**2. The anaesthesia situation**

In this first section, we will present the anaesthesia situation according to the temporal organisation of the anaesthetist's activity situation characteristics and constraints. Finally, risks linked to anaesthesia will be presented through epidemiological studies.

#### **2.1 The anaesthesia process**

Anaesthesia is commonly defined as the administration of medications, called anaesthetics, to control pain during a medical procedure or surgery. There are three kind of anaesthesia: the local one used for minor surgery; the regional one inducing sedation and, the general anaesthesia used for major surgeries and inducing unconsciousness and complete pain control. In this chapter, we will mainly focus on general anaesthesia.

Moreover, the anaesthetist has 3 additional roles. First, he has to work in collaboration with the surgeon to evaluate and prepare the patient. Second, he has to provide medical care to the patient during the surgery. Finally, he is directly involved in the patient's management after the surgical intervention (Chung & Lam, 1990). According to these roles, 3 phases are distinguished: the pre-operatory one, the per-operatory one and the post-operatory phase. The figure 1 presents the anaesthesia process in France.

Fig. 1. The anaesthesia process in France

#### **2.1.1 The pre-operatory phase**

206 Novel Approaches and Their Applications in Risk Assessment

patient's files according to their risk. While the second one consists in semi-structured interviews revolved on simulated cases using the information on request technique. Finally,

In this first section, we will present the anaesthesia situation according to the temporal organisation of the anaesthetist's activity situation characteristics and constraints. Finally,

Anaesthesia is commonly defined as the administration of medications, called anaesthetics, to control pain during a medical procedure or surgery. There are three kind of anaesthesia: the local one used for minor surgery; the regional one inducing sedation and, the general anaesthesia used for major surgeries and inducing unconsciousness and complete pain

Moreover, the anaesthetist has 3 additional roles. First, he has to work in collaboration with the surgeon to evaluate and prepare the patient. Second, he has to provide medical care to the patient during the surgery. Finally, he is directly involved in the patient's management after the surgical intervention (Chung & Lam, 1990). According to these roles, 3 phases are distinguished: the pre-operatory one, the per-operatory one and the post-operatory phase.

risks linked to anaesthesia will be presented through epidemiological studies.

control. In this chapter, we will mainly focus on general anaesthesia.

The figure 1 presents the anaesthesia process in France.

Fig. 1. The anaesthesia process in France

our results will be discussed in regard of theories used.

**2. The anaesthesia situation** 

**2.1 The anaesthesia process** 

This phase, the first box on the left, allows the anaesthetist to prepare the patient for the surgery and to highlight important elements to consider during this one (Anceaux & Beuscart-Zéphir, 2002). Due to the French specificity (for recall, the existence of a specific consultation). This first phase is divided into two different steps: the consultation and the anaesthetic visit.

The consultation takes place 2 weeks before the surgery. During this one, the anaesthetist looks for information about the patient's disease through interviews (medical history, comorbidities, medication and allergies) and physical examination, establishes a family medical history and tries to reassure the patient. This one allows the anaesthetist to prescribe additional tests, to adapt the treatment, and if needed, to optimize the patient's health.

The second step, the anaesthetic visit, takes place the day before or the same day of the surgery. Its major aim is to verify the absence of interfering elements occurred between the consultation and the visit.

#### **2.1.2 The per-operatory phase**

This second phase starts when the anaesthetist meets the patient at the entrance to the operating room. Two different steps compose this phase. The first one is the anaesthesia induction. It means the administration of anaesthetic agents and the establishment of adequate depth anaesthesia for surgery. It is an important step physically and mentally (McDonald & Dzwonczyk, 1988; McDonald et al*.*, 1990; Gaba & Lee, 1990). Physically because a lot of actions are needed (Xiao, 1994): preparation and injection of anaesthetic drugs, airway intubation, breathing circuit connections, programing a precise mechanical ventilation, etc. Mental workload is also observed. In fact, the patient's physiological status will change very quickly due to injected drugs. To interpret all the information relayed by the monitoring and the physical examination, the anaesthetist has to construct a specific representation of the patient's health state. Then, he has to check the data progress.

The second step of this phase is the maintenance of anaesthesia throughout the medical procedure. During this step, the anaesthetist's main task is to monitor the patient's vital signs and the progress of surgery.

A task analysis (Neyns, 2011) has highlighted 4 categories of anaesthetists' observable behaviours.

The first category relates to the anaesthetist's need to monitor the situation, assess the evolution of specific variables, understand the situation, etc. It is the **information-gathering step**. Five sources were underlined: *the patient's record* that includes both the pre-anaesthetic file (form the consultation and additional assessments) and the per-operative sheet that includes all patient data (drugs injected, vital signs, all relevant information); the *monitoring and its alarms*; the *surgery status and progress* allowing to explain changes in patient's vital signs or to restore the patient's anaesthesia; information taking on *patient* (apart from monitoring data) used to confirm hypothesis; and information related to *hour* either to check the schedule or to check the drugs effect, or for a personal need.

Risk Assessment in the Anaesthesia Process 209

Such situations are highly risky due to their dynamics characteristics. In the next section, we

As reported by the French Society of Anaesthesiology (SFAR, 2010), the rate of serious complications in the world is still 14 millions with a death rate of 2 millions. While some studies (Marty, 2003; Amalberti et al., 2005) have shown that anaesthesia was a safe system even so there are more difficult patients (older, more complicated diseases, etc.) and more complex surgeries, etc. Epidemiological studies have enabled to highlight risky factors in

Classically, two kinds of risks are distinguished: risks in the operating room and out of this one. Two predominant factors were highlighted by epidemiological studies (*e.g.* Cohen et al., 1988; Arbous et al., 2001). The first one is the most important in the operatory mortality and concerns the patient's features (physical state, age, gender). The second one is related to the surgery difficulty depending on the level of complexity (minor, intermediate, major) and the status of surgery (planned or urgent). A third one is also observed: the risk linked to the anaesthesia. Three kinds of variables in risks strictly linked to anaesthesia are identified: (1) the inadequate assessment and preparation of the patient to the surgery (25%); (2) the unsuitable choice of an anaesthetic technique (15%); (3) a wrong recovery action (56%). But this last risk is not really salient. More precisely, Lagasse (2002) estimates that the risk strictly linked to the anaesthesia is less than 1/10000, while the risk linked to the surgery is 1/100 to 1/1000 and the one linked to the patient is correlated to the patient's ASA score1

In summary, these results justify the relevance of anaesthetic consultation for the risk assessment and the patient's preparation to the surgery. However, this one is largely responsible of some complications (coupled with an inadequate choice of technique included in the anaesthetic evaluation). The study led by Zhang et al. (2002) has pointed out that the solution to many medical errors was to understand the operator's cognition and to adopt cognitive methodologies to prevent and avoid these errors. In the next section, we will present studies in cognitive psychology to highlight the cognitive mechanics involved in anaesthesia risks assessment (*e.g.* anticipation) and risks management (*e.g*. error detection

**3. Some psychological concepts about risk management in dynamic** 

In the management of a dynamic situation such as anaesthesia, the time management is important (*e.g.* Carreras et al., 2001). In fact, a dynamic environment can be managed before,

1 ASA is a physical status classification system for assessing the fitness of patients before surgery. It is a six-category classification system. ASA1 for a normal healthy patient; ASA2 for a patient with mild systemic disease; ASA3 for a patient with severe systemic disease; ASA4 for a patient with severe systemic disease that is a constant threat to life; ASA5 for a moribund patient who is not expected to survive without the surgery and ASA 6 for a declared brain-dead patient whose organs are being

will present epidemiological studies on anaesthetic security.

(e.g. ASA3 = 1/10000, ASA4 = 5/10000 and ASA5 =14/10000).

anaesthesia (*e.g.* Cohen et al., 1988 ; Sfez, 2002).

**2.2 Risks in anaesthesia** 

and recovery).

**situation** 

removed for donor purposes.

The second category refers to variables that are not directly controlled by the anaesthetist (*e.g.* the conduct of the intervention). He has then to refer to other members in the operating room (*e.g.* surgeon, nurses, surgical assistant, etc.) and to other health care professionals who are not directly involved in the surgical procedure (*e.g.* cardiologist, critical care services, recovery room, etc.). It is **functional communications** between the different members. These communications can go from the anaesthetist to other members but also to the anaesthetist. The communication objects can be multiple: to provide/exchange information, to help the anaesthetist to detect/identify a problem, to discuss plans for the previous/present/next patient, etc.

