Discussion

1114237 - Jones & Bartlett Learning ©

T

© ALMAGAMI/Shutterstock

CHAPTER 10

Classification and the Reduction of Medical Errors Donna Woods

Human fallibility is like gravity, weather, and terrain, just another foreseeable hazard.

— J. T. Reason (1997, p. 25)

he publication of the Institute of Medicine’s report, To err is human, highlighted that 44,000 to 98,000

deaths occurred each year from the care patients receive for a particular condition, rather than the condition itself (IOM, 2000). In fact, updated error

1114237 - Jones & Bartlett Learning ©

incidence findings from the 2010 Office of the Inspector General’s Report on Adverse Events in Hospitalized Medicare Beneficiaries (Levinson, 2010) now estimate that 180,000 deaths occur in hospitals per year in people over 65 years of age. Without even including the remainder of the population, this made medical errors the third-leading cause of death after cardiovascular disease and cancer.

Given this considerable excess mortality and harm, one of the key foci of the field of patient safety is to understand the underlying attributes, vulnerabilities, and causes in the systems and processes of care leading to errors and harm. If the mechanisms and causes are well understood, interventions to address these can be developed in order to prevent or mitigate harm.

When errors occur in high-risk environments, such as health care (IOM, 2000), there is a greater potential for harm. Health care is very complex, and an effective system of classification should accurately represent this level of complexity. By understanding human error, we can plan for likely, and potentially predictable, error scenarios and implement barriers to prevent or mitigate the occurrence of potential errors. This chapter will discuss the definitions, theories, and existing systems of classification that are used improve our understanding of the mechanisms of errors and patient safety events in order to apply this knowledge to reduce harm. Common definitions are essential to a standardized classification system. See TABLE 10.1 for some common definitions of some important terms in the field of patient safety.

1114237 - Jones & Bartlett Learning ©

TABLE 10.1 Terms and Definitions

Terms Definitions

Error “Error is defined as the failure of a planned action to be completed as intended or the use of a wrong plan to achieve an aim.” (IOM, 2000; QuIC, 2000)

Errors can occur in both the planning and execution stages of a task. Plans can be adequate or inadequate, and actions (behavior) can be intentional or unintentional. If a plan is adequate, and the intentional action follows that plan, then the desired outcome will be achieved. If a plan is adequate, but an unintentional action does not follow the plan, then the desired outcome will not be achieved. Similarly, if a plan is inadequate, and an intentional action follows the plan, the desired outcome will again not be achieved (IOM, 2000; QuIC, 2000).

“An act of commission (doing something wrong) or omission (failing to do the right thing) leading to an undesirable outcome or significant potential for such an outcome.” This definition highlights the fact that an error while it can lead to harm does not need to lead to harm to be so assessed (IOM, 2000; QuIC, 2000).

Adverse Event

“An adverse event is defined an injury that was caused by medical management (rather than the underlying disease) and that prolonged the

1114237 - Jones & Bartlett Learning ©

hospitalization, produced a disability at the time of discharge, or both.” (Thomas et al., 2000)

Preventable Adverse Event

A preventable adverse event is defined “as an adverse event that could have been prevented using currently available knowledge and standards.” (Thomas et al., 2000)

Sentinel Event

A sentinel event is a Patient Safety Event that reaches a patient and results in any of the following:

■   Death

■   Permanent harm

■   Severe temporary harm and intervention required to sustain life

(TJC, 2017).

Never Events

National Quality Forum described 29 serious events also called Serious Reportable Events, “which are extremely rare medical errors that should never happen to a patient.” (AHRQ, 2017b)

Incident “An unplanned, undesired event that hinders completion of a task and may cause injury, illness, or property damage or some combination of all three in varying degrees from minor to catastrophic. Unplanned and undesired do not mean unable to prevent. Unplanned and undesired also do not mean unable to prepare for. Crisis planning is how we prepare for serious incidents that occur that require response for mitigation.” (Ferrante, 2017b)

1114237 - Jones & Bartlett Learning ©

Near Miss “A near miss is defined as event that could have had adverse consequences but did not and was indistinguishable from fully fledged adverse events in all but outcome. In a near miss, an error was occurred, but the patient did not experience clinical harm, either through early detection or sheer luck.” (AHRQ, 2017a)

