFH2FG19

A b s t r a c t

We developed a laboratory incident report

classification system that can guide reduction of actual and potential adverse events. The system was applied retrospectively to 129 incident reports occurring during a 16-month period. Incidents were classified by type of adverse event (actual or potential), specific and potential patient impact, nature of laboratory

involvement, testing phase, and preventability. Of 129 incidents, 95% were potential adverse events. The most common specific impact was delay in receiving test results (85%). The average potential impact was 2.9 (SD, 1.0; median, 3; scale, 1-5). The laboratory alone was responsible for 60% of the incidents; 21% were due solely to problems outside the laboratory’s authority. The laboratory function most frequently implicated in incidents was specimen processing (31%). The preanalytic testing phase was involved in 71% of incidents, the analytic in 18%, and the postanalytic in 11%. The most common preanalytic problem was specimen transportation (16%). The average preventability score was 4.0 (range, 1-5; median, 4; scale, 1-5), and 94 incidents (73%) were preventable (score, 3 or more). Of the 94 preventable incidents, 30% involved cognitive errors, defined as incorrect choices caused by insufficient knowledge, and 73% involved noncognitive errors, defined as inadvertent or unconscious lapses in expected automatic behavior.

The Institute of Medicine report, To Err Is Human,1

generated widespread interest in medical error and adverse events in health care, as well as strategies for reducing them.2_{Existing studies have characterized the epidemiology}

and impact of actual and potential adverse events in a variety of settings.3-8 _{However, these studies, including}

benchmark population studies of adverse events among hospitalized patients,3,8 _{do not provide explicit information}

about adverse and potential adverse events related to clin-ical laboratory services.

Knowledge about actual and potential adverse events related to laboratory services comes from a small number of studies that have focused on the rate of laboratory errors and the classification of the errors by cause, phase of testing, responsible party, and extent of harm to the patient9-19_(for

review, see Bonini et al20_{). Two studies investigated}

prob-lems occurring across most hospital laboratory services,9,10

while other studies focused on problems within specific laboratory areas such as clinical chemistry,11,12 _blood

banking,13,14_{stat testing,}15_{and genetic testing.}16

The study of actual and potential adverse events related to laboratory testing would benefit from a commonly used, rigorous classification scheme that identi-fies preventable problems most likely to lead to patient injury and suggests solutions for those problems.20 _We

describe a tool that characterizes the nature of important errors and adverse events involving laboratory services. We evaluated the retrospective application of this tool to 129 incidents identified by internal laboratory incident reports or risk management reports as involving some aspect of the clinical laboratory services at a major aca-demic medical center.

by guest on July 5, 2016

http://ajcp.oxfordjournals.org/

(2)

Materials and Methods

Participating Institution

The participating institution is an academic medical center in the United States. The institutional review board of the host academic medical center approved the study.

The medical center laboratory system receives 750,000 requisitions and reports 3.8 million test results per year, with approximately 55% of the results pertaining to inpatients and 45% to outpatients. The laboratory system consists of a core laboratory that performs the majority of testing, an intermediate service laboratory in a smaller affiliated hospital, and 3 outpatient locations used for phlebotomy and specimen processing. Specimen transport between labora-tory locations is by automobile or foot courier. Inpatient specimens are transported to the core laboratory by foot or by a dumbwaiter located on each hospital floor. There is no pneumatic tube system.

Laboratory employees perform the majority of outpa-tient phlebotomy services, and hospital personnel not managed or employed by the laboratory (medical assistants organized as a mobile phlebotomy team, nurses, physicians, and a variety of trainees) perform the majority of inpatient phlebotomy services.

Incident Report Database

This retrospective study analyzed an existing database residing in the clinical laboratory. The database covered the period June 2000 (when the database was initiated) through September 2001, and included 168 laboratory-related inci-dents reported through the laboratory’s internal incident reporting system or through the hospital’s risk management reporting system. At the medical center studied, incident reports are encouraged when a problem occurs that actually or potentially has an important negative impact on patient care. The risk management incident reports were those referred by the medical center’s risk management team to the clinical laboratory director because the incidents were related to laboratory testing services.

Each case in the database relates to a specific incident report and contains the following information: (1) location of the patient involved in the incident, for example, the medical intensive care unit; (2) a synopsis of the incident from the incident report form; (3) key findings of the investigation of the incident report; (4) corrective action related to the incident; and (5) assessment of the impact to the patient.

This study includes cases from the database in which the incident report met all 4 of the following criteria: (1) not related to an employee injury; (2) not related to the adminis-tration of blood products; (3) related to a specific incident,

rather than a general complaint; and (4) related to an actual or potential patient injury and not solely a patient service complaint. Of the original 168 cases in the database, 129 (77%) met these criteria. Of these 129 cases, 44 (34%) were derived from risk management incident reports and 85 (66%) from internal laboratory incident reports.

