Handlers’ Representations on Therapy Dogs’ Welfare

Simple Summary Most research about Animal-Assisted Interventions (AAI) has focused on the benefits on human health. In contrast, very little has been made on the impact of this work on therapy dogs, although it is part of the ethics of the practice to ensure their welfare. This study aimed to contribute to the knowledge on the welfare of therapy dogs by interviewing 111 handlers through an online questionnaire. The qualitative assessment of handlers’ representations underlined that the welfare of therapy dogs is multidimensional and can be impacted by various variables. Its consideration is important for the quality and safety of the sessions, both for the dog and for the beneficiaries involved. Handlers have a central role in the welfare of their therapy dog and must be trained on stress-related behaviors. Research needs to focus on the impact of interactions on therapy dogs. Abstract While research on the benefits of animal-assisted interventions is beginning to build a significant body of work, studies on the well-being of therapy dogs are still in their infancy. Since handlers are the people responsible for their therapy dog’s welfare, we interviewed 111 French handlers through an online questionnaire. Our results underlined that (i) therapy dogs’ welfare is multidimensional when physical and psychological welfare, a balance between work and dog life and the settings and interactions of sessions are all taken into consideration. (ii) The response of our handlers emphasized that considering therapy dog welfare is important for the quality and safety of AAI. (iii) Three categories of risks factors were highlighted: the spatio-temporal framework (planning and environment), the interactions with beneficiaries and the handler themselves. It is particularly important that handlers talk about the negative impact of interactions with beneficiaries since they are at the heart of AAI, however there are few studies focusing on interactions as a stressor for dogs in this practice. Moreover, since there is a potential for positive bias in the handlers’ representations, it is important that they be trained to identify and manage the stress in their dogs. Future research is particularly needed on the impact of interactions during sessions on therapy dog welfare.


Introduction
During the last 5-10 years, the research interest in assessing the influence of human proximity and/or behavior towards animals on animal welfare has increased [1]. Currently, studies about animal welfare aim to understand the point of view of how animals feel about interactions with humans [2], and this should include working animals. Animal-Assisted Interventions (AAI) are defined as "a goal oriented and structured intervention that intentionally includes or incorporates animals in health, education and human services (e.g., social work) for the purpose of therapeutic gains in humans. It involves people with knowledge of the people and animals involved" [3]. These practices necessarily include an intervening dyad composed of a handler and his/her animal to benefit to a third party (the beneficiary); which is, in fact, different from assistance dogs where the benefits of the

Sample and Recruitment
This study is included in a doctoral project investigating the French practice of AAI. We questioned French handlers in AAI who agreed to answer to an online questionnaire shared on AAI-specialized social media accounts and sent by emails from April 2018 to May 2019. Our inclusion and exclusion criteria were very limited since we wanted to have a real vision of the field. We aimed to include workers from different fields of animal mediation (recreational, therapeutic, social, etc.). We have therefore included all handlers currently working in AAI with at least one dog and wishing to participate in our research on a voluntary basis, regardless of their professional background or affiliation to AAI associations. We focused mainly on canine-assisted interventions as the dog is the most common animal species in AAI [56][57][58][59]. The questionnaire covered several aspects of working with dogs in a therapeutic setting; this article focuses on identifying risks to the welfare of therapy dogs, based on an analysis of handlers' representations in this regard. See Mignot [18] for the representations of French handlers on their AAI practices and Mignot [26] for the characteristics of the French practice of AAI in relation to the IAHAIO model.

Ethics
The Ethics Committee of the UFR SPSE (Psychological and Educational Sciences University Paris Nanterre) approved this investigation, protocol code is 04-n • 5. Before accessing the questionnaire, handlers were required to complete a consent form that included an explanation of the study framework, objectives and the research ethics features.
Signing this consent form guarantees the confidentiality of their responses, the possibility of interrupting the research, respect for their integrity and their rights in accordance with the research ethics. The collection, processing and storage of personal data complied with the rules laid down by the General Data Protection Regulation [60].

