The “process” of anesthesia

Understanding human error requires us to learn about system vulnerabilities, resource constraints, pressures, and contradictions and understand how processes fail or succeed. There are several elements that make healthcare, in particular, anesthesia, susceptible to, and the people working within these areas prone to, error. Anesthesia requires the administration of rapidly acting drugs with significant physiological effects, the monitoring of patients in a vulnerable physiological state using advanced electronics, and the performance of highly technical procedures in a dynamic, unpredictable environment characterized by intense time pressure and high stakes. This situation may be further compounded by a concurrent teaching requirement. Gaba [15] and Reason [16] have discussed these factors at length, and their findings are summarized in Box 3.2.

Models for the development of safety incidents in high‐risk, complex industrial activities such as nuclear power have been adapted to model anesthetic processes. “Normal Accident Theory,” as described by Perrow in 1984, outlines two system attributes that commonly contribute to accidents: (1) the complexity of interactions between system components and (2) the tightness of coupling between these components [17].

A task or process is said to have complex interactions if there are many alternative and interrelated subtasks at any point in its completion. The more complex a system’s interactions, the more likely it is that something will go wrong as the consequence of interactions cannot always be predicted. Gaba et al. subsequently expanded on Perrow’s theory to outline how three different types of complexity (intrinsic, uncertainty, and proliferation) contribute to incidents in anesthesia [18].

Intrinsic complexity occurs due to the interconnected action of a multitude of individual agents within a system or process. The body can be considered an intrinsically complex system; huge numbers of cells and organs interact in a multitude of intricate ways, using a vast array of signals. In such a system, altering one component can have a knock‐on effect on other components, with dynamic and unpredictable consequences.

Uncertainty complexity occurs when cause and effect relationships are not always clear and are difficult to predict. Anesthesia is easily achieved by drug administration; however, the pharmacological, physiological, and pathophysiological interactions that occur are often poorly understood. In addition, devices used for measurement and monitoring the effect of these interactions can be imprecise and inaccurate, often have low signal‐to‐noise ratios, and are prone to error. Prediction of drug effects is based upon our knowledge of what a drug does in the general population, but the effect on a given individual cannot be predicted with certainty. Nor are we able to measure all affected parameters or predict all potential consequences; we only ever have a partial picture of what is happening.

Finally, if a process involves many components and tasks, this can cause proliferation complexity. Although a process and individual steps within it may be simple, the components and tasks may be interconnected in a complex fashion. For example, they may have to be performed in a precise order or at specific times for the process to succeed. Such proliferation is what can lead to lapses such as failing to open an APL valve in a breathing system following a machine leak check. The task is simple, but, because it is one task of many which need to be performed before anesthesia, it is easily forgotten. As more and more components are added to a system, the chance of one of them being missed, or failing, increases. Indeed, anesthetists typically add proliferation complexity to a case to help combat uncertainty; using multiple electronic monitoring devices, each of which has a risk of error, imprecision, and failure [18].

Coupling refers to the way components in a system are linked. In tightly coupled systems, the change in state or function of one component directly affects other components within a short period of time. Tightly coupled systems result in more accidents because even minor errors can become problematic before they can be corrected. In loosely coupled systems, there is redundancy or temporal lag built in the system allowing components to function normally despite changes in other components. The interaction of numerous homeostatic mechanisms means that physiology is inherently loosely coupled. Body systems have intrinsic buffers, which allow normal functioning in a variety of conditions; for example, renal blood flow and glomerular filtration rate are kept stable through a range of perfusion pressures. However, the redundancy and buffer mechanisms responsible for this loose coupling are hindered by the anesthetic process making components more tightly coupled. This has the effect of transferring responsibility for monitoring and maintaining an adequate state of health to the anesthetist, aided by technology and pharmaceuticals.