Classification of Each Incident Report

Each incident in the database was classified as an actual or a potential adverse event and further characterized in terms of impact on the patient, responsibility for the incident, and phase of testing (preanalytic, analytic, postanalytic) involved in the incident. In addition, preventability was assessed for each incident. Because we were interested in developing a classification tool that could be used easily by others, we asked reviewers (see “Review of Cases”) to assess how well the classification scheme captured the salient features of each incident. Similarly, reviewers were asked for each case to list information that would have improved the accuracy of the classification had it been provided. Last, reviewers were provided with a general comment field to record any other free text comments about the case that were not recorded by the classification system.

Adverse and Potential Adverse Events

An actual adverse event was defined as an injury to a patient caused by medical management rather than by a disease process, in which the injury resulted in disability or prolonged hospital stay.3,8 _A_{potential adverse event}_was

defined as an error or incident that produced no injury but had the clear potential to do so. Such errors may have been intercepted before producing harm or may have reached the patient but, by good fortune, produced no injury. When the occurrence of injury could not be determined, incidents were classified as potential adverse events.

Specific and Potential Impacts on Patient Care

All cases, whether adverse or potential adverse events, were scored for the presence or absence of the following specific impacts: (1) specimen redrawn or recollected, (2) delay in receiving test results (receiving results after the time the results would have been expected in the absence of a laboratory service problem), (3) incorrect test results sent to provider, (4) none of the above, or (5) unable to determine. More than 1 of the first 3 categories could be selected. In addition, for adverse events only, the specific injury was described with a text comment.

For potential adverse events, the potential impact to patient care was scored using a scale graded from 1, very unlikely to adversely affect the patient’s health, to 5, very likely to adversely affect the patient’s health. For cases with insufficient information, the case was scored as “unable to

(3)

determine.” Because the purpose of the classification system is to detect the problems most likely to injure patients, the following scoring rules were adopted regarding turnaround times, incorrect results, and patient care settings. Incidents involving the following conditions could be scored as no less than 3: (1) total turnaround times of longer than 2 hours for any stat request, (2) total turnaround times of longer than 8 hours for routine requests on hospitalized patients, and (3) incorrect results reported. These conditions could be scored a 4 or 5 if exacerbating factors (eg, critically ill patient, loss of irreplaceable specimen, incorrect treatment given) also were present. Any incident involving a critical care setting, such as the intensive care unit, the emergency room, or the oper-ating room, started with a score of 3 and could be moved up or down by exacerbating or mitigating factors.

Responsibility for Incident

The categories for responsibility for the incidents were as follows: (1) the laboratory, (2) outside the authority of the laboratory, and (3) unable to determine. An incident could be assigned to both the laboratory and outside the laboratory. The incident was assigned to the laboratory if the problems involved personnel, equipment, services, or policies for which the laboratory was responsible. An incident was assigned as outside the authority of the laboratory if the problems involved personnel, equipment, services, or poli-cies for which the laboratory was not responsible. In addi-tion, problems involving the transportation of the specimen to the laboratory from within the academic medical center were defined as outside the authority of the laboratory. If a problem was assigned to the laboratory, the laboratory func-tion associated with the problem also was chosen. The 12 functions were phlebotomy, processing, chemistry, hema-tology, coagulation, microbiology, virology, immunology, molecular diagnosis, blood banking, laboratory information system, or unable to determine. More than 1 laboratory func-tion could be chosen.

Phase of Laboratory Testing

The categories used for the general phase of laboratory testing were preanalytic, analytic, postanalytic, and unable to determine, and cases could involve more than 1 phase of testing. The preanalytic phase was defined as all procedures from the time the test was ordered until the specimen was analyzed by an instrument or other method. It included ordering the test by the provider, specimen collection, imen labeling, specimen transportation, logging of the spec-imen into the laboratory information system, and specspec-imen processing. The analytic phase was defined as analysis of the specimen by automated, semiautomated, or manual methods. The postanalytic phase was defined as all proce-dures from the time a result was produced in the laboratory

until the provider interpreted the result. It included results verification in the laboratory, entry into the laboratory infor-mation system, and communication of the result to the provider by any of a number of methods, including computer display, fax, remote printer, mailed report, or oral communication.