Data Collection
A five-section questionnaire was constructed based on a literature review [3,23,36,[61][62][63][64][65] and questioning raised during informal interviews and internship. As mentioned before, for this article, we focused our attention on 10 items (Table 1; complete questionnaire in Mignot [48]). Table 1. Description of the section "dog's welfare" of the questionnaire used in this study. Complete description of the questionnaire as a supplement data in Mignot [48].

Questions Type
Generalities on therapy dog welfare How would you define dog welfare in AAI? Open question Do you think that the dog's well-being plays on a good session? If so, why?

YES/NO Open question
The following is a cross-reference of the main guidelines found in the studies/associations. Indicate for each one whether you apply it: never, rarely, sometimes, often, all the time. 5 The first items concerned general questions about handlers' representations on therapy dog welfare: the definition of dog welfare in AAI, its impact on the quality of a session, the handlers' application of the main guidelines concerning therapy dog welfare, the limitations encountered to respect therapy dog welfare and their representations about the benefits for therapy dogs to work in AAI. The second part concerned the risk factors identified that could affect therapy dog welfare: the parameters influencing therapy dog welfare, the factors that stress their dogs, what handlers put in place to ensure the welfare of their therapy dog and also how they observe stress and pleasure in their dogs.

Analyses
A descriptive analysis was performed by calculating means and frequencies for numerous and categorical variables with the software GraphPad Prism 9 TM . We chose to categorize the answers to the open questions into themes: the parameters affecting therapy dog welfare, the risks factors identified, and the way handlers observed stress or pleasure in their therapy dogs.
A qualitative descriptive method was selected for the five open questions: the definition of dog welfare in AAI, the importance of therapy dog welfare for a good session, the benefits for the dogs working in AAI, what handlers put in place for their dog's welfare and the limitations encountered that may prevent the recommendations from being respected. We used a phenomenological method because it "describes the meaning for several individuals of their lived experiences of a concept or a phenomenon; describing what all participants have in common as they experience a phenomenon" [66]. Like Firmin et al. [61], we used an open coding strategy with a line-by-line analysis approach to explore data for recurring ideas [66][67][68], and developed clusters of meaning into themes. Only the principal investigator conducted the qualitative analysis and did it twice 3 months apart to confirm that the same themes emerged from the data. Also, the relevance of the themes and the mapping that emerged from the initial coding were discussed with the other researchers. We then counted the number of themes mentioned by each handler (a response could concern several themes) and finally calculated the percentages of each theme on the total number of themes cited for each question. These analyses respect the 15-point checklist of criteria for thematic analysis underlined by Braun and Clark [69]. Since it is a study that aims to show what stands out the most in the handlers' representations, we presented the major themes for each question, all themes are presented in the tables.
Handlers were able to answer for one or two therapy dogs. As a result, 57 handlers responded for only one dog and 54 for two dogs. From a total of 165 dogs, 63.63% were females (n = 93) of which 72.04% were sterilized (n = 67); 58.33% of males were neutered (n = 42). Their mean age was 5 years (±0.26) and they began AAI at 24.02 months (±2.05). Breeds were mostly retrievers (32.92%; n = 53), with only golden retrievers and labrador retrievers, shepherd dogs (24.22%; n = 39) with mostly Australian shepherd, Shetland shepherds and border collier and companion and toy dogs (13.66%; n = 22) with mainly bulldogs, cavalier king charles spaniels and poodle. Almost half of the dogs were certified

Handlers' Representations of Therapy Dog Welfare
In this section, we grouped handlers' answers to open questions regarding their definition of therapy dog welfare, how a dog's welfare can impact a "good session" and the benefits of working in AAI for their dogs. All themes are presented in the tables.

Definition of Dog Welfare in AAI
For the question about their definition of therapy dog welfare, handlers mentioned one to five themes in their answers. We identified seven major themes for the definition of dog welfare in AAI ( Table 2). The three major themes most cited by handlers were the psychological welfare of dogs, followed by the respect of the needs and rhythm of the therapy dog, and the consideration of therapy dog as an individual. Regarding the impact of dog welfare on a "good session", 108 handlers answered "yes" to this question and then answered the open question regarding the reasons they thought this. 97 handlers answered to this question and mentioned one to four themes in their answers. We identified eight major themes for the impact of dog's welfare on the session of AAI ( Table 3). The three major themes most cited by handlers were the quality of interactions, followed by the attentiveness of the dog during session and the emotion contagion.