The third category includes the anaesthetist's **actions on the environment** (*e.g.* actions on the monitoring interfaces to facilitate data interpretation or to set to the patient's vital constants, preparation of the equipment, etc.) and **the patient** (*e.g.* drugs injections, intubation, etc.). His actions are initiated by either obligatory procedure (*e.g.* prescribed tasks), or action plans, or by environment data (*e.g.* information gathering/communication/alarms). Concerning alarms, they play a role in detecting problems. Some alarms can be the result of a failure in the patient's vital signs. The anaesthetist will act to restore the situation in a "normal way". But others can be caused by interferences (*e.g.* the electric scalpel of the surgeon can cause false cardiac dysrhythmia). These alarms are ignored because they are not significant and the anaesthetist cannot do anything against them. Finally, other alarms prompt the anaesthetist to seek for further information (*e.g.* check the surgeon' advice about the surgery, an act painful for the patient can lead to an increased heart rate).

Finally, the fourth category includes two types of actions: **non-functional actions** that are not related to the intervention (*e.g.* reading articles or emails and communications with other members, telephone, etc.) and the **anaesthetist's absence** from the operating room who delegates the process supervising to other members in the operating room.

Thus, this task analysis shows that the anaesthesia management includes extensive information and various parameters can be taken into account. It demonstrates the system complexity and its relative reliability.

#### **2.1.3 The post-operatory phase**

This last phase mainly concerns the patient's awakening. Once the surgery is completed, the anaesthetist has to reverse the effect of anaesthesia as quickly as possible.

Another characteristic of the anaesthetic process in France is that the anaesthetist involved in the anaesthesia consultation will not be necessarily the one on call during the surgery. This cooperative activity implies that the anaesthetist in charge of the consultation collects relevant information about the patient, builds a representation of this case, plans activities and finally chooses the information to convey to his colleague. This information will be provided on an anaesthetic file which is used as an external memory (Anceaux et al., 2002 ; Thuilliez et al., 2005; Neyns et al., 2010).

In summary, anaesthesia process is a dynamic and complex work environment sharing common characteristics with other dynamics work environments as aviation, car driving and nuclear power plants management (Woods, 1988; Vicente, 1999).

Such situations are highly risky due to their dynamics characteristics. In the next section, we will present epidemiological studies on anaesthetic security.

#### **2.2 Risks in anaesthesia**

208 Novel Approaches and Their Applications in Risk Assessment

The second category refers to variables that are not directly controlled by the anaesthetist (*e.g.* the conduct of the intervention). He has then to refer to other members in the operating room (*e.g.* surgeon, nurses, surgical assistant, etc.) and to other health care professionals who are not directly involved in the surgical procedure (*e.g.* cardiologist, critical care services, recovery room, etc.). It is **functional communications** between the different members. These communications can go from the anaesthetist to other members but also to the anaesthetist. The communication objects can be multiple: to provide/exchange information, to help the anaesthetist to detect/identify a problem, to discuss plans for the

The third category includes the anaesthetist's **actions on the environment** (*e.g.* actions on the monitoring interfaces to facilitate data interpretation or to set to the patient's vital constants, preparation of the equipment, etc.) and **the patient** (*e.g.* drugs injections, intubation, etc.). His actions are initiated by either obligatory procedure (*e.g.* prescribed tasks), or action plans, or by environment data (*e.g.* information gathering/communication/alarms). Concerning alarms, they play a role in detecting problems. Some alarms can be the result of a failure in the patient's vital signs. The anaesthetist will act to restore the situation in a "normal way". But others can be caused by interferences (*e.g.* the electric scalpel of the surgeon can cause false cardiac dysrhythmia). These alarms are ignored because they are not significant and the anaesthetist cannot do anything against them. Finally, other alarms prompt the anaesthetist to seek for further information (*e.g.* check the surgeon' advice about the surgery, an act painful for the patient

Finally, the fourth category includes two types of actions: **non-functional actions** that are not related to the intervention (*e.g.* reading articles or emails and communications with other members, telephone, etc.) and the **anaesthetist's absence** from the operating room

Thus, this task analysis shows that the anaesthesia management includes extensive information and various parameters can be taken into account. It demonstrates the system

This last phase mainly concerns the patient's awakening. Once the surgery is completed, the

Another characteristic of the anaesthetic process in France is that the anaesthetist involved in the anaesthesia consultation will not be necessarily the one on call during the surgery. This cooperative activity implies that the anaesthetist in charge of the consultation collects relevant information about the patient, builds a representation of this case, plans activities and finally chooses the information to convey to his colleague. This information will be provided on an anaesthetic file which is used as an external memory (Anceaux et al., 2002 ;

In summary, anaesthesia process is a dynamic and complex work environment sharing common characteristics with other dynamics work environments as aviation, car driving

who delegates the process supervising to other members in the operating room.

anaesthetist has to reverse the effect of anaesthesia as quickly as possible.

and nuclear power plants management (Woods, 1988; Vicente, 1999).

previous/present/next patient, etc.

can lead to an increased heart rate).

complexity and its relative reliability.

Thuilliez et al., 2005; Neyns et al., 2010).

**2.1.3 The post-operatory phase** 

As reported by the French Society of Anaesthesiology (SFAR, 2010), the rate of serious complications in the world is still 14 millions with a death rate of 2 millions. While some studies (Marty, 2003; Amalberti et al., 2005) have shown that anaesthesia was a safe system even so there are more difficult patients (older, more complicated diseases, etc.) and more complex surgeries, etc. Epidemiological studies have enabled to highlight risky factors in anaesthesia (*e.g.* Cohen et al., 1988 ; Sfez, 2002).

Classically, two kinds of risks are distinguished: risks in the operating room and out of this one. Two predominant factors were highlighted by epidemiological studies (*e.g.* Cohen et al., 1988; Arbous et al., 2001). The first one is the most important in the operatory mortality and concerns the patient's features (physical state, age, gender). The second one is related to the surgery difficulty depending on the level of complexity (minor, intermediate, major) and the status of surgery (planned or urgent). A third one is also observed: the risk linked to the anaesthesia. Three kinds of variables in risks strictly linked to anaesthesia are identified: (1) the inadequate assessment and preparation of the patient to the surgery (25%); (2) the unsuitable choice of an anaesthetic technique (15%); (3) a wrong recovery action (56%). But this last risk is not really salient. More precisely, Lagasse (2002) estimates that the risk strictly linked to the anaesthesia is less than 1/10000, while the risk linked to the surgery is 1/100 to 1/1000 and the one linked to the patient is correlated to the patient's ASA score1 (e.g. ASA3 = 1/10000, ASA4 = 5/10000 and ASA5 =14/10000).

In summary, these results justify the relevance of anaesthetic consultation for the risk assessment and the patient's preparation to the surgery. However, this one is largely responsible of some complications (coupled with an inadequate choice of technique included in the anaesthetic evaluation). The study led by Zhang et al. (2002) has pointed out that the solution to many medical errors was to understand the operator's cognition and to adopt cognitive methodologies to prevent and avoid these errors. In the next section, we will present studies in cognitive psychology to highlight the cognitive mechanics involved in anaesthesia risks assessment (*e.g.* anticipation) and risks management (*e.g*. error detection and recovery).

#### **3. Some psychological concepts about risk management in dynamic situation**

In the management of a dynamic situation such as anaesthesia, the time management is important (*e.g.* Carreras et al., 2001). In fact, a dynamic environment can be managed before,

<sup>1</sup> ASA is a physical status classification system for assessing the fitness of patients before surgery. It is a six-category classification system. ASA1 for a normal healthy patient; ASA2 for a patient with mild systemic disease; ASA3 for a patient with severe systemic disease; ASA4 for a patient with severe systemic disease that is a constant threat to life; ASA5 for a moribund patient who is not expected to survive without the surgery and ASA 6 for a declared brain-dead patient whose organs are being removed for donor purposes.