After selecting the general phase of testing, 1 or more detailed classifications were selected within the general phase. For the preanalytic phase, the more detailed classifi-cations were as follows: (1) requisition incorrect; (2) patient injured during phlebotomy; (3) patient unhappy with phle-botomy customer service; (4) primary specimen or aliquot mislabeled or unlabeled; (5) no specimen collected; (6) incorrect tube used or “order-of-draw” problem; (7) spec-imen suboptimal or ruined; (8) specspec-imen lost or delayed in transport; (9) specimen delayed or lost in the laboratory; (10) specimen lost or delayed (unable to assign to inside or outside the laboratory); (11) failure to order, add, or modify a test; (12) data entry error or other information systems problem; and (13) other. For the analytic phase, the 3 more detailed classifications were (1) human error, (2) instrument error, and (3) other. For the postanalytic phase, the 3 detailed classifications were (1) results delayed, not reported, or reported to the wrong provider; (2) incorrect results reported because of postanalytic data entry error; and (3) other.

Preventability

A preventable problem was considered an error that was reasonably avoidable, in which the error was a mistake in performance or thought.21_{Preventability was}

scored on a scale of 1, definitely not preventable, to 5, definitely preventable. A score of 3 or more indicated a preventable incident.

Problems also were classified based on a cognitive psychology model because this classification helps design interventions for problems (see the “Discussion” section). Cognitive psychologists often divide tasks into 2 categories: attentional vs schematic.22,23_Attentional_{behaviors are}

asso-ciated with a strong mental effort, for example, learning a new piece of music, analyzing test results, or creating a treat-ment plan for a patient. Most defects in attentional behavior are cognitive errors (also known in the psychology literature as “mistakes”), which are defined as incorrect choices owing to insufficient knowledge, misinterpretation of available information, or application of the wrong cognitive rule. Examples from laboratory testing are choosing the wrong test because of lack of knowledge regarding diagnosing a particular disease and misinterpreting a Gram stain because of inexperience.

Schematic behaviors are automatic or unconscious, so that several different types of schematic behaviors often are carried out concurrently, for example talking to a patient

(4)

during phlebotomy while changing blood tubes. Most defects in schematic behavior are due to noncognitive errors (also known as “slips”), which are defined as inadvertent or unconscious lapses in expcteed automatic behavior. Exam-ples from the clinical laboratory include an experienced phlebotomist using an incorrect sequence of blood tubes when obtaining specimens and data entry errors by experi-enced processing technicians or technologists.

The significance of the distinction between cognitive and noncognitive errors lies in the different responses to these 2 types of errors. Guarding against noncognitive errors typically calls for the use of checklists, introduction of machine automation, and other such strategies for avoiding the excepted lapses in concentration and vigilance that char-acterize human behavior. By contrast, strategies for reducing cognitive errors typically involve changes in training or supervision. Incidents were scored as due to cognitive errors, due to noncognitive errors, or unable to determine. An incident could involve both cognitive and noncognitive errors.

Confidence in Classification

For each case, the reviewer was asked: “What is your confidence in the classification of this case using the model?” Confidence was scored on a scale from 1 to 5, in which 1 was “not confident” that the classification scheme captured the salient features of the incident and 5 was “very confident.”

Review of Cases

A board-certified clinical pathologist (M.L.A) reviewed all 129 cases, and the results from this pathologist are shown in the “Results” section below. In addition, 30% of the cases were chosen at random and then reviewed by 2 other board-certified pathologists (T.R.H., V.L.N.). Interrater reliability was measured by using the Cohen κ statistic. There were a large number of items in the classification scheme, and the κ statistic was calculated for the following 5 higher order cate-gories deemed to be most important for directing quality improvement activities: type of event, authority for the prob-lems described in the incident, general phase of laboratory testing, preventability, and error type (if the incident was preventable).

Results

Results Overview and Examples of Cases

An overview of the classification of the 129 incidents is given in ❚Table 1❚. ❚Appendix 1❚gives 4 illustrative cases from the database, along with the classification of the case.

Actual vs Potential Adverse Events and Their Impact on Patients

Of the 129 incidents, 122 (95%) were potential adverse events only, 6 (5%) were actual adverse events only, and 1 (1%) involved actual and potential adverse events (Table 1). The frequencies of specific impacts on patient care are listed in ❚Table 2❚. The most common specific impacts were delay in receiving test results, which occurred in 110 cases (85%), and redrawing specimens, which occurred in 51 cases (40%). The 7 cases (5%) involving adverse events were phlebotomy-related injuries.

