19.2.1 Frameworks for Welfare Assessment

Animal welfare has moral, ethical, and scientific aspects (Fraser 1995). Assessing welfare involves application of scientific knowledge, while determining what constitutes acceptable welfare is an ethical question guided by the moral standards of the community/society making the determination (Mench 1998). The field of animal welfare science encompasses a wide range of biological disciplines, including behavioral ecology, physiology, zoology, psychology, neuroscience, and ethology (Dawkins 2006). Therefore, a comprehensive review of animal welfare science is beyond the scope of this chapter. The reader is referred to Appleby et al. (2018) or Fraser (2008) for thorough overviews. What follows are brief summaries of several important or influential animal welfare assessment paradigms.

Concern for animal welfare reaches far back in history (Dawkins 2006), though the field of animal welfare science developed following a report by the Brambell Committee (1965), convened by the UK government in response to public concern regarding farm animal welfare. The concern stemmed from practices associated with the intensification of livestock farming and was spurred by the publication of Ruth Harrison’s book Animal Machines in 1964, which raised awareness of the behavioral restrictions placed on farm animals (Keeling et al. 2011). Most research has concentrated on livestock, including beef and dairy cattle, swine, and poultry. Animal welfare scientists are concerned with the identification and validation of behavioral and physiological welfare indicators, determining how discrete environmental or management factors impact one or a few welfare indicators, and also with the development of assessment protocols that incorporate multiple indicators to determine overall welfare at the individual or facility level (Duncan 2005; Webster et al. 2004). Inferences about how a particular factor influences animal welfare are then made, based on the difference in the behavioral or physiological indicator(s) between treatments or within subjects (in crossover studies). For example, a researcher may assess how space influences the behavior of dogs (Hubrecht et al. 1992).

Welfare research frameworks can be broadly classified as emphasizing biological functioning or subjective feelings (Keeling and Jensen 2002). Commonly used welfare indicators that fall under the “biological function” umbrella include body condition score, gait score, presence of injury and/or illness, production parameters (e.g., milk production in dairy cattle), normal reproduction, and physiological stress as measured by activation of the hypothalamic‐pituitary‐adrenal (HPA) axis (Mason and Mendl 1993). Indicators used to assess subjective state are mostly behavior and body language based but also include other measures such as spectral analyses of vocalizations (Duncan 2006). Most physiological indicators are not clearly positive or negative, making them of limited use in determining affective state (Polgár et al. 2019). Behavioral indicators include frequency and/or duration of behaviors associated with stress and fear, changes in responsiveness to reward that indicate anhedonia or reward sensitization (Cussen and Reid 2020), choice of and/or latency to acquire resources in preference tests, effort exerted to access resources in motivation tests (Kirkden and Pajor 2006), and the frequency and/or duration of behaviors, such as play, thought to be associated with positive affective state (Boissy et al. 2007).

Historically, animal welfare research focused on pain and physical suffering because their absence indicates minimally acceptable welfare and because they were less contentious concepts among scientists who eschewed the study of internal subjective states in the farm animal production industry (Duncan 2005; Mench 1998). However, animal welfare science has, since its inception, been concerned with the feelings of animals—even the Brambell Report stated that animals’ feelings were central to assessing animal welfare (Keeling et al. 2011). While much focus was on negative affective states, or “suffering” (Dawkins 2008), a great deal of research effort concerned what animals wanted (i.e., motivated behaviors; Mason and Burns 2011) and positive affect (Boissy et al. 2007). This was true prior to the advent of “positive welfare.” (For a clear discussion of historical context, see Lawrence et al. 2019.) Many welfare scientists, and certainly the general public, have a holistic view in which “animal welfare” means ensuring animals are “healthy and happy” (Keeling 2005). Thus, two questions are fundamental to animal welfare: “Q1: Will it improve animal health? and Q2: Will it give the animals something they want?” (Dawkins 2008).

Most animal welfare assessments evaluate the welfare of individuals but are conducted in production situations that make assessment of every individual impractical. Assessments in these settings must be brief enough to be feasibly completed but thorough enough to determine how well individual animals are coping with their environment. To achieve this, they usually include both input‐based and animal‐based measures. Input‐based measures consider the availability and quality of resources, the condition and safety of the housing environment, and husbandry or management parameters. For example, for group‐housed animals, the number of animals per square foot (stocking density) and access to resources determined by the number of drinkers or feeding troughs per animal might be considered. For individually housed animals, input‐based measures include access to a clean resting place, air quality, provision of veterinary care, and so forth. The benefit of input‐based measures is they are relatively easy to quantify and are objective (e.g., number of kennels with feces, presence or absence of clean drinking water). Animal‐based measures can also be objective but are more difficult to measure. For example, physiological parameters like cortisol require blood or saliva samples to be collected and analyzed. Behavioral animal‐based measures may also be used and, in general, are rated using a subjective scale or score. This is also true for some health‐related animal‐based measures like lameness, which is also subjectively determined. While input‐based measures are easier and more objective to score, they are less directly tied to welfare than animal‐based measures (Berteselli et al. 2019). Because of this, some assessments rely mostly or completely on animal‐based measures. The latter approach provides a snapshot of the animal in the moment but fails to provide information about the animal’s welfare over any period of time (Webster et al. 2004).