Risk Assessment in the Anaesthesia Process 211

However the limits of projection into the future described above (the knowledge, the rapid evolution and the level of uncertainty) also require a reactive mode. This reactive mode

All the risks cannot be anticipated. As we have seen before, certain risks are taken into account and actions are proposed (prediction, strategies of avoidance or management) and implemented, while others are ignored or not perceived, and their management is postponed (if they appear) in the process management. This type of management suggests a certain "allostasis risk" (Fuller et al., 2008). This second section aims to understand how

Although it is widely accepted that the operator commits many errors (Reason, 1990), he also detects the majority of them (at least 60% according to Allwood, 1984; Rizzo et al., 1987). Error detection is the first step in the risk management during the process. During this step, there is still no explanation of the error (Zapf & Reason, 1994). The operator is

Error detection can be done automatically, spontaneously and unconsciously or it is the result of a more aware comparative treatment with the initial objective (in terms of results, performance) or when the operator can no longer act on the system (Allwood & Montgmery, 1982; Allwood, 1984; Sellen, 1990; Sellen & Norman, 1992; Sellen, 1994; Rizzo et al., 1995;). Finally, the intervention of a third person is also a key factor in errors detection

Sellen (1990, 1994; et al., 1992) suggests that operator, before recovering the error, would be able to identify it. This identification process consists in a comparison between results obtained and those expected. Nevertheless, it is also possible that the operator detects an error without necessarily identify it or to identify it without necessarily recover it. Indeed, a significant part of detected but also not detected errors does not have any consequences (or little) on the process. Error recovery is often considered as the last step in risk management. It allows the operator to remove or reduce the gap between what he gets and what he wanted to get. Once the error is detected, recovery actions can take many forms (Sellen,

Regarding the anaesthesia situation, little attention has been paid to the mechanisms of error detection. Indeed the majority of research focuses on the study of the errors prevention (*e.g.* Cooper, et al., 1984; De Keyser & Nyssen, 1993; Cooper, et al., 1982; Finley & Cohen, 1991). In 2006, Nyssen & Blavier have shown that the majority of incidents were detected during the regular monitoring of parameters and on the basis of external signs, suggesting an "automatic" mode. The expertise seems also to play a role in error detection (Nyssen & De Keyser, 1998). The most experienced ones diagnose the problem with significantly less time

Finally, most studies about risk management (see Cooper et al., 1984) showed that anaesthetists do not taken seriously into account infrequent risks. Risk management depends thus mainly on the anaesthetist's ability to respond quickly and appropriately

aware of the existence of a gap between what he obtained and what he expected.

corresponds to the resilience (Hollnagel et al., 2006)

operators manage risks postponed in the process management.

1990). But the recovery process will not be detailed in this chapter.

**3.2 Risks management during the process** 

(Woods, 1984; Nyssen, 1997).

than novices.

when a problem occurs.

during and even after the process in itself (Morel et al., 2008a, 2008b). The operator has to learn to anticipate his actions and especially their effects. However, the different interactions between multiple variables do not always allow anticipating all the effects of operator's actions and the system evolution. So the operator must deal with uncertain situations in real time and adapt to variables involved in the process and their interactions. This adaptation refers to resilience that is defined as the ability of systems to find a stable state after a major disturbance and/or in a stressful situation (Hollnagel et al., 2006). According to Cook and Woods (1994), the first source of resilience is the human operator. Indeed Reason (2008) describes the human operator as "*an hero, a system element whose adaptations and compensations have brought troubled systems back from the brink of disaster on a significant number of occasions*" (p.3). In this chapter we will mainly discuss the risk management before and during the process.

#### **3.1 Risks management before the process**

As said earlier, the human operator can perform his task on a reactive mode but also on a proactive one. This last one permits to view the situation in the near future (limited by operator's knowledge and the rapid evolution of the process and its uncertainty) and makes possible the selection of information (filtering) (Hoc, 1995). We can observe such mode before the process with anticipatory activity (plans) but also during this one with the implementation of plans built, the preparation of the next action or a sequence of actions (Hoc, 2006). According to Cellier et al. (1996), anticipation in a dynamic work environment consists in (1) evaluating the future state of a dynamic process, (2) determining the type of actions to undertake and the moment when it must be done, and (3) mentally evaluating the impact of those actions. Denecker (1999) describes the anticipation on two different modes. The first one is considered as an explicit prediction or forecasting. The second one is implicit and expressed in a selective preparation to situations or an expectation of events. In this chapter, we will mainly focus on the explicit mode.

Regarding the anaesthesia situation, several studies (Xiao et al., 1997; Neyns et al., 2010) have shown that some strategies were built in advance during the pre-anaesthesia phase. Xiao et al. (1997) revealed that anaesthetists identified during this phase a list of "points for consideration". This list includes the specific conditions in the patient's physiology, anticipated events, and risks. This one functions then as a set of "warning" that guides the anaesthetist's attention in a dynamic, multi-tasking situation. However, Neyns (et al., 2010) showed that this list reflected a wide anticipation range (Boudes & Cellier, 1998). It means that this list does not contain planning elements per se, but they can guide attention in the anaesthesia activity where timing constraints are important. Thus the information recorded by anaesthetists concerns the plans adjustment and little information concerns the development or the selection of a plan (Anceaux, et al*.*, 2002). The information contained in the consultation files has two distinct purposes: (1) to assess the risk factors and to prevent them by the development of strategies (related to the patient and the surgery), and (2) the identification of specific problems (Neyns et al., 2010).

Finally, other studies (Anceaux et al., 2002; Anceaux et al., 2001; Beuscart-Zéphir et al., 2001; Thuilliez, et al., 2005) have pointed out that the way to anticipate during the first phase depended on the anaesthetist's expertise and the case complexity.

during and even after the process in itself (Morel et al., 2008a, 2008b). The operator has to learn to anticipate his actions and especially their effects. However, the different interactions between multiple variables do not always allow anticipating all the effects of operator's actions and the system evolution. So the operator must deal with uncertain situations in real time and adapt to variables involved in the process and their interactions. This adaptation refers to resilience that is defined as the ability of systems to find a stable state after a major disturbance and/or in a stressful situation (Hollnagel et al., 2006). According to Cook and Woods (1994), the first source of resilience is the human operator. Indeed Reason (2008) describes the human operator as "*an hero, a system element whose adaptations and compensations have brought troubled systems back from the brink of disaster on a significant number of occasions*" (p.3). In this chapter we will mainly discuss the risk management before and during the

As said earlier, the human operator can perform his task on a reactive mode but also on a proactive one. This last one permits to view the situation in the near future (limited by operator's knowledge and the rapid evolution of the process and its uncertainty) and makes possible the selection of information (filtering) (Hoc, 1995). We can observe such mode before the process with anticipatory activity (plans) but also during this one with the implementation of plans built, the preparation of the next action or a sequence of actions (Hoc, 2006). According to Cellier et al. (1996), anticipation in a dynamic work environment consists in (1) evaluating the future state of a dynamic process, (2) determining the type of actions to undertake and the moment when it must be done, and (3) mentally evaluating the impact of those actions. Denecker (1999) describes the anticipation on two different modes. The first one is considered as an explicit prediction or forecasting. The second one is implicit and expressed in a selective preparation to situations or an expectation of events. In this

Regarding the anaesthesia situation, several studies (Xiao et al., 1997; Neyns et al., 2010) have shown that some strategies were built in advance during the pre-anaesthesia phase. Xiao et al. (1997) revealed that anaesthetists identified during this phase a list of "points for consideration". This list includes the specific conditions in the patient's physiology, anticipated events, and risks. This one functions then as a set of "warning" that guides the anaesthetist's attention in a dynamic, multi-tasking situation. However, Neyns (et al., 2010) showed that this list reflected a wide anticipation range (Boudes & Cellier, 1998). It means that this list does not contain planning elements per se, but they can guide attention in the anaesthesia activity where timing constraints are important. Thus the information recorded by anaesthetists concerns the plans adjustment and little information concerns the development or the selection of a plan (Anceaux, et al*.*, 2002). The information contained in the consultation files has two distinct purposes: (1) to assess the risk factors and to prevent them by the development of strategies (related to the patient and the surgery), and (2) the

Finally, other studies (Anceaux et al., 2002; Anceaux et al., 2001; Beuscart-Zéphir et al., 2001; Thuilliez, et al., 2005) have pointed out that the way to anticipate during the first phase

process.

**3.1 Risks management before the process** 

chapter, we will mainly focus on the explicit mode.

identification of specific problems (Neyns et al., 2010).

depended on the anaesthetist's expertise and the case complexity.