1114237 - Jones & Bartlett Learning ©

▶ Why Classify Safety Events? A key component of improving patient safety is to conduct an assessment of safety incidents or events that occur, and through these assessments, to gain an understanding of the vulnerabilities that exist in the related systems and processes. Risk assessment is the science of risks and their probability (Haimes, 2004). This commonly involves error or event classification. A patient safety error/event classification is a method for standardizing language, definitions, and conceptual frameworks for optimal discussion, improvement, and dissemination of knowledge on patient safety (IOM, 2004; Woods et al., 2005). A classification system can be used within a reporting system or used to analyze events. Through the classification of events, trends can be assessed. The use of an effective system of classification is an important part of risk assessment and helps to group “like” (similar) risks and safety events as well as to distinguish characteristics in similar but “unlike” events for more effective and nuanced intervention development.

Event classification is a standardized method by which the event, contributors and causes of the event, including the root causes, are grouped into categories (Woods et al., 2005). Accident classification was initially used in nuclear power and aviation and has expanded into use in health care (Barach & Small,

1114237 - Jones & Bartlett Learning ©

2000). Additionally, by analyzing many events and applying the same standardized classification scheme, patterns can be detected in how these events develop and occur (Woods et al., 2008). An understanding of the different error types is critical for the development of effective error prevention and mitigation tools and strategies (Vincent, 2006). A variety of these tools and strategies must be implemented to target the full range of error types if they are to be effective. There are several different systems of classifications of error that describe differing characteristics of an error. Many of these systems of classification are built on the prevailing theoretical framework of Reason’s Swiss Cheese Model of accident occurrence (Reason, 2000).

1114237 - Jones & Bartlett Learning ©

▶ Skill-, Rule-, and Knowledge- Based Classification

James Reason posited a classification of human error (Reason, 1990). Errors result from a variety of influences however, believed the underlying mental processes that lead to error are consistent, allowing for the development of a human error typology. The terms skill-, rule-, and knowledge-based information processing refer to the degree of conscious control exercised by the individual over his or her activities. Failures of action, or unintentional actions, were classified as skill-based errors. This error type is categorized into slips of action and lapses of memory. Failures in planning are referred to as mistakes, which are categorized as rule-based mistakes and knowledge-based mistakes (Reason, 1990):

■   Skill-based errors—slips and lapses—when the action made is not what was intended.

Examples of skill-based errors in daily life and health care include: •   Daily life—a skilled driver stepping on the

accelerator instead of the brake. •   Health care—an experienced nurse administering

the wrong medication by picking up the wrong syringe.

1114237 - Jones & Bartlett Learning ©

■   Rule-based mistakes—actions that match intentions but do not achieve their intended outcome due to incorrect application of a rule or inadequacy of the plan. Examples of rule-based mistakes in daily life and in health care include: •   Daily life—using woodworker’s glue to mend a

broken plastic eyeglasses frame. •   Health care—proceeding with extubation before

the patient is able to breathe independently, because of incorrect application of guidelines.

■   Knowledge-based mistakes—actions which are intended but do not achieve the intended outcome due to knowledge deficits. Examples of knowledge- based mistakes in daily life and in health care include: •   Daily life—a failed cake because a novice baker

mistakenly thought baking soda could be used in place of baking powder.

•   Health care—prescribing the wrong medication because of incorrect knowledge of the drug of choice.

1114237 - Jones & Bartlett Learning ©

Slips, Lapses, Mistakes, and Violations Reason’s human error typology was further elaborated to describe slips as attentional failures and lapses as memory failures as depicted in FIGURE 10.1 (Reason, 1990). Violations were added as intended actions with further break-down into common, routine accepted violations (e.g., work-arounds, short cuts, potentially risky actions, movement into the margin of error) and exceptional violations that are not accepted (i.e., alcohol or drug use while providing patient care).

FIGURE 10.1 Generic Error Modeling System (GEMS)

Reproduced from Reason, J. Human Error. Cambridge University Press, Cambridge, UK. 1990.

1114237 - Jones & Bartlett Learning ©

Planning and Execution Error types can be further classified into whether the error involved taking an action or inaction. If a needed action is omitted, it is considered an error of omission. For active errors, there is a further breakdown into whether the error was the use of a wrong plan—an error of planning; or whether the plan was good, but the error was in the execution of the plan—an error of execution. An example of an error of planning would be a decision to order a medication that was wrong for a patient’s condition (e.g., an antiviral medication ordered for a person with a bacterial infection). An example of an error of execution would be administering a medication to one patient that was intended for a different patient.