Benefits for the Dogs to Work in AAI
For the question about the benefits of working in AAI for dogs, handlers mentioned one to five themes in their answers. We identified seven major themes for the benefits for the dogs to work in AAI ( Table 4). The three themes most cited by handlers were the positive interactions with humans, followed by being with their human and the cognitive and physical stimulations. Table 3. This table represents the answers to the question "Do you think that the dog's well-being plays on a "good" session? If so, why?". n = 97 complete answers. Total of 133 terms coded grouped into 8 themes; presented in descending order.

Themes
Description n %

Parameters Influencing Therapy Dog Welfare
In this section, we grouped questions on what handlers put in place to ensure their therapy dog welfare, their application of major recommendations and the limitations encountered when attempting to apply these recommendations.

Handlers' Actions to Respect Therapy Dog Welfare
For the question about handlers' actions to respect their therapy dog welfare, handlers mentioned one to six themes in their answers. We identified eight themes in their answers  Table 5). The three major themes cited by handlers were the times to unwind, their active role during session and the management of environmental equipment.

Recommendations
Regarding handlers' application of the main recommendations concerning therapy dog welfare, (Table 6), the highest mean scores were for the cessation of sessions when the dog is stressed, the recognition of stress signals, the use of positive training method, while the lowest mean scores were for "time-outs" during sessions, the time for adaptation before sessions and the cessation of sessions before appearance of stress signals. Table 6. This table represents the answers to the question "The following is a cross-reference of the main guidelines found in the studies/associations. Indicate for each one whether you apply it never, rarely, sometimes, often, all the time »; n = 111 complete answers.

Recommendations
Mean SEM

Limitations to the Application of These Recommendations
After answering the precedent question, handlers were able to elaborate on the limitations they encountered during the practice with regards to respecting the precedent recommendations. Handlers mentioned one to three themes. We identified seven major themes for the limitations to the application of major recommendations ( Table 7). The three major themes most cited by handlers were the environment, followed by the organization and the expectations of the institutions. Table 7. This table represents the answers to the question "What limitations do you encounter that prevent you from following these recommendations?". n = 57 complete answers. Total of 72 terms coded grouped into 7 themes; presented in descending order.

Risk Factors for Dog's Stress
The interrogation about the risk factors identified by handlers for therapy dogs' welfare was divided into two questions: "According to you, what are the three parameters that most influence the well-being of the dog in AAI?" and "With regard to each dog, what do you think stresses him/her?" We categorized all factors into 14 categories (see Table 8) and grouped both answers into one analysis.
For the questions about the risks factors for dogs' stress and welfare in AAI, we analyzed 435 factors ( Figure 1). The factors most cited were the beneficiary's emotional state, followed by the frame management and inappropriate behaviors. By grouping our data into general categories, we underlined four major factors: the beneficiaries, the environment, the handler (19.54%; n = 85) and the dog (6.90%; n = 30). Table 8. Description of the categories selected for the analysis of risks factors for therapy dog welfare.

Categories Description
Interactions Self-monitoring Management of the dog's frustration that it is linked to poorly formulated requests, inappropriate gestures, etc.

Shouting of the beneficiaries, incomprehensible noises
Type of interactions Abrupt/inappropriate gestures; positive vs. negative interactions

Restriction
It can be unwanted/unappreciated physical contact, intrusion into personal space, but also being surrounded by too many people. This notion thus includes the non-respect of one's limits, everything that is an obligation to interact but also to work.Not obliged to, not forced, freedom, being surrounded, hugs too tight, being on the knees

Beneficiary's emotional state
The mood of the group, agitation, restlessness, anger directed or not directed at the dog, excitability; also symptoms of pathology

Dog-handler relationship
Everything that is positive between them: trusting relationship but also feeling secure and listening to the dog, reading the dog, attention but also knowledge; Confidence in one's human being, feeling of security For the questions about the risks factors for dogs' stress and welfare in AAI, we analyzed 435 factors (Figure 1). The factors most cited were the beneficiary's emotional state, followed by the frame management and inappropriate behaviors. By grouping our data into general categories, we underlined four major factors: the beneficiaries, the environment, the handler (19.54%; n = 85) and the dog (6.90%; n = 30).