The 123 incidents involving potential adverse events were scored for potential impact on patient care. ❚Table 3❚ lists the frequency of occurrence of the 5 possible scores for potential impact and summarizes a typical incident

❚Table 1❚

Classification of 129 Laboratory-Related Incident Reports by Event Type, Responsibility, Phase of Testing,

Preventability, and Error Type if Preventable*

Category No. (%) of Incidents Event type

Actual adverse event 6 (5)

Potential adverse event 122 (95)

Actual and potential 1 (1)

Responsibility for incident

Laboratory 78 (60)

Outside laboratory 27 (21)

Laboratory and outside laboratory 16 (12)

Unable to determine 8 (6)

Phase of testing

Preanalytic 91 (71)

Analytic 21 (16)

Postanalytic 13 (10)

Preanalytic and analytic 1 (1)

Analytic and postanalytic 1 (1)

Preventable

Yes, noncognitive error 64 (50)

Yes, cognitive error 23 (18)

Yes, cognitive and noncognitive error 5 (4)

Yes, unknown error type 2 (2)

No 18 (14)

*_{See the “Materials and Methods” section for definitions.}

❚Table 2❚

Specific Impact on Patients as Described in 129 Incident Reports*

Specific Impact No. (%) of Incidents

Delay in receiving test results 110 (85)

Specimen redrawn or recollected 51 (40)

Incorrect results sent to provider 40 (31)

None 8 (6)

Specific injury (adverse event) 7 (5)

*_{The 7 cases (5%) with specific injuries (adverse events) were phlebotomy-related}

injuries and included hematoma (2 cases), hematoma possibly complicated by cellulitis (1 case), pain at the phlebotomy site (2 cases), excessive blood drawn from an infant (1 case), and fainting and sustaining a gash to the forehead a few minutes after phlebotomy (1 case).

(5)

associated with each score. The average score for potential impact was 2.9 (SD, 1.0), and the median score was 3.

Responsibility for the Incidents

The classification of the 129 incidents by responsible party and a summary of an incident from each of the 3 main categories are given in ❚Table 4❚. Of 129 incidents, 94 (73%) involved the laboratory, including 78 (60%) that involved only the laboratory and 16 (12%) that also had a component outside the authority of the laboratory.

The specific laboratory functions involved in the 94 incidents in which a laboratory problem was involved are given in ❚Table 5❚. Of the 94 incidents, 92 involved 1 labora-tory function, and 2 involved 2 functions. The most frequent functions implicated were processing (31% of 129 inci-dents), microbiology (19%), and phlebotomy (12%).

Phase of Testing

Categorization of the incidents by the general phase of laboratory testing and the specific components within each

❚Table 3❚

Potential Impact on Patient Care for 123 Incident Reports That Involved a Potential Adverse Event (PAE)* Potential No. (%)

Impact of PAEs Sample Incident

1 14 (11) Delay in a routine test for thyroglobulin in an ambulatory setting

2 20 (16) Provider relabeled a specimen on the hospital floor after realizing he had mislabeled it

3 46 (37) Physician questioned low hemoglobin and hematocrit values in the ambulatory setting because results were

inconsistent with other clinical data; specimen redrawn and hemoglobin and hematocrit values were within the reference range

4 35 (28) Turnaround time of 2 h for stat coagulation tests in the intensive care unit

5 1 (1) 3-d delay receiving positive blood culture result (Klebsiella pneumoniae) for a patient undergoing dialysis;

during the 3-d period, patient was untreated for 2 d and mistreated for 1 d

Unable to 7 (6) —

determine

*_{The table shows the frequency of occurrence of the 5 possible scores for potential impact and gives an example of a specific incident for each score. The potential impact on}

patient care was scored using a scale graded from 1 (very unlikely to adversely affect the patient’s health) to 5 (very likely to adversely affect the patient’s health). Percentages are the fraction of the 123 incidents.

❚Table 4❚

Classification of 129 Incidents by Responsible Party and Examples of Specific Incidents for the Three Major Categories Responsible Party No. (%) of Incidents Example

Laboratory only 78 (60) Processing staff failed to order a test indicated on the requisition

Outside the authority of the 27 (21) Nurse ordered a stat test on a routine requisition form rather than using the stat

laboratory only requisition form

Both laboratory and outside the 16 (12) Provider on the hospital floor used the incorrect blood tube for the test and laboratory

authority of the laboratory failed to notify the provider that the wrong tube was used

Unable to determine 8 (6) —

❚Table 5❚

Specific Laboratory Functions Involved in 94 Incident Reports in Which the Laboratory Was Responsible for the Incident* Laboratory No. (%) of

Function Incidents Sample Incident

Processing 40 (31) Specimen delayed in processing area because of inadequate staffing

Microbiology 24 (19) Technologist entered the results of antimicrobial susceptibility testing into incorrect patient’s electronic

record

Phlebotomy 16 (12) Phlebotomist forgot to draw specimen into lavender-top tube

Hematology 6 (5) Technologist phoned incorrect results to the operating room

Chemistry 6 (5) Technologist failed to notice that instrument had rejected a sample, leading to delay in completing stat

tests

Laboratory information 3 (2) Alphabetic flag (lower case “l”) in the results field can be misinterpreted as the number 1, leading to a

system misinterpretation of results

Other 0 (0) —

*_{Percentages are the proportion of the 129 incidents included in the study. Other includes coagulation, immunology, blood bank, virology, and molecular diagnostics. The}

number of incidents totals 96 because 2 incidents involved 2 functions.