Reason’s Swiss cheese model

The psychologist James Reason considers complex systems to contain fundamental elements that must work together harmoniously if efficient and safe operations are to occur [14]. This requires certain preconditions to be met, such as well‐maintained and functional equipment, and individuals with appropriate training, skills, and experience. Hazards are prevented from causing harm by a series of barriers, safeguards, and defenses. Some rely on people, while others are engineered into the system (e.g., alarms). In an ideal world, each layer would be intact; however, in reality, each has unintended weaknesses. These failings degrade the integrity of the system leading to vulnerabilities within different layers like holes in a slice of Swiss cheese. Unlike in Swiss cheese, these holes are continually opening, shutting, and shifting their location as conditions in the system change. The presence of holes in any single “slice” does not normally cause an incident or bad outcome. However, when the holes within these layers momentarily align, they allow an error to proceed through the layers of a system to reach the patient (Fig. 3.2).

Human factors

Human Factors (HF), sometimes referred to as “ergonomics,” is a scientific discipline concerned with understanding interactions between people and the other elements of a system. Applying HF approaches in healthcare can enhance the way it is delivered and received. HF views people as a vital component of any system and that their abilities and limitations must be considered when attempting to improve system performance. The basic principle of HF is to analyze tasks, systems, or processes to understand the limitations put upon the humans within them. In essence, this can be achieved through asking: Can this team/person reliably and safely perform these tasks, with this training and information, using this equipment, in this environment, with these constraints and pressures, to the required standard? Through understanding human limitations and how humans process information, workplaces, tasks, and equipment can be designed and engineered to allow for variability in human performance, making it easier for frontline workers to do their job safely and effectively.

Non‐technical skills

One area that the HF approach has highlighted is the importance of non‐technical skills. These are skills individuals use to perform their work that lie outside of the traditional medical competencies of knowledge and technical skills. Non‐technical skills are cognitive, social, and personal management skills necessary for an individual to perform a role, task, or job safely and effectively [20]. They include situation awareness, decision‐making, task management, teamwork, leadership, communication, and the management of stress and fatigue.

Until recently, little attention has been given to these competencies in medical professions. One of the early proponents of non‐technical skills in anesthesiology were Howard et al., who developed an “Anesthesia Crisis Resource Management (ACRM)” training course with the aim of introducing anesthesiologists to principles of dynamic decision‐making, human performance, and the development of countermeasures aimed at combating and reducing error [21]. More recently, the “Anaesthetists’ Non‐Technical Skills (ANTS)” behavioral marker system was developed following collaboration between anesthesiologists and psychologists [20]. Task analysis was used to define the critical non‐technical skills required for a safely functioning anesthetist, and an assessment scale was developed for ANTS training and observations in operating rooms or simulation settings (Table 3.1). Such training and observation rarely occur in the veterinary sector, but it is likely that non‐technical skills are similarly important for veterinary anesthesia.

Blame and no‐blame cultures

A blame culture is one where incidents are blamed on individuals making errors despite those individuals having little or no control over the conditions in which the error occurred [22]. It leads to norms and attitudes characterized by an unwillingness to take risks or accept responsibility for mistakes because of a fear of criticism or repercussions. Blame culture cultivates distrust and fear, with people tending to blame each other to avoid being blamed themselves. This can result in few new ideas and a lack of personal initiative because people do not want to risk being wrong. Safety incidents tend to stay unreported, and investigations tend to be brief, finding culpability in frontline workers. Such cultures tend to evolve in hierarchical, rule‐, and compliance‐based systems but where the work is highly variable, such as healthcare.

A no‐blame culture is in essence the opposite: a culture in which individuals are not held accountable for their actions. This should encourage people to report errors and provide an environment where innovation is encouraged [23]. However, a no‐blame culture is neither feasible nor desirable. Most people desire some level of accountability when a safety incident occurs, and individuals should take ownership of their decisions and actions and be accountable for being part of any solution [23].

Just culture

A just culture is one in which individuals are not punished for human errors if their actions, decisions, or omissions were appropriate to their experience and training, but where negligence, willful violations, and destructive acts are not tolerated [24]. Unlike a no‐blame culture, there is accountability; individuals are accountable for reporting incidents and organizations are accountable for implementing appropriate corrective actions to improve the system and reduce the risk of recurrence of incidents [23].

If staff members perceive that their reports are treated fairly and lead to positive change, and that people who willfully violate safety rules and take unnecessary risks are held accountable, the willingness to report will increase. This encourages reporting and investigation of incidents and improves safety.