Furthermore, these elements of error classification can interact with the classification of slips, lapse, mistakes, and violations described above and as seen in Figure 10.1.

Runciman and colleagues conducted a review of adverse events in a random selection of admissions in Australia and developed a classification system called the Quality in Australia Health Care Study (QAHCS) (Wilson et al., 1995; Runciman et al., 2009). Risk classification has been an important tool for risk reduction and safety improvement in high risk high reliability industries such as nuclear power, aviation, chemical manufacturing and others. Valid and reliable classifications of risk and details of hazards and risks is a very important tool for exploring risk.

Reason hypothesized that most accidents can be traced to one or more of four failure domains: organizational

1114237 - Jones & Bartlett Learning ©

influences, supervision, preconditions, and specific acts. For example, in aviation, preconditions for unsafe acts include fatigued air crew or improper communications practices. Unsafe supervision encompasses for example, pairing inexperienced pilots on a night flight into known adverse weather. Organizational influences encompass such things as organizational budget cuts in times of financial austerity.

1114237 - Jones & Bartlett Learning ©

The Swiss Cheese Model James Reason (1990) proposed the Swiss Cheese Model as a theory of accident occurrence, which has become the dominant paradigm of accident occurrence in health care. Reason’s “Swiss Cheese” model describes four levels within which active failures and latent failures may occur during complex operations.

In the Swiss Cheese Model, an organization’s defenses against failure are modeled as a series of barriers, represented by slices of cheese. The holes in the cheese slices represent weaknesses in individual parts of the system and are continually varying in size and position across the slices. The system produces failures when a hole in each slice momentarily aligns, permitting (in Reason’s words) “a trajectory of accident opportunity,” so that a hazard passes through holes in all of the slices, leading to a failure (Reason, 1990). The Swiss Cheese model, as Reason conceived of it, includes both active and latent failures. Active failures encompass the unsafe acts that can be directly linked to an accident, such as human error. This is not about blame for the action, it is intended to depict the actions taken most proximal to the event whether a violation or an error. Latent failures include contributory factors that may lie dormant for days, weeks, or months until they contribute to the accident. Latent failures span the first three domains of failure in Reason’s (1990) model (FIGURE 10.2).

1114237 - Jones & Bartlett Learning ©

FIGURE 10.2 Reason’s Swiss Cheese Model

Reproduced from Reason, J. Human Error. Cambridge University

Press, Cambridge, UK. 1990.

1114237 - Jones & Bartlett Learning ©

Human Factors Analysis and Classification System (HFACS) In the early 2000s, Wiegmann and Shappell were tasked by the U.S. Navy with developing a classification of risks to reduce the high rate of error. They developed the Human Factors Analysis and Classification System (HFACS). HFACS was designed as a broad human error framework heavily based on Reason’s Swiss Cheese Model and was used initially by the U.S. Navy to identify active and latent failures that combine to result in an accident with the goal of understanding the causal factors that led to the accident (Weigmann & Shappell, 2000, 2003).

The HFACS framework describes human error at each of four levels of failure (FIGURE 10.3):

1114237 - Jones & Bartlett Learning ©

FIGURE 10.3 Depiction of HFACS Classification System

Reproduced from Patterson J, Shappell SA. Operator error and

system deficiencies: Analysis of 508 mining incidents and accidents

from Queensland, Australia using HFACS. Accident Analysis and

Prevention. 42 (2010) 1379–1385.

1. Unsafe acts of operators (e.g., aircrew) 2. Preconditions for unsafe acts 3. Unsafe supervision 4. Organizational influences

1114237 - Jones & Bartlett Learning ©

Within each HFACS level, causal categories were developed that identify the active and latent failures that occur. In theory, at least one failure will occur at each level leading to an adverse event. If one of the failures is corrected at any time leading up to the adverse event, it is thought that the adverse event would be prevented.

As can be seen in Figure 10.3, the HFACS classification further defines an incident in relation to Reason’s categories of Errors and Violations, with skill-based errors, decision errors, and perceptual errors under the Errors classification and routine violations and exceptional violations under the Violations classification. Likewise, further defining each of the four primary classifications. Using the HFACS framework for accident investigation, organizations are able to identify the breakdowns within the entire system that allowed an accident to occur and provide insight for the development of potential interventions for improvement.