Dog's Communication of His/Her Limits
For the question about the way handlers perceived that their dogs communicate their limits, we coded the answers into six categories (Table 9). Handlers mostly noticed the limits of their dogs by the fact that their dog goes away, followed by the calming signals and the avoidance of interactions with beneficiaries. Handlers also mentioned the contact seeking with them that can be physical or visual. Finally, handlers mentioned the decreased in attention (7.02%; n = 12) and the agitation (5.26%; n = 9).

Dog's Communication of His/Her Limits
For the question about the way handlers perceived that their dogs communicate their limits, we coded the answers into six categories (Table 9). Handlers mostly noticed the limits of their dogs by the fact that their dog goes away, followed by the calming signals and the avoidance of interactions with beneficiaries. Handlers also mentioned the contact seeking with them that can be physical or visual. Finally, handlers mentioned the decreased in attention (7.02%; n = 12) and the agitation (5.26%; n = 9).

Signs Showing Pleasure in Dogs
For the question about how handlers perceive pleasure in their dogs in AAI, we coded 168 terms. Five major themes were identified in handlers' answers (Table 10). The search for interactions was the most represented, followed by the pleasure signals, which included cheerfulness, tail wagging, a playful attitude, jumping and barking. Handlers also noticed the motivation to go to work with the excitement at seeing work -related paraphernalia or running to the care facility etc. Finally, they mentioned active participation and a relaxed attitude.

Discussion
The aim of this study was to interrogate handlers in order to understand how they grasp the welfare of their therapy dog (i) to highlight the importance of addressing the well-being of therapy dogs both for dogs and beneficiaries as well as (ii) to point out the environmental and interactional risk factors to this well-being. To this end, we considered handlers' representations as it may allow the identification of risk factors for the well-being of therapy dogs in complementarity to the results of studies using physiological indicators and/or behavioral assessment.
We assume that our sample is representative of the French practice of AAI because handlers were dispatched in various localizations in France. In addition, as there is no official data about the number of handlers who practice AAI in France, this study can be considered as a pilot for further investigations. Our sample represented a variety of French practices in AAI regarding the characteristics of the human-dog teams. Indeed, handlers were mostly trained in AAI whereas it is not mandatory in France [23,26,70]. It is important to note that they were trained in different structures, which can impact their representations and management of their therapy dog welfare [48]. Only 37.84% of them undertaken some initial training within the animal field, which was represented by various professions (i.e., veterinarian and breeder). Therapy dogs also represented various profiles regarding their gender, age, starting age in AAI and breeds. Beyond the heterogeneity of human-dog teams that corresponds with the heterogeneity of AAI in France, this highlights that any action for the well-being of therapy dogs must target the different characteristics of handler and dogs involved in AAI.
Since we questioned the welfare of therapy dogs in different ways, we chose to present handlers' general representations on the definition of therapy dog welfare first, then their representations about the importance of considering the welfare of therapy dogs, and finally the risk factors identified by them coupled with how they identify stress in their dogs. Once again, our objective was to show what stands out most in their representations in order to find areas for study and intervention.