Despite extensive research efforts, welfare indicators are still contentious. There is no “gold standard” measure that definitively indicates good or poor welfare (Mason and Mendl 1993). This is especially true for indicators of positive welfare, which have been more recently developed and less empirically tested than indicators associated with stress (Forkman 2009; Lawrence et al. 2019). With few exceptions, no single measure or indicator is sufficient to assess an animal’s overall welfare (Spoolder et al. 2003). An animal may have a disease process, even a painful one, but have adequate or even positive welfare if pain is managed. Additionally, pain itself is subjective, and what may be considered very painful by one animal may be well tolerated by another—so the same injury or illness may have different welfare consequences. Conversely, a healthy animal may have poor welfare if they spend extended periods of time afraid or frustrated by thwarted goal‐oriented behavior (Duncan 2006; Jensen and Toates 1993). Further complicating a single‐parameter approach, age, sex, breed, and personality may all influence how animals respond to acute and chronic stress, which has implications for how and when they show stress‐related behavior(s) (Stephen and Ledger 2005). To assess an animal’s overall welfare requires integrating information from indicators of both physical and psychological health (Webster et al. 2004). This can prove difficult for several reasons, including that not all indicators covary, even if they purportedly measure the same thing (e.g., stress) and some indicators vary in the same way in both positive and negative situations (e.g., cortisol may increase with eustress or distress) (Mason and Mendl 1993).

What indicators are included and how they are integrated lies at the heart of animal welfare assessment (Botreau et al. 2007a, 2007b). The Brambell Report laid out minimum criteria for farm animals: the ability to stand up, lie down, turn around, stretch their limbs, and groom their body (Brambell 1965). These original freedoms were later revised by the Farm Animal Welfare Council (FAWC) into the well‐known Five Freedoms: freedom from thirst, hunger and malnutrition; freedom from thermal and physical discomfort; freedom from pain, injury, and disease; freedom from fear and distress; and freedom to express normal behavior (Farm Animal Welfare Council 1992). The Five Freedoms framework emphasizes avoiding physical and mental suffering. The freedoms include physical health, behavior, and subjective experience; they are broad categories into which many individual welfare indicators can be placed (Webster 2016; Webster et al. 2004). An early model of animal welfare depicted three interlocked circles representing physical well‐being, mental well‐being, and natural behavior. Together, these three categories encompass factors associated with each of the Five Freedoms. Each circle overlaps to some degree with each of the others but also includes space unique to that domain (Appleby and Hughes 1997). Natural behavior was initially emphasized because it was thought animals had certain behavioral needs related to performing internally motivated behaviors (see Duncan 1998 for an overview). The concept was later challenged on the grounds that thwarting goal‐directed behavior could cause frustration regardless of the loci of motivation (Jensen and Toates 1993) and because “natural” does not necessarily equate with positive as, for example, in the case of aggression. From a practical standpoint, however, the repeated frustration of highly motivated behaviors likely causes animals to suffer (Jensen and Toates 1993), so providing opportunities to express those behaviors is important for animal welfare (Duncan 1998; Mason and Burn 2011). In addition to being intrinsically important to welfare, behavior is relevant to assessing welfare as an indirect indicator of the animal’s subjective state—for example, the presence or absence of play behaviors or the frequency and duration of rest.

Unlike the Five Freedoms, which focus on suffering, the criteria included in the Welfare Quality^® four principles include the presence of positive physical and mental states. The Welfare Quality (WQ) approach to animal welfare assessment was developed by a multi‐year, multi‐institution EU‐backed research initiative. The project was motivated in part by the need to assess welfare on‐site where the animals were housed and to develop relevant, reliable, and practicable indicators of animal welfare (Veissier and Evans 2015). The WQ four principles are good feeding, good housing, good health, and appropriate behavior (Forkman 2009). Twelve criteria were developed, each corresponding to one of the four principles. Multiple welfare measures were included to assess performance on each criterion, with 30 measures in total. For each principle, at least one animal‐based measure was included. On‐farm assessments use those measures to calculate criteria scores, the criteria scores are then used to calculate a score for each of the four principles, and these principle scores are then used to determine the overall assessment of animal welfare for the facility (Veissier et al. 2009). Welfare assessments are conducted by third‐party auditors that are trained in using the WQ assessment, which ensures consistency across facilities and accuracy of implementation.