While the HFACS classification system was developed by the U.S. Navy, the HFACS framework has been used to assess root causes within the military, commercial, and general aviation sectors to systematically examine underlying human causal factors and to improve aviation accident investigations as well as in many other high-risk industries such as rail, construction, and mining and has demonstrated acceptable to high reliability. Studies have shown that the results from the HFACS classification system are reliable in capturing the nature of, and relationships among, latent conditions and active failures. The kappa statistic is reported in these studies to assess agreement (kappa is a standard

1114237 - Jones & Bartlett Learning ©

statistical measure of inter-rater agreement that ranges from 0.00–1.00). In one study, a post-hoc analysis was conducted on a sample National Transportation Safety Board (NTSB) serious incident reports and achieved an inter-rater reliability kappa of 0.85. In another study in the biopharmaceutical industry, three pairs of coders used the HFACS classification system to code a sample of 161 reports of “upstream” incidents, “downstream” incidents, and incidents in “operational services.” In this study, they achieved an overall agreement of 96.66 with a 0.66 kappa, suggesting that this classification system has reasonable reliability for use (Cintron, 2015).

1114237 - Jones & Bartlett Learning ©

The Joint Commission Patient Safety Event Taxonomy Improvement in patient safety involves risk assessment to understand the attributes of a safety incident. The Joint Commission (TJC) encouraged hospitals to conduct risk assessments on events that occurred in hospitals, particularly serious adverse events or “sentinel events” to reduce harm. To support and facilitate the conduct of hospital safety risk assessments, in 2005 TJC developed the Patient Safety Event Taxonomy to put forward a standardized terminology and classification schema for near miss and adverse events to develop a common language and definitions (Chang et al., 2005). This taxonomy was developed through application of the results from a systematic literature review, evaluation of existing patient safety terminologies and classifications, assessment of the taxonomy’s face and content validity, patient safety expert review and comment, and assessment of the taxonomy’s comparative reliability. Five root nodes were proposed based on the review of other classification schemes. The root nodes include: impact, type, domain, cause, and prevention and mitigation (Chang et al., 2005).

The taxonomy was applied qualitatively to reports submitted to TJC’s Sentinel Event Program. According to the developers, the taxonomy incorporated works by Reason (1990), Rasmussen (1986), Runciman (1998, 2002), and Hale (1997) as well as contributions from aviation and other high-risk/high-reliability industries. Root nodes were divided into 21 subclassifications,

1114237 - Jones & Bartlett Learning ©

which were then subdivided into more than 200 coded categories and an indefinite number of noncoded text fields to capture narrative information about specific incidents.

1114237 - Jones & Bartlett Learning ©

World Health Organization International Classification for Patient Safety (ICPS) The Patient Safety Event Taxonomy was disseminated internationally through TJC International and then in response to the growth in the recognition of the magnitude of excess morbidity and mortality from patient safety, the World Health Organization (WHO) launched the World Alliance for Patient Safety to “‘pay the closest possible attention to the problem of patient safety and to establish and strengthen science-based systems necessary for improving patients’ safety and quality of care” (WHO, 2002). Concurrent with the development of TJC (then JACHO) Patient Safety Event Taxonomy, since 2005, WHO was tasked with developing global norms and standards to support and inform the development of effective patient policies and practices. Runciman et al. (2009) drafted the International Classification for Patient Safety (ICPS), drawing on many sources and terms including TJC Patient Safety Taxonomy. The accumulated terms were analyzed and assessed through responses to a Delphi process, with the focus on patient safety classification (Thomson, 2009). The conceptual framework for the ICPS, consisting of 10 high level classes (FIGURE 10.4) (WHO, 2009):

1114237 - Jones & Bartlett Learning ©

FIGURE 10.4 International Classification for Patient Safety Framework

Reproduced from The World Alliance For Patient Safety Drafting

Group, Sherman H, Castro G, Martin Fletcher M on behalf of The

World Alliance for Patient Safety, Hatlie M, Hibbert P, Jakob R, Koss R,

Lewalle P, Loeb J, Perneger T, Runciman W, Thomson R, Van Der

Schaaf T, Virtanen1 M. Towards an International Classification for

Patient Safety: the conceptual framework. Int J Qual Health Care.