However the limits of projection into the future described above (the knowledge, the rapid evolution and the level of uncertainty) also require a reactive mode. This reactive mode corresponds to the resilience (Hollnagel et al., 2006)

#### **3.2 Risks management during the process**

All the risks cannot be anticipated. As we have seen before, certain risks are taken into account and actions are proposed (prediction, strategies of avoidance or management) and implemented, while others are ignored or not perceived, and their management is postponed (if they appear) in the process management. This type of management suggests a certain "allostasis risk" (Fuller et al., 2008). This second section aims to understand how operators manage risks postponed in the process management.

Although it is widely accepted that the operator commits many errors (Reason, 1990), he also detects the majority of them (at least 60% according to Allwood, 1984; Rizzo et al., 1987). Error detection is the first step in the risk management during the process. During this step, there is still no explanation of the error (Zapf & Reason, 1994). The operator is aware of the existence of a gap between what he obtained and what he expected.

Error detection can be done automatically, spontaneously and unconsciously or it is the result of a more aware comparative treatment with the initial objective (in terms of results, performance) or when the operator can no longer act on the system (Allwood & Montgmery, 1982; Allwood, 1984; Sellen, 1990; Sellen & Norman, 1992; Sellen, 1994; Rizzo et al., 1995;). Finally, the intervention of a third person is also a key factor in errors detection (Woods, 1984; Nyssen, 1997).

Sellen (1990, 1994; et al., 1992) suggests that operator, before recovering the error, would be able to identify it. This identification process consists in a comparison between results obtained and those expected. Nevertheless, it is also possible that the operator detects an error without necessarily identify it or to identify it without necessarily recover it. Indeed, a significant part of detected but also not detected errors does not have any consequences (or little) on the process. Error recovery is often considered as the last step in risk management. It allows the operator to remove or reduce the gap between what he gets and what he wanted to get. Once the error is detected, recovery actions can take many forms (Sellen, 1990). But the recovery process will not be detailed in this chapter.

Regarding the anaesthesia situation, little attention has been paid to the mechanisms of error detection. Indeed the majority of research focuses on the study of the errors prevention (*e.g.* Cooper, et al., 1984; De Keyser & Nyssen, 1993; Cooper, et al., 1982; Finley & Cohen, 1991). In 2006, Nyssen & Blavier have shown that the majority of incidents were detected during the regular monitoring of parameters and on the basis of external signs, suggesting an "automatic" mode. The expertise seems also to play a role in error detection (Nyssen & De Keyser, 1998). The most experienced ones diagnose the problem with significantly less time than novices.

Finally, most studies about risk management (see Cooper et al., 1984) showed that anaesthetists do not taken seriously into account infrequent risks. Risk management depends thus mainly on the anaesthetist's ability to respond quickly and appropriately when a problem occurs.

Risk Assessment in the Anaesthesia Process 213

Each patient's file contained the following information in the same order: surgery information (name, operating position, surgery duration, type of intervention, risk of infection, latex allergy); patient's information (gender, age, weight/height, medications, comorbidities, other allergies, dentures, medical and surgical histories); data form the clinical examination (cardiovascular, pulmonary, intubation, risk of post-operative nausea and vomiting, remarks);

Sixteen patient's records (4 records by kind of cases) were constructed and presented in a random order. Anaesthetists were asked to sort them into several distinct groups such as they "go well together" using the think aloud method. At the end of the task, they had to

Twenty anaesthetists volunteers participated: 10 experienced anaesthetists (average years of experience=21.2, SD=6.23) and 10 novices (average years of experience=4.2, SD=0.4). They all came from different departments and different medical structures. They were recruited

Data will be first analysed according to the produced categories: number of categories and titles given to each category. Then, information used to produce categories will be recorded and compared according to the level of experience. Finally, verbalizations will be analysed as a material language and exploited through an approach of natural language processing (NPL). This use of NPL is in line with works on texts profiling (see Planes, 2011). We wanted to highlight experts' and novices' profiles answering the following question: What kind of information the two groups use when they assess a patient's file? Lexico, a language tool for analysing corpus, will be used (see Lebart & Salem, 1994). This tool allows to manipulate some statistical calculation on corpus (more particularly frequencies) and to calculate specificities (in other words, "are some terms more specific to experimented anaesthetists than novices?").

In this second subsection, we will firstly present categories produced by the two groups of anaesthetists according to their number and their titles. Finally, information used from their

Anaesthetists construct different categories according to their level of experience. The more experienced anaesthetists build more categories (M=4.4, SD=0.52) than novices (M=3.5, SD=0.53). The t-test applied to the data confirms that the level of experience influences the

Regarding the titles of categories, a difference is observed according to the level of experience. The table 1 and 2 show labels given to the categories formed by the two groups, the participants' number giving this label and the average number of reports in each

Looking to the first table, categories made by experimented anaesthetists, it seems that they mainly use the surgery difficulty and the risk level associated to patient's health. Some other

verbalisation data will be explained with the Natural Language Processing (NLP).

differentiation between patient's records (t(18)=3.86, *p<0.001*, *d*=1.8).

conclusion made according to information from consultation (ASA score).

by mail or by phone. The entire experience lasted up to 30 minutes.

explain their choices of categories giving them a title.

**4.2 Results** 

**4.2.1 Produced categories** 

category formed.

In summary, various studies have focused on highlighting either prevention strategies in advance or management strategies in real time. But none of them has yet made the connection between what is evaluated during the consultation and in the operating room. The French anaesthesia situation permits to study this assessment at both phases. More specifically, we want to answer the following questions: (1) what kind of risk is used when an anaesthetist has to assess a patient's file? Some studies have emphasized the effect of certain variables on the process of risk assessment. For example, Anceaux et al*.* (2002) showed that the case complexity could influence how to proceed with the patient during the consultation. Epidemiological studies (*e.g.* Lagasse, 2002) have also identified three kinds of risks as the cause of problems: those related to the patient, surgery and anaesthesia itself. It would be interesting to know how anaesthetists assess risks involved according to their complexity. (2) Epidemiological studies have shown on the one hand that infrequent risks are not taken into account by an anaesthetist (*e.g.* Cooper et al., 1982) and on the other hand that there are often problems when the anaesthetist can assess and prepare a patient (*e.g.* Arbous et al., 2001). How important are the risk frequency and the risk predictability in the assessment and the management in real time?

The next two studies tend to answer these questions and to understand the link between consultation and the anaesthesia in itself.

### **4. Study one: What kind of criteria does anaesthetist use to assess risks linked to a patient based on consultation information?**

This first study aimed to understand how anaesthetists gather patient's files according to their risks. We supposed that information gathering and plans production would depend on operators' level of experience and risks level, without indication about the kind of risks used (patient, surgery or anaesthetic). First, we will present the method used to answer our question and the method used to treat our data. Finally, we will present results obtained. A brief discussion will be proposed to introduce the next study.

### **4.1 Method**

We proposed a card-sorting task using patient's records. All the patient's files were constructed using real patients with the help of an experimented anaesthetist. Four kinds of patient's files were constructed according to the level of complexity defined by 2 combined variables: the surgery complexity (easy and complex surgery) and the complexity related to the patient's health (healthy and unhealthy patient).

An easy surgery is defined as a short-term surgery, minimally invasive (*e.g.* surgery for a breast cancer). A difficult surgery is defined as a long duration surgery (6 to 12hours) that is invasive or mutilating (*e.g.* surgery of the larynx cancer). A healthy patient is described as a normal healthy patient (*e.g.* ASA1) or with mild systemic disease (*e.g.* ASA2). The patient with a complex health status is a patient with important comorbidities and taking many drugs (*e.g.* ASA3 or 4).

Combining these 2 variables, 4 kinds of cases were obtained: easy surgery with healthy patient (called OSPS); easy surgery with unhealthy patient (called OSPC); complex surgery with healthy patient (called OCPS) and, complex surgery with unhealthy patient (called OCPC).