(6)

phase is given in ❚Table 6❚. Of the 129 incidents, 92 (71%) involved the preanalytic phase, 23 (18%) the analytic phase, and 14 (11%) the postanalytic phase. In 2 (2%), the phase could not be determined. The most common preanalytic problems involved specimen transportation to the laboratory (16% of incidents), specimens lost or delayed inside the laboratory (13%), and data entry errors (12%). Only 1 inci-dent (1%) was caused by an instrument error.

Preventability and Cognitive vs Noncognitive Errors

The classification of 129 incidents by preventability on a scale of 1 to 5 and summaries of an incident associated with each score are given in ❚Table 7❚. The average score for preventability was 4.0 (range, 1-5), and the median was 4. Of the 129 incidents, 94 (73%) were preventable.

The results obtained when the 94 preventable inci-dents were classified as noncognitive or cognitive errors and summaries of incidents from each category are given in ❚Table 8❚. There were 69 incidents (73% of 94 preventable incidents) that involved a noncognitive error, including 64 (68%) that involved a noncognitive error only and 5 (5%) that also involved a cognitive error. There were 28 incidents (30%) that involved a cognitive error, including 23 (24%) that involved a cognitive error only and the same 5 (5%) that also involved a noncogni-tive error.

Confidence in Classification and Missing Data

The average score for confidence in the classification of the 129 incident reports was 3.7 (range, 1-5), and the median was 4. The most commonly missing data were patient care setting and patient outcomes.

Interrater Reliability

Interrater reliability was near perfect for the general phase of laboratory testing (κ= 0.87) and excellent for the type of event (ie, actual vs potential adverse event, κ= 0.79) and responsibility for the incident (κ= 0.74). By contrast, reviewer agreement about preventability was only slightly better than that expected by chance (κ= 0.10). Thus, we assessed error type only in cases unanimously judged as preventable. In these 22 cases, agreement was fair for error type (ie, cognitive vs noncognitive, κ= 0.38).

Discussion

We developed a classification system to characterize incident reports related to laboratory services and tested the system by applying it retrospectively to 129 incident reports. Our overall objectives were to develop a tool that character-izes the nature of important laboratory errors,20 _implement

the tool, and intervene to reduce the problems identified. The results suggest that the system is feasible to implement and produces results that can guide an individual laboratory’s quality improvement activities.

The specific results obtained are similar to those of other studies of hospital-based laboratories regarding the phase of laboratory testing most frequently involved in errors.9-12,15,16,19,20 _{The main theme in these studies is the}

predominance of preanalytic errors. To our knowledge, there has been 1 other study of laboratory-related incident reports,9_{and the results of that study are similar to ours. In}

that study, 133 laboratory-related incident reports from 1 hospital were analyzed. The laboratory was responsible for 66% of the incidents, and the most common problems for which the laboratory was responsible were incorrect results reported or delays in reporting results, and the most common problems outside the laboratory were wrong patient identifi-cation and requisitions lacking the test request.

For interrater reliability, we found that general features of events, such as phase of laboratory testing and authority for the incident, could be classified with high reliability, but the reliability for judgments related to error type and preventability was poor. These findings recapitulate the pattern seen in larger studies of medical error and adverse

❚Table 6❚

Classification of the 129 Incidents by General Phase of Testing (Preanalytic, Analytic, Postanalytic) and the Specific Classification Within Each General Phase*

No. (%) of Phase of Laboratory Testing Incidents

Preanalytic 92 (71)

Requisition incorrect 6 (5)

Patient injured during phlebotomy 7 (5)

Patient unhappy with phlebotomy customer service 7 (5)

Specimen or aliquot not labeled or mislabeled 13 (10)

No specimen collected 5 (4)

Incorrect tube used 4 (3)

Specimen suboptimal or ruined 8 (6)

Specimen lost or delayed in transport to laboratory 20 (16)

Specimen lost or delayed in laboratory 17 (13)

Specimen lost or delayed (not assignable) 2 (2)

Failure to order, add, or change a test request 14 (11)

Data entry error or other laboratory information 16 (12)

system problem Other 3 (2) Analytic 23 (18) Instrument error 1 (1) Human error 18 (14) Other 4 (3) Postanalytic 14 (11)

Results not reported, delay in reporting, or reported to 7 (5)

wrong provider

Incorrect results reported because of postanalytic data 5 (4)

entry error

Other 2 (2)

*_{The sum of the general items is 131, since 2 incidents involved 2 general phases.}

The sum of the specific items is 159, since many incidents were associated with multiple preanalytic problems. All percentages are based on 129 incidents.