2009 Feb; 21(1): 2–8

1. Incident type 2. Patient outcomes 3. Patient characteristics

1114237 - Jones & Bartlett Learning ©

4. Incident characteristics 5. Contributing factors/hazards �. Organizational outcomes 7. Detection �. Mitigating factors 9. Ameliorating actions

10. Actions taken to reduce risk

According to the conceptual framework for the ICPS, “A patient safety incident in this context is an event or circumstance that could have resulted, or did result, in unnecessary harm to a patient. A patient safety incident can be a near miss, a no harm incident or a harmful incident (adverse event). The class, incident type, is a descriptive term for a category made up of incidents of a common nature grouped because of shared, agreed features, such as “clinical process/procedure” or “medication/IV fluid” incident. Although each concept is clearly defined and distinct from other concepts, a patient safety incident can be classified as more than one incident type. A patient outcome is the impact upon a patient, which is wholly or partially attributable to an incident. Patient outcomes can be classified according to the type of harm, the degree of harm and any social and/or economic impact. Together, the classes incident type and patient outcomes are intended to group patient safety incidents into clinically meaningful categories” (The World Alliance for Patient Safety Drafting Group, 2009).

In an NIH-funded research study, the author of this chapter tried to classify the Incident Types and the Contributing Factors through application of the WHO

1114237 - Jones & Bartlett Learning ©

ICPS for clinician described incidents that occurred during the course of surgical care, which included operating room set-up, the surgery itself, and anesthesia management. We developed a web-based confidential Debriefing Tool, which would proactively collect issues, challenges, problems, incidents, near misses, and preventable adverse events from everyone involved in specific surgeries. The web-based tool is emailed to all of those involved in the procedure immediately following the procedure with 2 reminders within 3 days if needed. Both open-ended questions and specific prompts are used. We sought to develop a method which would overcome many of the barriers to incident reporting in a surgical context, including developing a method that would be perceived as safe for all involved, to report anything that occurred, large or small, without fear of personal reprisal or reprisals for others on the team. The web-based Debriefing Tool was deployed across four large transplant centers. See examples of reported incidents and the application, function, and results of the use of the WHO ICPS system of classification to classify emblematic incidents reported from the surgical care teams below.

Unlike the HFACS which can model the complexity of multiple causal inputs to an accident, as represented by the Swiss Cheese Model of accident causation, the WHO ICPS is represented as a linear hierarchical system of classification that requires reviewers to select one sole Incident Type and one sole Contributing Factor. Within the Incident Type classification, there are 13 primary Incident Type classifications. However, the

1114237 - Jones & Bartlett Learning ©

requirement to select only one classification created forced choices. For example, for an incident in which the consent was not available, the primary Incident Type classification could be either or both Clinical Administration and/or Documentation, where both could actually be true error types related to this incident. Furthermore, each Primary Incident Type has different downstream classification options which could stunt resulting understanding of the event. Selecting one or the other will lead the classification down very different paths where the secondary classification would be very different. See TABLE 10.2 for a representation of this challenge.

1114237 - Jones & Bartlett Learning ©

TABLE 10.2 Depiction of WHO Classification Challenges

1114237 - Jones & Bartlett Learning ©

For example, for Documentation, the secondary classification choices would be, Document Involved— Forms/Certificates, and the Problem classification could be Document Missing or Delay in Accessing Document as compared with Clinical Administration where the secondary classification of the Process would be Consent and the Problem could be either Not Performed When Indicated and/or Incomplete/Inadequate. When a classification system allows only exclusive classification where the classification of the same event could be accurately classified as involving multiple codes, information in the classification results of the incident will be lost and as well the results will likely lack strong reliability and accuracy which can confuse the development of interventions to improve care and can lead to loss of important information. Likewise, the primary incident type for ABO Incompatibility can be classified as Documentation, Clinical Administration, or Blood Products; Incorrect counts or a lost needle can be classified as a Clinical Process/Procedure or as Medical/Device, Equipment/Property. In our study, we developed a more specific set of definitions that created a forced choice selection of one of the classification pathways for similar events, but this was somewhat arbitrary, lacked precision and lost potentially important information about the incident. The linear nature of the classification patterns limited the classification of incidents, as it did not accurately reflect the reality of the emergence of the incidents (McElroy, 2016; Khorzad, 2015; Ross, 2011)