Generalities on Dog Welfare in AAI
Answers about the definition of therapy dog welfare varied from one handler to another, but their answers converged on the importance of a balance between worktime and the dog's life, as well as their psychological and physical welfare and the considerations of therapy dogs as individuals with specific needs and preferences.
The first theme cited by handlers concerned the psychological welfare of therapy dogs (32.37%) with the absence of negative emotions (including mostly stress) and the presence of positive emotions during sessions. It corresponds to the definition of animal welfare that included the absence of stress [71][72][73][74] and the presence of positive emotions; that has been included more recently in the assessment of animal welfare [11,75,76]. On a smaller percentage, handlers cited the "physical" welfare of therapy dogs (7.69%) that included mostly the absence of injury during sessions but also the respect of the specific needs of the canine species. Some handlers mentioned, on the borderline of psychological and physical well-being, that it is important for therapy dogs to participate without coercion either through coercive tools or mental restraints [11]. To guarantee the absence of stress during sessions, it is important for handlers to be able to recognize stress signals in dogs. Indeed, the active role of the handler was the fourth theme (12.56%) cited by handlers, notably with noticing stress signals and removing the dog from interactions. However, this can be difficult when handlers are not trained and may also be biased because they want their dog to enjoy accompanying them to work. This positive bias has been highlighted in the article of Zenithson [10] and confirmed in the study of Haubenhofer and Kirchengast [35] where handlers used positive words (satisfied, relaxed, happy mood) to describe how their dogs feel after a AAI session.
The second theme mentioned by handlers in their definition of their dog's welfare in AAI was the fact that the dog should have a dog's life outside of work (21.74%). This theme mostly regrouped time away from work to unwind, blow off steam or rest. It underlines that, for the handlers, it is important that their dog also has a "dog's life" outside of the work and that the needs of the canine species are taken care of (walking, playing, meeting fellow canines). Developing upon these thoughts, the evocation of the importance of a balance between work and time-off to unwind suggests that handlers consider that working in AAI can be stressful for their dogs and that they must decompress. Indeed, some studies [34][35][36]77] have underlined a higher level of stress on work days than on a day "off"; supporting handlers' perceptions.
The third theme mentioned by handlers was to consider their dogs as individuals who have their own limits and preferences and the ability make choices (14.78%). This means that what suits one dog will not necessarily suit another and that an adaptation of the working conditions is necessary. This individuality must also be considered in relation to the characteristics of the setting and all the variables with which the dog must cope [22]. Furthermore, they evoked the freedom of their therapy dog that included not only the possibility for them to make choices such as whether or not to come to the sessions and participate but also in the management of the space during sessions (i.e., having a place to rest). Therefore, handlers also evoked a mutual comprehension with their dogs and the necessary adjustment during sessions based on their dog's signals.
Finally, since the goal of this study was to get a complete picture of the welfare of therapy dogs, we also asked handlers about the perceived benefits of AAI for their dogs. It seems that the handlers see the work in AAI as something positive for their dogs because it allows them to not be alone at home, to experience positive interactions with other humans, to become familiar with different environments and to have their cognitive needs met. However, it is important to note that it is normal for handlers to see benefits for their dogs, but this has to be put in perspective with each dog's personality and the way the sessions are conducted. For instance, handlers mentioned being in the presence of their humans and/or other humans as a primary benefit for therapy dogs, which could be discussed because interactions between human and dogs could be positives for both [78,79] but also stressful [27][28][29]43,80].Handlers also cited the physical and cognitive benefits for therapy dogs with training and stimulation before and during sessions. Indeed, the behavioral training for AAI provide positive mental stimulation [31]. Consequently, some handlers made the comparison with the classic "pet dog" that stays alone all day, which can indeed affect such a dog's welfare [81].
To summarize, handlers' answers underline the fact that dog welfare in AAI involves common prerequisites for all therapy dogs (time-off, psychological and physical integrity) but also specificities linked to each dog. Consequently, it seems important to consider the balance between the costs and benefits of each session for each dog.

Importance to Take into Account Dog's Welfare in AAI
The interviews with the handlers about how therapy dog welfare can affect a session of AAI underlined three main points: the influence of therapy dog welfare on the quality of work in AAI, the risks for beneficiaries of interacting with a stressed therapy dog and the risks for the therapy dog him/herself. The first point mentioned by handlers concerned the influence of therapy dog welfare on the quality of the session, which was mostly represented by the quality of interactions, the attentiveness and the motivation of the dog In the handlers' discourses, we found the general idea that a dog in discomfort will be focused on managing his stress rather than on meeting the beneficiaries and listening to their requests. Consequently, they evoked the avoidance of interactions as a sign that their dog is in discomfort, and the search for interactions and the motivation to go to work as a sign of pleasure. This links with the idea that the well-being of the therapy dogs plays on the quality of their work. Indeed, the main expectation on therapy dogs is to initiate interaction and bonding with the beneficiaries [12,[82][83][84]. Handlers also mentioned the possible impact of negative therapy dog welfare on the safety of the beneficiaries with the risk of agitation or biting in a stressed dog. Indeed, a depiction of the gestures that any dog will give in response to an escalation of perceived stress and threat has been illustrated by the Canine ladder of aggression of Sheperd [85]; from yawning, blinking, and nose licking to biting. In the behaviors identified by handlers as a communication of limits, handlers mentioned appeasement behaviors that have been identified as signs of stress in the literature [86,87] and are at the lower end of the aggression scale. These behaviors must therefore be perceived by handlers and considered to guaranty the safety of beneficiaries. Some handlers also mentioned that the stress of dogs can be "contagious" for beneficiaries if they feel the dog's discomfort. Finally, handlers mentioned the importance of considering a therapy dog's welfare because due to the risks for the dogs in relation to the ethics in AAI and a parallel with the burnout of caregivers after chronic stress. It is interesting that it was not the prior theme mentioned by handlers about the impact of dog welfare on a good session, evoking that a session can be good even if the dog is in poor welfare.