Learning culture

Learning culture is concerned with the sustainability of learning from failure through the reporting and analysis of errors and incidents and the implementation of systems level interventions. It requires an atmosphere of psychological safety, that is, a supportive work environment in which team members believe that they can question existing practices, express their concerns, or dissent, and admit mistakes without suffering ridicule or punishment [25]. It requires open communication between all levels of staff within an organization, a flattening of hierarchy, and transparency in incident management.

Senge [26] has described five disciplines which together make a learning culture: (1) self‐mastery, (2) shared mental models, (3) shared vision, (4) team learning, and (5) systems thinking [27]. Self‐mastery involves realistic thinking about one’s abilities. Shared mental models require team members to have a common understanding of the task being performed and of the involved teamwork, whereas a shared vision means all team members are working toward the same goal. Team learning represents a situation where all team members are continuously learning from each other and their successes and failures. Systems thinking, on the other hand, is an attempt to understand the way the system works and how this influences the behavior and performance of individuals working within it.

Safety I versus Safety II

Modern views of safety consider that people bring the innovation, creativity, flexibility, and resilience required for processes to succeed in systems with huge variation [28]. People can adjust what they do to match the conditions in which they work [28]. It can be argued that everyday performance variability provides the adaptations that are needed to respond to varying conditions and is the reason why systems are generally successful. Humans can consequently be seen as a resource necessary for system flexibility and resilience rather than a cause of the problem. Safety I can be viewed as the approach to safety, which concentrates on avoiding things going wrong. Safety II focuses on how things go right under varying conditions, exploring all possible outcomes to understand how flexibility in a system allows success. This concept is still in its infancy within medicine; however, it may give a better overall view of performance within the healthcare system [29].

Assessing organizational safety culture

Assessing safety culture can reveal potential issues in communication, teamwork, resources, and management strategies, delineating areas for targeted improvement efforts. Several validated measurement scales have been developed to assess safety culture in human healthcare organizations. One of these, the “Safety Attitudes Questionnaire” [30], has been modified for use in veterinary medicine. The resultant “Nottingham Veterinary Safety Culture Survey” was developed for use in veterinary practices within the United Kingdom and was subsequently further adapted for use in the US [31,32].

Incident reporting systems

One of the main sources of information about patient safety has come from the systematic collection and analysis of incident reports. Moreover, incident reporting is central to the development of a learning culture, wherein errors are discussed, and systems analysis is applied to improve outcomes. There is also considerable value in reporting near misses and no harm incidents since they are harbingers of potential patient harm. To capture a range of different perspectives and build a more complete picture of an incident, information should be collected from all involved. Often, it is more junior team members (e.g., interns), animal health technicians, and support staff who are most exposed to weaknesses and flaws within a system.

Early reporting is important. As time passes after an incident, memories fade, bias creeps in, and an incident may never be reported as caregivers move on to other tasks. In the view of the authors, reporting should be completed within 12–48 h of an incident occurring, with harmful incidents reported toward the shorter end of this range (by the end of the workday is a good rule of thumb).

Information collected on incidents should focus on creating an account of events, describing the “5 Ws”: “Who” (was involved), “What” (happened), “When” (the incident happened), “Where” (it happened), and “Why” (and how it happened).

Existing reporting systems in human medicine are effectively voluntary, with perhaps the exception of harmful incidents when there is a stronger expectation of reporting. While voluntary reporting systems have their drawbacks, they can be highly effective when properly designed and administered. Importantly, the success of any system is predicated upon its acceptance and use by caregivers. Several concepts to voluntary reporting system design facilitate their adoption and usefulness for incident analysis. These are confidentiality, ease of use, accessibility, independent analysis, and release of findings [33,34].

Confidentiality means individuals should be able to report incidents without the risk of personal or professional consequences [35]. Confidentiality should not be confused with anonymity. While anonymity confers further protection to those reporting, it removes the ability to follow up a report with the individual involved. To fully understand a reported incident, particularly when attempting to collate multiple reports of the same incident or simply to clarify circumstances, some degree of follow‐up is often necessary. Also, under the protection of anonymity, there may be the tendency to use reporting systems as a means of personal attack or to vent frustrations with aspects of the system (e.g., supervisors and administration) drawing focus away from a more complete report of the incident.