1114237 - Jones & Bartlett Learning ©

Similarly, in the classification of contributing factors and hazards, an incident reviewer must choose between the included factors, even though frequently many of the listed factors can contribute to the occurrence of an incident (McElroy, 2016; Khorzad, 2015; Ross, 2011). The listed primary classifications for contributing factors in the ICPS are:

1. Staff Factors 2. Patient Factors 3. Work Environment Factors 4. Organizational Service Factors 5. External Factors �. Other

Assigning a single contributing factor in many cases oversimplifies the classification of an incident and is not an accurate representation of the complexity of the interacting components of these types of events. In many safety incidents, patient factors are interacting with staff factors in the context of work environment factors, and organizational factors. There can be plural factors related to each component. The requirement in the ICPS to select one Contributing Factor also sets up a false choice for the individual conducting the classification, reliability is compromising reliability, and introducing subjective bias to the process of classification.

1114237 - Jones & Bartlett Learning ©

Systems Engineering Initiative for Patient Safety (SEIPS) Vincent, a leading thought leader in the field of human error and safety, described seven elements that influence safety (Vincent, Taylor-Adams, & Stanhope, 1998; Vincent, 2006):

1. Organization and management factors 2. Work environment factors 3. Team factors 4. Task factors 5. Individual factors �. Patient characteristics 7. External environment factors

Building upon Vincent’s model of human error and safety related elements, Carayon and colleagues proposed a Systems Engineering Initiative for Patient Safety (SEIPS) model for human error and design in health care (Carayon et al., 2006, 2014; Carayon & Wood, 2010). In the SEIPS classification model, components of the system are helpfully depicted with intersecting arrows that illustrate how these components in the system can interact with one another, representing the emergent properties of simultaneously interacting risks depicted through the Swiss Cheese Model. In the SEIPS 2.0 Framework, shown in FIGURE 10.5, accidents arise from the interaction among humans, machines, and the environment in the sociotechnical system. Similar to Vincent’s seven elements that influence safety, in the SEIPS Model 2.0 (Holden, 2013) the five interacting elements within the work system include:

1114237 - Jones & Bartlett Learning ©

FIGURE 10.5 Systems Engineering Initiative for Patient Safety (SEIPS) Model 2.0

Reproduced from Carayon P, Hundt AS, Karsh B, et al. Work system

design for patient safety: The SEIPS model. Qual Safe Health Care.

2006;15(Suppl I):i50–i58.

■   People—patient and clinical professionals ■   Tasks ■   Technology and tools ■   The organization ■   External environment

This framework is able to represent the dynamic work system and work system risks to effectively model the complexity of medical work. This framework and depiction facilitates the ability to understand the characteristics and specific patterns of the simultaneous interactions of activities and systems leading to risk and an incident. This depiction enables

1114237 - Jones & Bartlett Learning ©

the representation of both safety and risk as dynamic, not static, and represents an incident as having not just one input but multiple inputs to creates a pattern, a perfect storm, that is frequently hard to foresee prospectively. This form of classification is then represented as patterns of interactions of the components in the system and can then provide a fuller, more complete characterization of the incident. Johnson, Haskell, and Barach (2016) portray actual patient safety cases to provide an exercise using this framework for readers to conduct an analysis of the Lewis Blackwell case.

Carayon and colleagues (2006, 2014) apply this kind of incident classification and analyses and portrayed this in what they call the configural work systems concepts and configural diagrams, which provide a more nuanced and more accurate representation of the incident. This method for incident analysis and classification enables modeling of both proximal factors (sharp end) as well as latent factors (blunt end) and can provide insights regarding how they interact as well as the specific impact.

By using the configural work system diagram for incident analysis, it is possible to investigate the types of incidents that occurred when a particular work system factor (e.g., excessive workload) or combination of factors (e.g., workload and worker fatigue) were an active part of the configuration. The diagram can also be used to analyze and compare how two or more units or organizations have configured their work system, by design or otherwise, for the same process or processes.

1114237 - Jones & Bartlett Learning ©

Ideal types of configurations could be identified for improved safety and resilience and safety planning. For example, as Holden (2013) notes, “If we introduce technology A versus technology B, which new interactions will become relevant between each of those technologies and the work system’s people, task, other tool and technology, organization, internal environment and external environment factors?”