(7)

events. For example, in the Harvard Medical Practice Study,3

physician reviewers exhibited substantial agreement in iden-tifying the presence of adverse events (κ = 0.61) but only “fair” agreement in identifying negligent care (κ= 0.24). In a recent study that focused on the issue of interrater reliability, the authors noted that “if one reviewer rated a death as defi-nitely or probably preventable, the probability that the next reviewer would rate that case as definitely not preventable (18%) was actually slightly higher than the probability that the second reviewer would agree with the first (16%).”24_The

one study to achieve even fair to modest interrater reliability for judgments related to preventability and error type required intensive training for reviewers and eliminating reviewers who consistently yielded divergent judgments.8

One reason for poor agreement among physicians about preventability is that physician reviewers often regard errors as reflecting ubiquitous quality problems, and, therefore, they perceive the errors as unpreventable.25 _{To avoid this}

problem, we supplemented the question about preventability with characterization of each incident as involving a cognitive

or noncognitive error. Physician reviewers exhibited greater agreement for error type than for the implicit judgment of preventability. In general, noncognitive errors require inter-ventions that help people avoid lapses in concentration, and cognitive errors require interventions that involve additional training or supervision. In this study, noncognitive errors occurred more frequently (73% of the preventable incidents) than cognitive errors (30% of preventable incidents). Noncognitive errors were particularly common in the prean-alytic phase of testing (Table 8 shows 7 examples of noncog-nitive errors made in the preanalytic phase of testing).

In our experience, a common mistake in clinical labora-tories is that corrective actions focus on additional training and supervision for noncognitive errors. Examples include suggesting additional training for an experienced phle-botomist who forgot to draw 1 blood tube in a series or for an experienced specimen processor who committed a data entry error at specimen login. In these cases, training is not the most effective intervention. Rather, it is more appropriate to concentrate on improving error checking and decreasing

❚Table 7❚

Classification of 129 Incidents by Their Preventability* No. (%) of

Preventability Incidents Sample Incident

1 3 (2) Patient insisted on leaving after phlebotomy despite feeling nauseous and being advised to remain

seated; fainted on departing

2 15 (12) Delayed transport of stat specimens owing to delay on dumbwaiter; dumbwaiter problems were

infrequent and unpredictable

3 9 (7) Insufficient staffing in processing led to delays in processing a stat specimen

4 32 (25) Processing technician misread an uncommonly used requisition form and entered the wrong test

request into the laboratory information system

5 53 (41) Technologist failed to call provider with critical result

Unable to 17 (13) —

determine

*_{The preventability scale used was 1, definitely not preventable, to 5, definitely preventable.}

❚Table 8❚

Classification of 94 Preventable Incidents by Psychological Category* No. (%) of

Error Category Incidents Examples

Noncognitive only 64 (68) Failure to enter information from the requisition into the laboratory information system; mislabeled

specimens; experienced laboratory staff member selected wrong specimen tube for a common test; specimen placed in the incorrect transportation bin; wrong requisition form chosen for a stat test; laboratory staff member failed to notify provider that specimen was inadequate; common test ordered on the wrong requisition from

Cognitive only 23 (24) Mistaking yeasts for host cells on a Gram stain; failure to identify an organism growing on a culture

plate; failure to arrange for proper staffing of the processing division; not knowing that critical results cannot be communicated by voice mail only (a provider must be contacted and orally accept the result); failure to implement a downtime policy for notifying hospital floors when the dumbwaiter is not functioning; laboratory staff rejecting a specimen for malaria owing to lack of knowledge about acceptable specimens

Cognitive and 5 (5) Provider mislabeled specimen (noncognitive), and laboratory staff misinterpreted existing policy and

noncognitive refused to run the specimen even though the provider was willing to take responsibility for the

result (cognitive)

*_{See the “Materials and Methods” section for definitions. Percentages are the fraction of the 94 preventable incidents.}

(8)

work interruptions, task complexity, and the volume of work. For example, decreasing the number of different types of requisitions potentially could reduce the number of errors at specimen login. Increasing staffing or improving the deploy-ment of staff can reduce the probability of work overload.