Reporting systems must be easy to use, including clear language, use of checkboxes for some data entry, automatic field population, minimal mandatory fields, and a section for a narrative description of events. The latter is important in gathering an account of the incident [36]. The use of open questions encourages descriptions from the reporter’s perspective.

Reporting systems must be easily accessible to all healthcare workers. Online reporting systems should be easy to find and preferably accessed through a “single click,” from computers in and out of the hospital network. There remains an argument for offering a paper‐based system in combination with a secure drop‐off box for staff members that do not have private access to computers.

To reduce bias, reports should be analyzed by independent investigators not involved in the incident. Releasing the findings of investigations, as well as a general report of types and outcomes of incidents reported, “closes the loop,” building faith in the system. This demonstrates that reports are taken seriously and acted upon and allows caregivers to be involved in any changes introduced by providing a period for comment. When an incident reveals significant system weaknesses that are likely to cause further incidents, it is critical that this information can be shared quickly and widely using multiple means (e.g., team meetings, email alerts, website notifications, and notice boards).

Voluntary reporting systems are limited by the quality of material collected and the extent to which they are representative of all incidents. Material collected is largely determined by the design of the reporting system. It is widely recognized that voluntary reports underestimate the true incidence of safety incidents [13]. However, if users understand the role and importance of submitting reports, the types of incidents to report, in combination with a reporting system that is easy to access and use, and can do so without fear of repercussion, uptake is likely to be improved. Education can be provided (e.g., as part of orientation for new employees) and reinforced within the system (e.g., regularly releasing findings emphasizes the value of reporting). There are a huge number of incident reporting systems in medicine with one of the first being the “Australian Incident Monitoring System (AIMS)” [36]. There are far fewer incident reporting systems in veterinary medicine, most being developed using digital forms and cloud‐based software [11,37].

Interviews

A senior member of the clinical team who was not involved with the incident may interview individuals involved in an incident. The primary advantage to one‐on‐one interviews is the opportunity to elicit responses without the potential for pressure from discussing an incident in a group environment. However, this advantage is countered by several important disadvantages including the inherent power dynamic of the interview process and the potential to create significant bias through the format and scope of the questioning. The use of structured or semi‐structured interview techniques, such as critical incident technique, and the use of open, non‐accusatory language can help reduce this bias.

Morbidity and mortality conferences

Morbidity and mortality conferences (M&MCs) can promote learning from incidents and drive improvements in patient safety [38]. However, despite being in existence for well over 50 years, it is unclear to what extent they take place in veterinary medicine [39]. Conceptually, M&MCs can serve as a forum for collaborative review and investigation of incidents without fear of negative personal or professional consequences. When done well, M&MCs represent an opportunity to improve patient safety and maximize learning from an incident through open reflection and discussion [40–44].

In human medicine, M&MCs have successfully driven improved outcomes in patient care and management, including improving safety culture and care quality and reducing mortality and malpractice claims [45]. For M&MCs to fulfill their potential, key considerations are case selection, duration and frequency, roles of moderator and presenter, presentation format, incident analysis technique, and outcome and follow‐up.

All cases of mortality resulting from error or potential error should be reviewed through an M&MC. In larger clinics/hospitals, where the number of cases presenting exceeds the capacity to review cases in a timely manner, it may be necessary to screen and select cases for M&MCs. This can be based on the number of systemic factors implicated, the number of future patients that may benefit, and educational value. Screening can be performed as part of the voluntary reporting system process.

Duration of M&MCs varies widely, from 20 to over 60 min, probably reflecting time available in many instances, as well as case complexity [40–42]. Similarly, the frequency of M&MCs is variable though once monthly is commonly reported in the literature [38]. There is a case to be made for having a regular schedule to emphasize that M&MCs are a normal and accepted part of clinical governance, rather than special events reflecting personal failings. Cases selected for M&MCs should be presented as soon as possible as a timely acknowledgment of the incident and to maximize the likelihood of collecting all pertinent information.