Risks Factors
We grouped what handlers cited as stress factors and/or parameters influencing therapy dog welfare into four categories: the interactions with beneficiaries, the space-time framework, the handler's responsibility, and the dogs' characteristics. We discuss only the first three because as we don't have sufficient details regarding the characteristics of dogs as risks factors for their welfare, the dog's emotional state during sessions and the dog's skills more generally.

Interactions
The influence of interactions represented 43.68% of identified stress factors by handlers. As mentioned before, interactions between human and animals could be a benefit for dogs, but it can also be stressful [27][28][29]43,80]. Interrogated handlers evoked three types of human behaviors that can affect therapy dog's welfare Firstly, handlers evoked human behaviors that are not directly threatening to the therapy dogs but can be stressful such as restrictive behaviors and behaviors that require the dog to control him/herself. Indeed, being approached, petted on the head, kissed or hugged by strangers can play a role in the comfort level of therapy dogs [12,88], whereas these situations are common in AAI [89]. Secondly, handlers mentioned the presence of behaviors that can be perceived as stressful and/or painful by therapy dogs such as agitated or even aggressive behaviors, which could be linked to the beneficiaries' pathologies. Even if these behaviors are not directly directed toward the dog, they can be interpreted as a threat by the dog. Finally, handlers mentioned inappropriate behaviors directed towards the therapy dog such as abrupt gestures, rough handling and direct interactions. The mistreatment of therapy dogs is rarely mentioned in studies [12] or by handlers [10] although it is a reality in the field. As Beck & Katcher [90] point out, research needs to identify populations or situations where contact with therapy animals may be problematic or inappropriate for both the animals and the people involved. Therefore, this may call into question the idea of handlers that working in AAI represents a social benefit for therapy dogs. The positive or negative impact of the beneficiary-dog interactions must then be put into perspective with the type of interactions that take place in the session. In addition, in our study the presence of a referent responsible for the beneficiaries in the AAI was random, which puts the handler in the position of being the only one responsible for these interactions. It is important that handlers work in collaboration with a referent that regulates beneficiary behaviors and who is also trained to recognize a dog's stress.

Space-Time Framework
The space-time framework represented 28.89% of risks factors cited by handlers. They mentioned the length of work (frequency and duration of sessions) and the environment (size of the room, number of beneficiaries) as important parameters that influence therapy dog welfare. It is congruent with the literature that mentioned an effect of the duration of sessions [13,34,36], the frequency of sessions [13,34,40,41] and the number of beneficiaries [40] on a therapy dog's welfare. We can also include in the framework of AAI the mention of a balance between work and time-off that should be considered to be part of the dog's work schedule. Finally, the medical staff and the institutions were pointed out by handlers as a limit for dog welfare. Indeed, they evoked the difficulty of respecting the needs and choices of the dog the expectations and organization of institutions. This underlines the fact that these institutions do not always seem to understand that AAI is a practice with living beings and that expecting a certain "profitability" from the animal is inappropriate. Indeed, some authors mentioned that the economic interests could lead to a conflict of interest if they outweigh the welfare status of the animal [10,12]. It is therefore important that institutions understand the importance of considering therapy dog welfare and that there is no pressure to be cost-effective with AAI. On the other hand, handlers must be trained to be able to meet the expected objectives, even without the active presence of their animal.