In summary, various studies have focused on highlighting either prevention strategies in advance or management strategies in real time. But none of them has yet made the connection between what is evaluated during the consultation and in the operating room. The French anaesthesia situation permits to study this assessment at both phases. More specifically, we want to answer the following questions: (1) what kind of risk is used when an anaesthetist has to assess a patient's file? Some studies have emphasized the effect of certain variables on the process of risk assessment. For example, Anceaux et al*.* (2002) showed that the case complexity could influence how to proceed with the patient during the consultation. Epidemiological studies (*e.g.* Lagasse, 2002) have also identified three kinds of risks as the cause of problems: those related to the patient, surgery and anaesthesia itself. It would be interesting to know how anaesthetists assess risks involved according to their complexity. (2) Epidemiological studies have shown on the one hand that infrequent risks are not taken into account by an anaesthetist (*e.g.* Cooper et al., 1982) and on the other hand that there are often problems when the anaesthetist can assess and prepare a patient (*e.g.* Arbous et al., 2001). How important are the risk frequency and the risk predictability in the

The next two studies tend to answer these questions and to understand the link between

This first study aimed to understand how anaesthetists gather patient's files according to their risks. We supposed that information gathering and plans production would depend on operators' level of experience and risks level, without indication about the kind of risks used (patient, surgery or anaesthetic). First, we will present the method used to answer our question and the method used to treat our data. Finally, we will present results obtained. A

We proposed a card-sorting task using patient's records. All the patient's files were constructed using real patients with the help of an experimented anaesthetist. Four kinds of patient's files were constructed according to the level of complexity defined by 2 combined variables: the surgery complexity (easy and complex surgery) and the complexity related to

An easy surgery is defined as a short-term surgery, minimally invasive (*e.g.* surgery for a breast cancer). A difficult surgery is defined as a long duration surgery (6 to 12hours) that is invasive or mutilating (*e.g.* surgery of the larynx cancer). A healthy patient is described as a normal healthy patient (*e.g.* ASA1) or with mild systemic disease (*e.g.* ASA2). The patient with a complex health status is a patient with important comorbidities and taking many

Combining these 2 variables, 4 kinds of cases were obtained: easy surgery with healthy patient (called OSPS); easy surgery with unhealthy patient (called OSPC); complex surgery with healthy patient (called OCPS) and, complex surgery with unhealthy patient (called OCPC).

**4. Study one: What kind of criteria does anaesthetist use to assess risks** 

**linked to a patient based on consultation information?** 

brief discussion will be proposed to introduce the next study.

the patient's health (healthy and unhealthy patient).

assessment and the management in real time?

consultation and the anaesthesia in itself.

**4.1 Method** 

drugs (*e.g.* ASA3 or 4).

Each patient's file contained the following information in the same order: surgery information (name, operating position, surgery duration, type of intervention, risk of infection, latex allergy); patient's information (gender, age, weight/height, medications, comorbidities, other allergies, dentures, medical and surgical histories); data form the clinical examination (cardiovascular, pulmonary, intubation, risk of post-operative nausea and vomiting, remarks); conclusion made according to information from consultation (ASA score).

Sixteen patient's records (4 records by kind of cases) were constructed and presented in a random order. Anaesthetists were asked to sort them into several distinct groups such as they "go well together" using the think aloud method. At the end of the task, they had to explain their choices of categories giving them a title.

Twenty anaesthetists volunteers participated: 10 experienced anaesthetists (average years of experience=21.2, SD=6.23) and 10 novices (average years of experience=4.2, SD=0.4). They all came from different departments and different medical structures. They were recruited by mail or by phone. The entire experience lasted up to 30 minutes.

Data will be first analysed according to the produced categories: number of categories and titles given to each category. Then, information used to produce categories will be recorded and compared according to the level of experience. Finally, verbalizations will be analysed as a material language and exploited through an approach of natural language processing (NPL). This use of NPL is in line with works on texts profiling (see Planes, 2011). We wanted to highlight experts' and novices' profiles answering the following question: What kind of information the two groups use when they assess a patient's file? Lexico, a language tool for analysing corpus, will be used (see Lebart & Salem, 1994). This tool allows to manipulate some statistical calculation on corpus (more particularly frequencies) and to calculate specificities (in other words, "are some terms more specific to experimented anaesthetists than novices?").

#### **4.2 Results**

In this second subsection, we will firstly present categories produced by the two groups of anaesthetists according to their number and their titles. Finally, information used from their verbalisation data will be explained with the Natural Language Processing (NLP).

#### **4.2.1 Produced categories**

Anaesthetists construct different categories according to their level of experience. The more experienced anaesthetists build more categories (M=4.4, SD=0.52) than novices (M=3.5, SD=0.53). The t-test applied to the data confirms that the level of experience influences the differentiation between patient's records (t(18)=3.86, *p<0.001*, *d*=1.8).

Regarding the titles of categories, a difference is observed according to the level of experience. The table 1 and 2 show labels given to the categories formed by the two groups, the participants' number giving this label and the average number of reports in each category formed.

Looking to the first table, categories made by experimented anaesthetists, it seems that they mainly use the surgery difficulty and the risk level associated to patient's health. Some other

Risk Assessment in the Anaesthesia Process 215

The table 3 shows the average number of information used according to the cases and the anaesthetists' level of experience. Even if there is an augmentation of the information used according to the level of complexity (F(3,54)=66.09, *p*<0.0001, *f=1.92*), there is no effect of the

OSPS OSPC OCPS OCPC Average

level of experience on the number of information used (F(1,18)=1,09, NS, *f=0.25*).

 N M SD M SD M SD M SD Experimented 10 19.8 3.08 30.3 2.94 31 3.22 37.3 4.95 27.1 Novices 10 24.9 4.68 36.4 5.03 26.7 4.24 33.5 3.82 30.38 Table 3. Average number of information used according to the level of complexity and the

According to the NLP, the figure 2 shows frequencies of criteria used by the two groups (experimented and novices) in their verbalization data. Results show that experimented anaesthetists use 4 different criteria. They use more frequently the patient criteria (frequency=406), followed by the surgery criteria (frequency=217), the risk criteria (frequency=138) and finally, the problem criteria (frequency=118). Novices use also 4 different criteria. They use more frequently the surgery criteria (frequency=373), directly followed by the patient criteria (frequency=321) and less often the risk (frequency=100) and

Fig. 2. Frequencies of criteria used to assess patient's files according to the level of

As showed on the figure 1, two major differences appear between the two groups. Experimented anaesthetists mainly use the patient criteria when they have to assess a case,

**4.2.2 Information used by the two groups** 

anaesthetists' level of experience

the problems criteria (frequency=57).

experience (verbalization data)

This subsection concerns the information use to assess patient's files.

categories can highlight specific problems (*e.g.* neurological problems, attention to the induction time).


Table 1. Categories made by experimented anaesthetists


Table 2. Categories made by novices

Regarding the second table, categories made by novices, it seems that they manly used the overall risks and do not really distinguish what is related to the surgery and to the patient. There is just one category that includes this differentiation, the category named "long surgery and low-risk patient", but it is poorly used (just 2 novices with an average of 3 reports). Finally, they also use some categories highlighting specific problems (*e.g.* additional tests, post-operative management, particular intubation).

In summary, 90% of experimented anaesthetists distinguish the difficulty of the surgery and risks linked to the patient's health state, while 60% of novices suggest the overall risk associated with patient's files. According to the number of categories formed and qualitative analysis of labels, experimented anaesthetists use more differentiations between records than do novices. In the next subsection, we will present how anaesthetists analyse a patient's file through the natural language processing.

categories can highlight specific problems (*e.g.* neurological problems, attention to the

Categories Label Participants'

Easy surgery, healthy patient 10 5.8 Easy surgery but patient with mild systemic disease 8 3.4 Surgery moderately complex with minor risks 8 2.1 Complex surgery with major risks 10 4.4 Neurological problems 1 4 Very difficult case 2 2.5 Attention to the induction time 2 2.5

Categories Label Participants'

Not at risk 8 5.5 Low level of risk 6 4.5 Long surgery, low-risk patient 2 3 Very high risk 10 4.7 Intermediate risks 4 4.25 Additional tests 1 2 Post-operative management 1 1 Particular intubation 1 3

Regarding the second table, categories made by novices, it seems that they manly used the overall risks and do not really distinguish what is related to the surgery and to the patient. There is just one category that includes this differentiation, the category named "long surgery and low-risk patient", but it is poorly used (just 2 novices with an average of 3 reports). Finally, they also use some categories highlighting specific problems (*e.g.*

In summary, 90% of experimented anaesthetists distinguish the difficulty of the surgery and risks linked to the patient's health state, while 60% of novices suggest the overall risk associated with patient's files. According to the number of categories formed and qualitative analysis of labels, experimented anaesthetists use more differentiations between records than do novices. In the next subsection, we will present how anaesthetists analyse a patient's

additional tests, post-operative management, particular intubation).

Table 1. Categories made by experimented anaesthetists

Table 2. Categories made by novices

file through the natural language processing.

number

number

Average number of reports

Average number of reports

induction time).