A significant limitation of our study was that many cases were missing data such as specific test information, patient care setting, and patient outcomes. Data were missing owing to incomplete incident reports, incomplete extraction of the incident report data into the research database, and incomplete investigation of the incident owing to delays in receiving incident reports. Specific examples of missing data were (1) knowing that antibiotic sensitivities were reported incorrectly for a hospitalized patient but not knowing the impact on the patient’s therapy and condition and (2) knowing the reason and duration for a delay in reporting results but not knowing the specific tests ordered or the patient impact. The missing data have a number of implica-tions for the study, most notably an underestimation of the number of adverse events.

Another limitation of the study is that incident reports, despite their widespread use, are a flawed method for detecting adverse and potential adverse events because they tend to underreport events and may not randomly sample events26_(for

review see Wald and Shojania27_{). This has led to a number of}

efforts to improve event surveillance using 1 or more methods such as active physician solicitation, information system–based

screening methods, and active chart review.27-31_{The flaws in}

incident reporting make it impossible to use the results in this study to calculate the rate of laboratory-related adverse and potential adverse events. In addition, it is possible that events characterized herein are not representative of the insti-tution that was studied.

Our future efforts will concentrate on using an enhanced version of the classification system in a prospective study of laboratory-related incident reports. This study will use pathologists and pathology residents to rapidly capture complete sets of data that include patient outcomes. In addi-tion, to minimize the problem of underreporting associated with incident reports, we are exploring other criteria for identifying laboratory-related adverse and potential adverse events. Examples of other screening criteria include reviewing cases in which corrected laboratory reports were issued in a critical care setting and reviewing stat tests asso-ciated with prolonged turnaround times. The application of improved screening criteria should lead to a more accurate determination of the rate and characteristics of adverse events.

We developed a classification tool that can be used to characterize the nature of laboratory-related adverse events. Our plans are to improve the classification system, apply it prospectively to significant events, and use the results to guide our quality improvement efforts.

❚Appendix 1❚

Four Incidents and Their Classification*

Impact

Specific Responsi- Prevent- Error

Synopsis Key Investigative Findings on Patient Potential bility ability Type Testing Phase

CSF Gram stain reported as no Inexperienced technologist Incorrect results 4 Laboratory 4 Cognitive Analytic, human organisms seen; moderate thought yeasts were sent; delay in (microbiology) error Cryptococcus organisms host cells receiving results

were present

Positive blood culture result left 3-d delay in receiving positive Delay in receiving 5 Laboratory 5 Cognitive Postanalytic, delay on physician's voice mail; blood culture result (Klebsiella test results (microbiology) in reporting laboratory policy is to phone pneumoniae) for patient results physicians and have them orally undergoing dialysis resulting

accept the result in no treatment for 2 d and mistreatment for 1 d

Physician notified laboratory that Test was handwritten clearly on Redraw; delay in 1 Laboratory 5 Noncognitive Preanalytic, failure microalbumin test ordered requisition form; processing receiving test (processing) to order, add, or was not performed technician in laboratory failed results change a test

to enter test order request

4-h turnaround time on stat Laboratory processing staff Delay in receiving 4 Laboratory 5 Noncognitive Preanalytic, data

chemistry tests for a patient entered an incorrect receipt test results (processing entry error; in intensive care unit time (2 h ahead of actual and chemistry) analytic, human

time) and technologist did not error

recognize that the chemistry instrument had rejected the specimen

CSF, cerebrospinal fluid.

*_{All incidents were identified as potential adverse events, and the confidence in the classification was scored as 4 for all incidents. See the “Materials and Methods” section for}

definitions of terms and an explanation of the scoring systems used.

(9)

From the Departments of 1_{Laboratory Medicine and}4_Family Medicine, University of Washington School of Medicine, Seattle; and 3Laboratory Medicine and 2_{Medicine, University of} California San Francisco School of Medicine.

Address reprint requests to Dr Astion: Dept of Laboratory Medicine, University of Washington, Box 357110, Seattle, WA 98195-7110.

References

1. Kohn L, Corrigan J, Donaldson M, eds. To Err Is Human: Building a Safer Health System. Washington, DC: National Academies Press; 2000.

2. Shojania KG, Duncan BW, McDonald KM, et al, eds. Making Health Care Safer: A Critical Analysis of Patient Safety Practices. Rockville, MD: Agency for Healthcare Research and Quality; 2001. Evidence Report/Technology Assessment No. 43. AHRQ publication 01-E058.

3. Brennan TA, Leape LL, Laird NM, et al. Incidence of adverse events and negligence in hospitalized patients: results of the Harvard Medical Practice Study I. N Engl J Med.

1991;324:370-376.