Handlers' Responsibility
Handlers considered themselves as an important factor (19.54%) that can affect their dogs' welfare because they are the ultimate gatekeeper of animal welfare in AAI [58]. Indeed, all factors mentioned above are under the responsibility of handlers (except the influence of the institution in which they work). Therefore, they evoked a role in prevention by constructing a framework that respects the dog's needs, their preferences and limits; and by supervising the interactions with beneficiaries during sessions. For the framework, the handler is not only the responsible of the planning but also for the management of the space. For instance, handlers can organize the space of the session with a safe place for their dogs to escape from interactions if they are stressed or tired [15,91]. As mentioned before, handlers must adjust to their dogs, not force interactions with the beneficiaries and leave a margin of control to their dogs. During interactions, handlers have a role of mediator between their dog and the beneficiaries and must intervene in the early stages of negative arousal [12] and give breaks before discomfort appears [10]. In our "recommendations" section, the highest mean score was for recognizing signals of stress and subsequently stopping sessions when the dog is already stressed, but a low mean score for stopping sessions before the stress appears. This can be due to the variation of knowledge on animal behavior between handlers that can impact their perceptions [91]. As we mentioned before, handlers having training in AAI from various structures leads to uneven knowledge on canine behavior and welfare. Moreover, the perception of stress in their dogs can vary between humans [92], especially when speaking about subtle signals of stress. In our sample, handlers underlined four types of signals indicating stress in their dogs: the avoidance of interactions, appeasement signals, agitation and physical or visual contact seeking. It underlined the variety of signs of stress in dogs, which is congruent with the literature [93,94]. Finally, regarding their relationships, handlers mentioned the influence of their own emotional state on their dog's stress. It has been mentioned recently in a paper that handlers may suffer from compassion fatigue, which can compromise their own welfare but also the ability to lead AAI [95].
Therefore, a lack of training in dog behavior can be a risk factor for a therapy dog's welfare; they therefore must be accompanied by professionals such as dog trainers or specialized ethologists.

Limitations
The representativeness of our sample could be questioned because the handlers who chose to answer to our questionnaire were concerned by their practice in AAI and the selection of their mediation dog. In addition, we chose to study the representations of the handlers as an approach to understanding the well-being of the therapy dogs, however it is possible that their answers are biased by a desire that the work pleases their animal as well as a social desirability bias. By looking at the social desirability bias, we can think that, since the questionnaire was entirely anonymous, the handlers did their best to transcribe the reality of their AAI practices. The combination of interviewing handlers and directly observing their practices would have countered these biases. Therefore, these data should not be understood as truth but as possibilities for exploration for future research. Future research should focus on behavioral analyses of various animal-mediated devices with a focus on interspecific interactions and their potential for stressing dogs.

Conclusions
The interview of handlers about therapy dogs welfare underlined their concerns about their dogs' well-being. They have ideas about the factors that can positively or negatively affect well-being and they adjust their practices according to these ideas. Handlers' representations about therapy dogs welfare highlight the need to consider it as multidimensional with all the variables it encompasses such as the spatio-temporal framework (planning and working area), the individuality of each dog, the beneficiaries and interactions, and the handler's own knowledge and skills. This underlines the importance of thinking about therapy dog welfare within a context that encompasses all of the variables that can affect it. Furthermore, handlers' answers emphasized the importance to integrate the concept of one welfare in the considerations and studies about AAI [96][97][98]. Developing upon these thoughts, it is important to note that there may be a bias in the handlers' representations related to the fact that they want their dog to enjoy working with them. This bias must be considered particularly in relation to the interactions with beneficiaries but also the choice of the setting in which the dog works. It is important that the handler is trained to recognize the signs of stress in their dog and not leave them in an uncomfortable situation. As handler's are the principal person responsible for their dog welfare it is important to make sure that they are trained in dog behavior and that the institutions give them full control over the management of the framework based on their expertise of their dog. Research in applied ethology must be developed to provide science-based criteria to assess therapy dog welfare [62,99] to give clear guidelines to handlers and, on the other side, as experts of their dogs handlers' voices must be legitimize in research about AAI.