#### **4.2.2 Information used by the two groups**

This subsection concerns the information use to assess patient's files.

The table 3 shows the average number of information used according to the cases and the anaesthetists' level of experience. Even if there is an augmentation of the information used according to the level of complexity (F(3,54)=66.09, *p*<0.0001, *f=1.92*), there is no effect of the level of experience on the number of information used (F(1,18)=1,09, NS, *f=0.25*).


Table 3. Average number of information used according to the level of complexity and the anaesthetists' level of experience

According to the NLP, the figure 2 shows frequencies of criteria used by the two groups (experimented and novices) in their verbalization data. Results show that experimented anaesthetists use 4 different criteria. They use more frequently the patient criteria (frequency=406), followed by the surgery criteria (frequency=217), the risk criteria (frequency=138) and finally, the problem criteria (frequency=118). Novices use also 4 different criteria. They use more frequently the surgery criteria (frequency=373), directly followed by the patient criteria (frequency=321) and less often the risk (frequency=100) and the problems criteria (frequency=57).

Fig. 2. Frequencies of criteria used to assess patient's files according to the level of experience (verbalization data)

As showed on the figure 1, two major differences appear between the two groups. Experimented anaesthetists mainly use the patient criteria when they have to assess a case,

Risk Assessment in the Anaesthesia Process 217

The first study was only focused on the risk assessment that can occur during the consultation. This second study aims to investigate it during all the anaesthesia process. We wanted to see how anaesthetists assess a predictable risk and how they do when they face an unpredictable risk. We supposed that anaesthetists would pay more attention to a predictable risk than to an unpredictable one. This attention would be clearly observed in the risk assessment during the consultation and also, in the risk management (during the surgery) by a better detection/identification. First, we will present the method chosen to answer our question and finally, we will present and discuss briefly a part of our results.

This second study consists of semi-structured interviews revolved on simulated cases (already validated in Sfez, et al., 2008) using the technique of information on request (see Rimoldi, 1963). These cases were presented on vignettes that described an anaesthetic

Cases were selected based on two criteria: the frequency of occurrence and the predictability. Four cases were used in the experiment: the case of difficult intubation (predictable or not) and the case of malignant hyperthermia (predictable or not). The first case, difficult intubation, is a common situation. Moreover anaesthetists have to evaluate the intubation criteria during the consultation through patient's body inspection (mouth opening, length of upper incisors, relation of maxillary and mandibular incisors during normal jaw closure and during voluntary protrusion of mandible, intercisor distance, visibility of uvula, shape of palate, compliance of mandibular space, thyromental distance, length and thickness of neck, and range of motion of head and neck) and patient's interview (familial history). They also have a difficult airway algorithm (developed by the French Society of Anaesthesia) allowing them to choose an adequate technique according to the patient's criteria. In this simulation, the case of difficult intubation was coupled with gastric oesophageal reflux, which occurs if the anaesthetist does not practice preventive

The second case, malignant hyperthermia, was chosen because it is a rare life threatening condition in anaesthesia but it can be detected during the consultation. It is usually triggered by exposure to certain drugs used for general anaesthesia. There are also algorithms that manage this risk (strategies for prevention, avoidance and recovery actions). In the case presented, the anaesthetist can either detect the problem during the consultation, either during the anaesthetic induction with warning signs as muscle stiffness, difficulty of intubation due to spasms, breathing problems, increasing of the patient's body temperature

Each vignette was divided into 3 phases. The first one was related to the preoperative phase during which the anaesthetist has to be aware of the patient's physical conditions before to put him to sleep and has to determine an action plan to intervene. This phase corresponded to the anaesthesia consultation and to the risk that a problem could appear.

**5. Study two: How does an anaesthetist assess risks during a surgical** 

**intervention?** 

**5.1 Method** 

manoeuvre.

and arrhythmias.

situation that deviated form its normal course.

while novices use the surgery and the patient criteria without significant distinction. Finally, experimented anaesthetists and novices assess not only risks but they also highlight some specific problems to consider, even if the experimented ones use it more often. This last difference has a particular interest for us: either novices do not look at these particular problem as much as experts, or they use other words to describe these elements due to a difference in the vocabulary used.

The figure 3, called problems frequencies underlined by experimented anaesthetists and novices, shows that the two groups highlight all the problems. But the experimented ones underline problems more often than do novices (4 kinds of problems are more often cited by experimented anaesthetists while novices identified only two kinds most frequently).

Fig. 3. Problems frequencies underlined by experimented anaesthetists and novices

In summary, it appears that novices look the same information in the records than experimented anaesthetists. However, novices consider them less frequently as problems and analyse it as a whole. This suggests the existence of different strategies among experience level. With experience, anaesthetists can focus only on the potentially problematic aspects of patient's files.

#### **4.3 Brief discussion of the first study**

This first study permits to put in light differences in the risk assessment. In fact, the level of experience seems to be a relevant factor in it. As developed through epidemiological studies, our results insist on difficulties linked to the patient's health state and the surgery. Anaesthetists, whatever their level of experience, use these two criteria more frequently. However, the more experimented anaesthetists seem to insist on patient's risks factors than novices. These last ones seem to use both of these criteria to assess risk linked to one case. Finally, risk assessment is also a way to put some warning flags on points for consideration. That confirms Xiao's results (1996).

#### **5. Study two: How does an anaesthetist assess risks during a surgical intervention?**

The first study was only focused on the risk assessment that can occur during the consultation. This second study aims to investigate it during all the anaesthesia process. We wanted to see how anaesthetists assess a predictable risk and how they do when they face an unpredictable risk. We supposed that anaesthetists would pay more attention to a predictable risk than to an unpredictable one. This attention would be clearly observed in the risk assessment during the consultation and also, in the risk management (during the surgery) by a better detection/identification. First, we will present the method chosen to answer our question and finally, we will present and discuss briefly a part of our results.

#### **5.1 Method**

216 Novel Approaches and Their Applications in Risk Assessment

while novices use the surgery and the patient criteria without significant distinction. Finally, experimented anaesthetists and novices assess not only risks but they also highlight some specific problems to consider, even if the experimented ones use it more often. This last difference has a particular interest for us: either novices do not look at these particular problem as much as experts, or they use other words to describe these elements due to a

The figure 3, called problems frequencies underlined by experimented anaesthetists and novices, shows that the two groups highlight all the problems. But the experimented ones underline problems more often than do novices (4 kinds of problems are more often cited by experimented anaesthetists while novices identified only two kinds most frequently).

Fig. 3. Problems frequencies underlined by experimented anaesthetists and novices

In summary, it appears that novices look the same information in the records than experimented anaesthetists. However, novices consider them less frequently as problems and analyse it as a whole. This suggests the existence of different strategies among experience level. With experience, anaesthetists can focus only on the potentially

This first study permits to put in light differences in the risk assessment. In fact, the level of experience seems to be a relevant factor in it. As developed through epidemiological studies, our results insist on difficulties linked to the patient's health state and the surgery. Anaesthetists, whatever their level of experience, use these two criteria more frequently. However, the more experimented anaesthetists seem to insist on patient's risks factors than novices. These last ones seem to use both of these criteria to assess risk linked to one case. Finally, risk assessment is also a way to put some warning flags on points for consideration.

difference in the vocabulary used.

problematic aspects of patient's files.

That confirms Xiao's results (1996).

**4.3 Brief discussion of the first study** 

This second study consists of semi-structured interviews revolved on simulated cases (already validated in Sfez, et al., 2008) using the technique of information on request (see Rimoldi, 1963). These cases were presented on vignettes that described an anaesthetic situation that deviated form its normal course.

Cases were selected based on two criteria: the frequency of occurrence and the predictability. Four cases were used in the experiment: the case of difficult intubation (predictable or not) and the case of malignant hyperthermia (predictable or not). The first case, difficult intubation, is a common situation. Moreover anaesthetists have to evaluate the intubation criteria during the consultation through patient's body inspection (mouth opening, length of upper incisors, relation of maxillary and mandibular incisors during normal jaw closure and during voluntary protrusion of mandible, intercisor distance, visibility of uvula, shape of palate, compliance of mandibular space, thyromental distance, length and thickness of neck, and range of motion of head and neck) and patient's interview (familial history). They also have a difficult airway algorithm (developed by the French Society of Anaesthesia) allowing them to choose an adequate technique according to the patient's criteria. In this simulation, the case of difficult intubation was coupled with gastric oesophageal reflux, which occurs if the anaesthetist does not practice preventive manoeuvre.