4. Wilson RM, Runciman WB, Gibberd RW, et al. The Quality in Australian Health Care Study. Med J Aust. 1995;163:458-471.

5. Bates DW, Cullen DJ, Laird N, et al. Incidence of adverse drug events and potential adverse drug events: implications for prevention. ADE Prevention Study Group. JAMA. 1995;274:29-34.

6. Fischer G, Fetters MD, Munro AP, et al. Adverse events in primary care identified from a risk-management database. J Fam Pract. 1997;45:40-46.

7. Kaplan HS, Battles JB, Van der Schaaf TW, et al. Identification and classification of the causes of events in transfusion medicine. Transfusion. 1998;38:1071-1081. 8. Thomas EJ, Studdert DM, Burstin HR, et al. Incidence and

types of adverse events and negligent care in Utah and Colorado. Med Care. 2000;38:261-271.

9. Goldschmidt HMJ, Lent RW. Gross errors and work flow analysis in the clinical laboratory. Klin Biochem Metab. 1995;3:131-140.

10. Stahl M, Lund ED, Brandslund I. Reasons for a laboratory’s inability to report results for requested analytic tests. Clin Chem. 1998;44:2195-2197.

11. Lapworth R, Teal TK. Laboratory blunders revisited. Ann Clin Biochem. 1994;31:78-84.

12. Chambers AM, Elder J, O’Reilly DS. The blunder-rate in a clinical biochemistry service. Ann Clin Biochem. 1986;23(pt 4):470-473.

13. Boone DJ, Steindel SD, Herron R, et al. Transfusion medicine monitoring practices. Arch Pathol Lab Med. 1995;119:999-1006. 14. Linden JV, Paul B, Dressler KP. A report of 104 transfusion

errors in New York State. Transfusion. 1992;32:601-606.

15. Plebani M, Carraro P. Mistakes in a stat laboratory: types and frequency. Clin Chem. 1997;43:1348-1351.

16. Hofgärtner WT, Tait JF. Frequency of problems during clinical molecular-genetic testing. Am J Clin Pathol. 1999;112:14-21. 17. Nutting PA, Main DS, Fischer PM, et al. Problems in

laboratory testing in primary care. JAMA. 1996;275:635-639. 18. Witte DL, VanNess SA, Angstadt DS, et al. Errors, mistakes,

blunders, outliers, or unacceptable results: how many? Clin Chem. 1997;43:1352-1356.

19. Simpson JB. A unique approach for reducing specimen labeling errors: combining marketing techniques with performance improvement. Clin Leadersh Manag Rev. 2001;15:401-405.

20. Bonini P, Plebani M, Ceriotti F, et al. Errors in laboratory medicine. Clin Chem. 2002;48:691-698.

21. Leape LL, Brennan TA, Laird N, et al. The nature of adverse events in hospitalized patients: results of the Harvard Medical Practice Study II. N Engl J Med. 1991;324:377-384.

22. Shojania KG, Wald H, Gross R. Understanding medical error and improving patient safety in the inpatient setting. Med Clin North Am. 2002;86:847-867.

23. Reason JT. Human Error. New York, NY: Cambridge University Press; 1990.

24. Hayward RA, Hofer TP. Estimating hospital deaths due to medical errors: preventability is in the eye of the reviewer. JAMA. 2001;286:415-420.

25. Weingart SN, Davis RB, Palmer RH, et al. Discrepancies between explicit and implicit review: physician and nurse assessments of complications and quality. Health Serv Res. 2002;37:483-498.

26. Cullen DJ, Bates DW, Small SD, et al. The incident reporting system does not detect adverse drug events: a problem for quality improvement. Jt Comm J Qual Improv. 1995;21:541-548.

27. Wald H, Shojania KG. Incident reporting. In: Shojania KG, Duncan BW, McDonald KM, et al, eds. Making Health Care Safer: A Critical Analysis of Patient Safety Practices. Rockville, MD: Agency for Healthcare Research and Quality; 2001:chap 4. Evidence Report/Technology Assessment No. 43. AHRQ publication 01-E058.

28. Weingart SN, Ship AN, Aronson MD. Confidential clinician-reported surveillance of adverse events among medical inpatients. J Gen Intern Med. 2000;15:470-477. 29. O’Neil A, Petersen L, Cook E, et al. Physician reporting

compared with medical record review to identify adverse medical events. Ann Intern Med. 1993;119:370-376. 30. Classen DC, Pestotnik SL, Evans RS, et al. Computerized

surveillance of adverse drug events in hospital patients. JAMA. 1991;266:2847-2851.

31. Wolff AM, Bourke J, Campbell IA, et al. Detecting and reducing hospital adverse events: outcomes of the Wimmera clinical risk management program. Med J Aust. 2001;174:621-625.