Moderators set the tone of M&MCs and must be familiar with the M&MC format, have a good understanding of analysis techniques, have sufficient experience and expertise to guide the presenter and audience, and be respected by the attendees [40,46,47]. Cases should be presented by someone directly involved as they are best placed to present the events and answer questions. The presenter can be a senior or junior team member. It is helpful to have senior members intermittently present to show their investment in the process and to demonstrate to junior/new team members the expected format and standard. Involvement of junior members as presenters is invaluable as an educational tool for critical evaluation of case management and in maintaining a just and learning culture. Audiences for M&MCs should reflect the personnel of the clinic (i.e., all members of staff should be invited). Attendance by senior team members shows support for the process and helps create a positive, productive discussion [39]. A multidisciplinary audience can enrich the discussion by bringing different perspectives and disseminate lessons learned more widely [39].

Standardized presentation formats help ensure that information is presented in a systematic, organized way, minimizing the risk of bias [39,41,42,48]. One such model is the “Situation, Background, Assessment, Recommendation (SBAR)” format (Table 3.2), a structured method for efficiently transferring information [41,42]. This method has been effectively used when information is being passed between personnel who occupy different positions in a hierarchy (e.g., senior clinician and intern) [49,50].

Root cause analysis

Root cause analysis (RCA) is a generic term used to describe a range of techniques which aim to identify problems and then work toward establishing the problem’s root cause(s). A common misconception of RCA springs from the word “cause” appearing as singular, leading to the belief that a problem results from a single cause, whereas multiple causes are almost always involved, particularly in healthcare [51]. The simplest RCA method is the “five whys” technique, which involves asking “why?” five times, with each question following on from the previous answer to identify root causes of a problem. However, this technique promotes linear thinking, in that other contributing factors may be overlooked. Consequently, such techniques have now been replaced with more structured techniques, which ensure that the entire system is considered within the analysis.

Human Factors analysis techniques

Modern incident analysis techniques involve investigating the role of the entire system in the evolution of an incident. The “London Protocol” is a model of healthcare incident evolution based upon Reason’s accident model and Human Factors approaches (Fig. 3.3) [52,53]. Reason’s “unsafe acts” are termed “care delivery problems” and are affected by contributory factors, which can be classified into patient, task, individual team, work environment, and organizational/cultural levels. The framework provides a systematic and conceptually driven approach to accident investigation and to risk assessment in healthcare.

Another system commonly used for incident analysis in healthcare is the “Human Factors Analysis and Classification System (HFACS)” (Fig. 3.4). This system, based on Reason’s model and developed for the investigation of aviation accidents [54], has been adapted to the healthcare setting [51,55]. Used correctly, this approach is believed to have the “potential to identify actionable systemic causes of error, focus specific performance improvement efforts, and ultimately improve patient safety” [51].

A fishbone diagram can be used as a visual representation of an analysis and an accessible entry into a system‐based approach for incident investigation (Fig. 3.5) [34,38,56].

Checklists

Checklists are an organized list of action items or criteria that a user can record as present/absent as each item is considered or completed [60]. Fundamentally, the purpose of a checklist is to reduce error and improve performance [60]. A checklist achieves this by reinforcing accepted safety practices and fostering communication and teamwork, reducing the risk of error, and improving patient outcomes [61–64].

Much of the literature on checklists use comes from aviation, reflecting its early acknowledgment of the relationships between humans and complex systems. In aviation, checklists are a standard part of flight protocol; not completing checklists, or completing a checklist from memory, are considered violations [65]. By comparison, the widespread use of checklists is relatively recent in medicine but has been shown to be highly effective. This is exemplified by two well‐known examples: the “Keystone Intensive Care Unit (ICU) Project” and the “Surgical Safety Checklist (SSC)” [61,62]. In the “Keystone ICU Project,” catheter‐related bloodstream infections were reduced from a mean baseline rate of 7.7 infections per 1000 catheter days to 1.4 infections per 1000 catheter days at 16–18 months after the introduction of a simple five‐point checklist. These improvements were sustained over 5 years of follow‐up, with estimated savings of $2–3 billion annually, and prevention of 30,000–60,000 deaths [62,63]. The introduction of the “SSC” in a diverse patient population from eight hospitals in eight countries, representing both high‐ and low‐income settings, reduced mortality within 30 days of non‐cardiac surgery from 1.5% to 0.8% and the overall complication rate (including surgical site infections, sepsis, and pneumonia, among others) from 11.0% to 7.0% [61]. The positive outcomes from this study led to its adoption by the WHO, its use in over 120 countries, representing 90% of the world’s population, and over 230 million surgeries per year. The checklist comprises three sections: “sign in” (before induction of anesthesia), “time out” (before skin incision), and “sign out” (before patient leaves operating room), with five to seven checklist items per section [66]. The mechanisms underlying the improvements achieved were “likely multifactorial,” reflecting both systems and behavioral changes [61]. These can be described in terms of technical and adaptive (cultural) aspects. Technical aspects encompass education and evaluation, and adaptive aspects encompass engagement and execution [62,63,66,67]. All must be considered for a checklist to be successful in design and implementation [63].