The second case, malignant hyperthermia, was chosen because it is a rare life threatening condition in anaesthesia but it can be detected during the consultation. It is usually triggered by exposure to certain drugs used for general anaesthesia. There are also algorithms that manage this risk (strategies for prevention, avoidance and recovery actions). In the case presented, the anaesthetist can either detect the problem during the consultation, either during the anaesthetic induction with warning signs as muscle stiffness, difficulty of intubation due to spasms, breathing problems, increasing of the patient's body temperature and arrhythmias.

Each vignette was divided into 3 phases. The first one was related to the preoperative phase during which the anaesthetist has to be aware of the patient's physical conditions before to put him to sleep and has to determine an action plan to intervene. This phase corresponded to the anaesthesia consultation and to the risk that a problem could appear.

Risk Assessment in the Anaesthesia Process 219

Fig. 4. The infrequent but predictable problem: the malignant hyperthemia vignette

After a brief description of the case including age, weight, height, the surgery planned and the position required, four questions were asked to the anaesthetist. (1) What information do you need to assess risks linked to this patient? (2) On a scale of 0 to 10 (where 0 means "very easy" and 10 "very difficult") can you estimate the difficulty level for this case? (3) What are the difficulties envisaged? (4) Do you recommend some specific actions for this patient? (Additional strategies: surveillance, specific preparation, etc.). In summary, the anaesthetist was asked to assess risk level associated to this patient and to highlight problems expected.

The second phase was related to the anaesthesia induction. During this phase, the problem really appeared. After reading the description of the situation, three questions were asked to the anaesthetist. (1) Do you need additional information? (2) Are you concerned about some elements of the current patient's state and the surgery? Could you explain the reason? (3) Are there specific actions to implement? In summary, the anaesthetist had to provide details about his situation awareness, to formulate hypotheses and to suggest recovery actions.

The last phase concerned the anaesthesia maintenance. This one is related to the recovery situation. During this third phase, if the anaesthetist did not make anything, the situation will not be anymore recoverable. After reading the vignette, four questions were asked to the anaesthetist. (1) Do you need more information? (2) Do you need to do something in this situation? (3) On a scale of 0 to 10 (where 0 means "very easy" and 10 means "very difficult") can you estimate the level of difficulty of the case you just faced? (4) Do you think you could change the course of this event? If so, could you explain how?

The figure 4 presents an example of vignette about an infrequent but predictable problem, the malignant hyperthermia case.

All the interviews were recorded. Each anaesthetist was faced with a vignette presented in a random order. Thirty-four French anaesthetists participated to our study (average years of experience = 17.12, SD = 11.02). They all came from different departments and different medical structures. They were recruited by mail or by phone. The duration of the entire experience lasted up to 10 minutes or 15 minutes.

Data will be analysed as follow. First we will present the anaesthetists risk assessment (the estimation scale of 0 to 10 at the beginning and at the end of the simulation) of the patient according to the two variables: predictability and frequency. There will be also a qualitative treatment of this assessment to demonstrate if the anaesthetist had taken into account or not difficulties presented in the vignette (*e.g.* in the case of difficult intubation, if the anaesthetist diagnosed it in the first stage, he has adequately understood the problem. It is noted 1. Conversely, if the real risk is not taken into account, it is noted 0). Finally, the risk management will be apprehended by 2 kinds of variables: the quality of the detection, identification and recovery actions proposed by anaesthetists and, the number of hypotheses generated. It tends to answer the following questions: "Does the anaesthetist identify correctly the problem?" and "Does the anaesthetist use the correct recovery actions according to specific algorithms?". It means that if the proposed diagnosis and recovery actions are correct, they are rated 1. In the opposite, if they are false, they are rated 0.

After a brief description of the case including age, weight, height, the surgery planned and the position required, four questions were asked to the anaesthetist. (1) What information do you need to assess risks linked to this patient? (2) On a scale of 0 to 10 (where 0 means "very easy" and 10 "very difficult") can you estimate the difficulty level for this case? (3) What are the difficulties envisaged? (4) Do you recommend some specific actions for this patient? (Additional strategies: surveillance, specific preparation, etc.). In summary, the anaesthetist was asked to assess risk level associated to this patient

The second phase was related to the anaesthesia induction. During this phase, the problem really appeared. After reading the description of the situation, three questions were asked to the anaesthetist. (1) Do you need additional information? (2) Are you concerned about some elements of the current patient's state and the surgery? Could you explain the reason? (3) Are there specific actions to implement? In summary, the anaesthetist had to provide details about his situation awareness, to formulate hypotheses

The last phase concerned the anaesthesia maintenance. This one is related to the recovery situation. During this third phase, if the anaesthetist did not make anything, the situation will not be anymore recoverable. After reading the vignette, four questions were asked to the anaesthetist. (1) Do you need more information? (2) Do you need to do something in this situation? (3) On a scale of 0 to 10 (where 0 means "very easy" and 10 means "very difficult") can you estimate the level of difficulty of the case you just faced? (4) Do you think

The figure 4 presents an example of vignette about an infrequent but predictable problem,

All the interviews were recorded. Each anaesthetist was faced with a vignette presented in a random order. Thirty-four French anaesthetists participated to our study (average years of experience = 17.12, SD = 11.02). They all came from different departments and different medical structures. They were recruited by mail or by phone. The duration of the entire

Data will be analysed as follow. First we will present the anaesthetists risk assessment (the estimation scale of 0 to 10 at the beginning and at the end of the simulation) of the patient according to the two variables: predictability and frequency. There will be also a qualitative treatment of this assessment to demonstrate if the anaesthetist had taken into account or not difficulties presented in the vignette (*e.g.* in the case of difficult intubation, if the anaesthetist diagnosed it in the first stage, he has adequately understood the problem. It is noted 1. Conversely, if the real risk is not taken into account, it is noted 0). Finally, the risk management will be apprehended by 2 kinds of variables: the quality of the detection, identification and recovery actions proposed by anaesthetists and, the number of hypotheses generated. It tends to answer the following questions: "Does the anaesthetist identify correctly the problem?" and "Does the anaesthetist use the correct recovery actions according to specific algorithms?". It means that if the proposed diagnosis and recovery actions are correct, they are rated 1. In the opposite, if they are

you could change the course of this event? If so, could you explain how?

and to highlight problems expected.

and to suggest recovery actions.

the malignant hyperthermia case.

false, they are rated 0.

experience lasted up to 10 minutes or 15 minutes.

Fig. 4. The infrequent but predictable problem: the malignant hyperthemia vignette

Risk Assessment in the Anaesthesia Process 221

Fig. 5. Percentage of risks correctly identified according to their predictability and their

In average, anaesthetists have advanced 2.24 hypotheses before giving the correct diagnosis (SD=1.16). As shown in the table 5, whatever the frequency or the predictability, there is no

Predictable 2 (SD=0.93) 2.89 (SD=1.36) 2.47 (SD=1.23) Unpredictable 1.89 (SD=0.93) 2.13 (SD=1.25) 2 (SD=1.06)

Table 5. Means of hypotheses generated according to the predictability and the frequency

Concerning the recovery, three results are interesting to quote. First, 3 anaesthetists did not attempt to recover the problem. Second, even if 97% of the anaesthetists identified the risks, only 67.6% of them were able to recover the problem presented. As shown on the figure 6, anaesthetists recovered correctly 94,44% of frequent risks and more often when the risk was predictable (50%). Finally, the less managed risk is the infrequent one (40,97% of anaesthetists were able to provide adequate recovery actions) regardless of its predictability

Mean (SD) 1.94 (SD=0.9) 2.53 (SD=1.33)

Frequent Infrequent Mean (SD)

frequency

(SD)

(22,22%) or not (18,75%).

difference in the number of hypotheses.

### **5.2 Results**

We will present in this chapter only the results concerning risk assessment before and during the anaesthesia.

### **5.2.1 Risk assessment during the patient's consultation**

The table 4 shows the average level of difficulty estimated by the anaesthetists according to the frequency and the predictability. As expected, a predictable event is always considered as more risky (M=4.06, SD=1.52) than if it was not predictable (M=2.18, SD=1.19). In fact, the t-test shows that this difference is extremely statistically significant (t(32)=4.02, *p<0.0005*, *d*=0.468). In general, the same is observed for the frequency: a common risk is assessed as more difficult (M=3.35, SD=1.9) than an infrequent one (M=2.88, SD=1.36). But by conventional criteria, this difference is not statistically significant according to the t-test (t(32)=0.83, *NS*, *d*=0.57).