1114237 - Jones & Bartlett Learning ©

Using Error Classification for Mitigation and Safety Improvement

1114237 - Jones & Bartlett Learning ©

High Reliability Organizations A high reliability organization (HRO) is an organization that has succeeded in avoiding catastrophes in an environment where normal accidents can be expected due to risk factors and complexity. In studying these organizations, Weick and Sutcliffe (2015) found several common features across these industries that they called HRO principles:

■   Preoccupation with Failure ■   Reluctance to simplify ■   Sensitivity to operations ■   Commitment to resilience ■   Deference to expertise

These principles are very important to safety risk reduction and mitigation and apply well to the task of error classification. HROs do not ignore failures no matter how small as well as how things could fail. Error reporting systems are designed to collect this type of information (adverse events, preventable adverse events, near-misses, etc.) for classification and analysis. In the analysis of these events, it is important not to overly simplify what has occurred and to represent the complexity of interactions within the system that lead to an event not only from the perspective of the clinician but also the perspective of the patients and their families (Travaglia & Robertson, 2013). In the recommendation to be sensitive to operations is the encouragement to understand real processes and their inputs as they actually occur as opposed to relying on an idealized process or operation that may not include various work-arounds. This is also important for error

1114237 - Jones & Bartlett Learning ©

classification and risk mitigation. Through a commitment to resilience there is also attention to the potential of unintended consequences even while applying interventions to improve care based on an analysis of events. Deference to expertise rather than authority in the context of classification means that the person with the most information about an incident or a process to provide relevant contextual knowledge.

In order to effectively improve the safety of processes in health care and reduce harm, it is necessary to accurately represent the complexity of the incident, various contributors and contextual circumstances. HRO Principles are very helpful for this. Establishing a culture of safety with these principles will increase the number of events reported and the use of these events to inform safety improvement (Sollecito & Johnson, 2013).

The SEIPS framework is effective at representing the appropriate complexity, contributors, and inputs in the occurrence of an incident or event such that an intervention can address the actual problems. There have been numerous studies in which the SEIPS framework was used to improve health care design and safety, for example, a study of the timeliness of follow- up of abnormal test results in outpatient settings (Singh et al., 2009); a study on the safety of the Electronic Health Record (EHR) technology (Sittig & Singh, 2009); a study improving electronic communication and alerts (Hysong et al., 2009); characterizing patient safety hazards in cardia surgery (Gurses et al., 2012); and a study to improve patient safety for radiotherapy (Rivera & Karsh, 2008). The SEIPS framework has been used

1114237 - Jones & Bartlett Learning ©

effectively to examine patient safety in multiple care settings (Intensive Care Units, pediatrics, primary care, outpatient surgery, cardiac surgery, and transitions of care). Many studies that applied the SEIPS framework not only to demonstrate its effectiveness for understanding error and incident causation but also for how to apply this learning to safety improvement and safety planning.

1114237 - Jones & Bartlett Learning ©

▶ Conclusions Safety events are very frequent in health care. In this chapter, several systems of classification in use have been described. Some are descriptive and others a based on the psychology of error, and the strengths and weakness of each have been presented. When the nature of the system of classification can represent the varying patterns of incident occurrence, it can be used as an important and effective tool for risk reduction and safety improvement. These are frameworks of event classification that can be used in Incident Reporting Systems and to provide a framework for analysis of incidents capture trends and inform quality and safety improvement. The value of event classification has been demonstrated by numerous other high risk, high reliability industries as well as in health care, that systems of error and risk classification are important to understanding patient safety risks in health care to improve safety (Weigmann & Shappell, 2003; Carayon et al., 2014; Busse & Johnson, 1999). We further understand that High Reliability Organizations apply the practice of “Reluctance to Simplify Operations” (Weick & Sutcliffe, 2007). However, if the system for classification does not accurately describe the nature and patterns of medical work and the resulting specific risks or over simplifies the understanding of the interactions of elements of medical work or risks in the

1114237 - Jones & Bartlett Learning ©

work system, important information will be lost and attempts at using the resulting information for safety improvement will be flawed. In this review of systems of error and incident classification, the HFACS and the SEIPS frameworks, both of which rely heavily on the theoretical underpinnings of Reason’s work, have effectively enabled classification systems that model the complexity of health care and error occurrence and have demonstrated their use for effective risk identification and safety improvement in health care.