Other cognitive aids

There are many other cognitive aids that can be used to improve human performance and safety. Evidence for their use in settings outside of medicine is compelling and has led to much interest especially in acute medicine, surgery, and anesthesia [73]. Their use in emergency situations has perhaps received the most attention as it appears that failure to adhere to best clinical practice often occurs when time for decision‐making is limited [74]. Many cognitive aids have been developed for emergency situations ranging from single‐treatment algorithms and drug charts to large sets of checklists and cognitive aids grouped together in a “crisis manual.”

Some of the earliest cognitive aids developed and put to widespread use were cardiopulmonary resuscitation (CPR) algorithms. In CPR, adherence to Advanced Cardiovascular Life Support protocols is associated with increased patient survival, whereas non‐compliance and omissions of indicated steps are associated with decreased survival [75]. Error rates during CPR may be halved by using cognitive aids, at least in a high‐fidelity simulator setting, suggesting that the use of cognitive aids may have significant effects on outcomes [76].

The Reassessment Campaign on Veterinary Resuscitation (RECOVER) developed a set of consensus guidelines for veterinary CPR alongside cognitive aids encompassing a CPR algorithm, a post‐cardiac arrest care algorithm, and a quick reference chart of CPR drugs and doses [77]. Little evidence has been published as to the success of these guidelines to date; however, it appears the guidelines have changed CPR practices [78], although they still vary considerably [79]. One study of 141 dogs suffering cardiac arrest demonstrated that dogs resuscitated using the RECOVER guidelines and cognitive aids had better outcomes than dogs resuscitated using traditional CPR procedures, with improved return to spontaneous circulation (43% versus 17%) and higher survival rates (5% versus 0%) [80].

Several crisis manuals have now been developed for use in human anesthesia and surgery. Based upon information from the “Australian Incident Monitoring System (AIMS),” Runciman et al. developed a set of 24 crisis algorithms for managing anesthetic crises [81]. It is claimed that 60% of 4000 incidents reported to AIMS would be addressed in 40–60 s using the manual [81]. In a pilot trial of a surgical crisis manual developed by collaboration between medical and aviation experts, two surgical teams were exposed to eight crisis simulations, four with and four without the checklist [82]. Manual use led to a sixfold reduction in failure to perform critical steps and tasks [82]. Later, 17 operating room teams were exposed to 106 surgical crisis simulations. All teams performed better when using the manual and there was almost a 75% reduction in omissions of critical steps during crisis management [83].

Communication tools: briefings, debriefings, and patient hand‐offs

Communication is a significant component of perioperative checklists such as the WHO SSC, and in many organizations, the surgical “time‐out” has developed into a miniature pre‐procedural briefing. A recent study including over 8000 surgical procedures investigated outcomes before and after the introduction of a structured intraoperative briefing and found that mortality, unplanned re‐operations, and prolonged hospital stays were all reduced [84].

A point of weakness for transfer of critical patient‐related information is during hand‐offs (handovers) [85,86]. Hand‐offs occur whenever a transfer and acceptance of patient care responsibility between two caregivers or teams is required, for example, when a patient is moved from the operating theater to the recovery suite or ward. Effective hand‐offs require passing safety‐critical and patient‐specific information from one caregiver to another and are key in facilitating the continuity and safety of the patient’s care. Unstable recovering patients, competing demands upon caregivers, multitasking, and time limitations make postanesthetic hand‐offs particularly vulnerable to error [86]. Such errors lead to fragmented postoperative care, delays in treatment and diagnosis, and cause significant patient harm [86]. A systematic review of the literature identified broad strategies to reduce risk around postanesthetic hand‐offs which are summarized in Box 3.3 [85].