Table 4. Average level of difficulty according to the frequency and the predictability

Specifically, two major differences appear. Predictable risks are considered as more difficult when they are frequent (M=4.88, SD=1.25) than when they are infrequent (M=3.33, SD=1.41). This difference is statically significant by the t-test applied (t(15)=2.38, *p<0.05*, *d*=*0.65*). The same is observed concerning the frequent risk. They are considered as more difficult when they are predictable (M=4.88, SD=1.25) than the reverse (M=2, SD=1.22). By conventional criteria, this difference is considered to be extremely statistically significant (t(15)=4.8, *p<0.0005*, *d*=0.6).

Regarding the quality of their risk assessments, only 35.3 % of the anaesthetists correctly assess the risk they face and mainly when the risk is frequent and predictable (7 anaesthetists have a correct assessment against 1 who is wrong). Typically, risks identified are mostly distorted by other problems than those presented in vignettes.

#### **5.2.2 Risk assessment during the surgery**

Although risks are not correctly identified at the consultation, when problem arises, the majority of the anaesthetists (97%) identified that there was a problem but only 79.4% of the anaesthetists identified correctly the problem they faced.

In general, all the risks are correctly identified. As shown on the figure 5, frequent risks are always identified correctly (100%). Anaesthetists frequently recognized predictable and unpredictable risks (respectively, 83,33% and 75%). The most difficult risk to identify is the one that is infrequent (58,33%) and mainly when it could not be foreseen during the consultation (25%).

We will present in this chapter only the results concerning risk assessment before and

The table 4 shows the average level of difficulty estimated by the anaesthetists according to the frequency and the predictability. As expected, a predictable event is always considered as more risky (M=4.06, SD=1.52) than if it was not predictable (M=2.18, SD=1.19). In fact, the t-test shows that this difference is extremely statistically significant (t(32)=4.02, *p<0.0005*, *d*=0.468). In general, the same is observed for the frequency: a common risk is assessed as more difficult (M=3.35, SD=1.9) than an infrequent one (M=2.88, SD=1.36). But by conventional criteria, this difference is not statistically significant according to the t-test

Frequent Infrequent **General mean (SD)** 

Predictable 4.88 (SD=1.25) 3.33 (SD=1.41) 4.06 (SD=1.52) Unpredictable 2 (SD=1.22) 2.38 (SD=1.19) 2.18 (SD=1.19)

Specifically, two major differences appear. Predictable risks are considered as more difficult when they are frequent (M=4.88, SD=1.25) than when they are infrequent (M=3.33, SD=1.41). This difference is statically significant by the t-test applied (t(15)=2.38, *p<0.05*, *d*=*0.65*). The same is observed concerning the frequent risk. They are considered as more difficult when they are predictable (M=4.88, SD=1.25) than the reverse (M=2, SD=1.22). By conventional criteria, this difference is considered to be extremely statistically significant

Regarding the quality of their risk assessments, only 35.3 % of the anaesthetists correctly assess the risk they face and mainly when the risk is frequent and predictable (7 anaesthetists have a correct assessment against 1 who is wrong). Typically, risks identified

Although risks are not correctly identified at the consultation, when problem arises, the majority of the anaesthetists (97%) identified that there was a problem but only 79.4% of the

In general, all the risks are correctly identified. As shown on the figure 5, frequent risks are always identified correctly (100%). Anaesthetists frequently recognized predictable and unpredictable risks (respectively, 83,33% and 75%). The most difficult risk to identify is the one that is infrequent (58,33%) and mainly when it could not be foreseen during the

Table 4. Average level of difficulty according to the frequency and the predictability

**5.2.1 Risk assessment during the patient's consultation** 

**General Mean (SD**) 3.35 (SD=1.9) 2.88 (SD=1.36)

are mostly distorted by other problems than those presented in vignettes.

**5.2 Results** 

during the anaesthesia.

(t(32)=0.83, *NS*, *d*=0.57).

(t(15)=4.8, *p<0.0005*, *d*=0.6).

consultation (25%).

**5.2.2 Risk assessment during the surgery** 

anaesthetists identified correctly the problem they faced.

Fig. 5. Percentage of risks correctly identified according to their predictability and their frequency

In average, anaesthetists have advanced 2.24 hypotheses before giving the correct diagnosis (SD=1.16). As shown in the table 5, whatever the frequency or the predictability, there is no difference in the number of hypotheses.


Table 5. Means of hypotheses generated according to the predictability and the frequency (SD)

Concerning the recovery, three results are interesting to quote. First, 3 anaesthetists did not attempt to recover the problem. Second, even if 97% of the anaesthetists identified the risks, only 67.6% of them were able to recover the problem presented. As shown on the figure 6, anaesthetists recovered correctly 94,44% of frequent risks and more often when the risk was predictable (50%). Finally, the less managed risk is the infrequent one (40,97% of anaesthetists were able to provide adequate recovery actions) regardless of its predictability (22,22%) or not (18,75%).

Risk Assessment in the Anaesthesia Process 223

surgery and anaesthesia risk, even if their verbalizations show that they use the patient and the surgery criteria with almost the same frequency. Both of them use more information according to the files complexity. Finally, the two groups of anaesthetists use some warning

Through this study, our results show that the more experimented anaesthetists build more functional representations because they analyse the situation more broadly. This confirms the results of Cellier et al. (1997) and Hoc (1989). In other words, experimented anaesthetists consider two levels of risk (the patient and the surgery) even if they use more often the criteria related to the patient. That should enable them to anticipate potential incidents

Our results also confirm that two variables can influence the consultation (Anceaux et al., 2001, 2002, 2005): the level of experience and the case complexity. Finally, regarding information used by anaesthetists (all experience level combined) to assess a file, they all

All these elements show that the risk assessment during the consultation leads to a schematic representation that can be then specified later during the surgery (Hoc, 1987).

The second research permits to answer the second question (How important are the risk frequency and the risk predictability in the assessment and the management in real time?). Our results indicate that the risk predictability increases the perceived difficulty level associated with it. Moreover this perceived difficulty is higher when the risk is both frequent and predictable. The risk frequency seems to be important for the identification

Overall, few anaesthetists assess the problem correctly at the end of the consultation (for recall: only 35.9% of anaesthetists assess the real risk). Anaesthetists correctly assess only frequent and predictable risks at this early stage. Concerning the other risks, the anaesthetist's representation is distorted by other problems. Even if the risk is not really perceived during the consultation, the identification is almost correct for all the anaesthetists. We also find that anaesthetists identified and recovered less correctly the

Finally our results point out that anaesthetists suggest that anaesthetists propose several hypotheses before reaching the right result whatever the nature of the risk according to its

These results confirm epidemiological and psychological studies previously cited. In fact, several explanations can be given to our results. Firstly, previsability permits to assess (consultation) and to identify (surgery) correctly the risks. Secondly, as Cooper's results (et al., 1982) showed, anaesthetists not taken into account infrequent risks. These ones are not correctly assessed, identified and recovered because there is a misunderstanding. Indeed the majority of anaesthetists faced with a infrequent risk explained that they have been rarely confronted with this kind of risk and most often through simulations. Thus, when the patient presented signs of this problem, they tended to minimize the facts (Amalberti, et al., 2005). Moreover, their representation seems distorted by highlighting other problems (points for consideration, Xiao, et al., 1997). Finally, frequent and predictable risks are most

point out specific problems. This last point confirms Xiao's results (et al., 1997).

flags to underline some specific problems.

associated with the patient and the surgery.

during the surgery and also for the recovery actions.

infrequent risks.

frequency and its predictability.

Fig. 6. Percentage of correct recovery actions according to the predicatbility and the frequency

#### **5.3 Brief discussion of the second study**

In this second study, we wanted to see how anaesthetists assessed risks before and during the surgery. We were interested by two kinds of variables: the predictability and the occurrence frequency. Our results showed that anaesthetists assessed problems as more risky when they can anticipate it than when they are unforeseeable. However, few anaesthetists were able to correctly assess the risks during the consultation. In fact, their representation were often distorted by others problems. It means that they put a lot of warning flags for each case. One variable seem relevant in the risk assessment and management: the frequency. A frequent risk is always correctly identified and recovered.

In the following section, all the results will be explained regarding the theories used.