The Five Domains Model is similar to both the Five Freedoms and the WQ four principles inasmuch as it attempts to group factors relevant to animal welfare under categorical headings to arrive at a picture of overall animal welfare. Similar to the WQ, it includes both the absence of unpleasant states and the presence of pleasant experiences. The domains are nutrition, environment, health, behavior, and mental state. The first four domains are conceived of as inputs to the fifth domain, which represents the affective state of the animal (Mellor 2017). Fraser and Duncan hypothesized that negative affective states underpin “need situations,” whereas positive affective states underpin “opportunity situations” (Fraser and Duncan 1998). Similarly, the Five Domains Model segregates survival‐related domains from the behavior domain, which is mostly concerned with positive welfare. But behavior clearly encompasses nutrition (appetitive and consummatory behavior), environment (frustration‐related behavior), and health (illness behavior). The model is not an assessment protocol but a framework in which to consider multiple aspects relevant to animal welfare (Mellor 2017). In this respect, it is similar to the Five Freedoms, which Webster (2016) refers to as “guideposts” for considering animal welfare. The Five Freedoms and Five Domains Model differ in the emphasis on absence of suffering versus presence of positive affect, respectively, and the specificity of welfare inputs/outputs. Webster (2016) suggests the Five Domains approach may be more useful in an intellectual than a practical capacity, but it has been widely discussed in the literature.

19.2.2 Assessing Welfare in Animal Shelters

In‐shelter welfare assessments share many of the same challenges as on‐farm welfare assessments. Typically, large numbers of animals are housed in a facility where the daily schedule cannot be interrupted for extended periods of time. Staff time is needed to perform the assessment in the absence of third‐party auditors. This means the welfare assessment must be relatively brief, and animal‐based measures should reflect welfare state while also being straightforward to observe within the practical constraints of the assessment. As in production species, welfare indicators include inputs (environment and management) and outputs (animal‐based measures).

Welfare inputs include resource‐based and management‐based measures (Barnard et al. 2016). Examples of relevant input‐based measures for dogs and cats in shelters are the physical environment, including sound, air quality, temperature and humidity, and housing size and substrate; resources, including nutrition, food enrichments, and social resources such as interactions with humans and conspecifics; and management, including routine veterinary care, the facility’s operations schedule, opportunities for exercise, grooming, and handling (Timmins et al. 2007). Welfare outputs include health parameters such as body condition score and disease, heart rate, and so on, and physiological and behavioral indicators of affective state. The indicators are generally the same as discussed in the previous section, with species‐specific differences (e.g., the preferred social structure or the form of abnormal behaviors exhibited). Recent articles on animal‐based welfare indicators in dogs (Polgár et al. 2019) and cats (Arhant et al. 2015) provide valuable resources for practitioners interested in learning more.

Shelter‐based welfare studies date back almost 30 years, examining environmental factors such as the effect of space allowance on the behavior of dogs (Hubrecht et al. 1992). Most shelter studies examine how specific interventions influence one or a few behavioral or physiological welfare indicators of dogs or cats housed in the shelter (Protopopova 2016). Less research effort has been applied to developing companion animal welfare assessments, where a series of input‐ or animal‐based measures are used to determine the welfare of animals within a given “typical” shelter environment.

Work from an Italian research group extends WQ indicators (see Section 19.2.1) from on‐farm assessment of livestock to in‐shelter welfare assessment of dogs. Dubbed “Shelter Quality” after its farm animal counterpart, the assessment protocol includes input‐ and animal‐based metrics that are aggregated to arrive at an overall facility‐level welfare determination (Barnard et al. 2014). Shelter Quality maps welfare measures onto the same 12 criteria and four principles developed by the WQ project: good feeding, good housing, good health, and appropriate behavior (Forkman 2009). Measures are a mixture of animal‐ and input‐based; they include continuous (e.g., space dimensions), binary (e.g., presence‐absence of disease condition), and ordinal (e.g., emotional state ratings) variables (Berteselli et al. 2019). Measures are split out across the shelter, pen (or kennel), and individual level. A random subsample of pens and individuals are observed for the assessment (Barnard et al. 2014).

Shelter Quality was developed and refined for European shelters, where several countries have legislation prohibiting the euthanasia of physically healthy animals. A surplus of animals led to warehousing dogs, with shelters holding large numbers of animals for long periods—up to the remaining years of the dogs’ lives (Barnard et al. 2016). Length of shelter stay is relevant to welfare assessment because the same welfare indicator may be differentially impacted by acute and chronic stress. An example is cortisol, where acute stressors activate the HPA axis, causing an increase in circulating cortisol. Chronic stress, however, can lead to dysregulation of the HPA axis and hypo‐secretion of cortisol in response to an acute stressor (Hennessy 2013). Additionally, length of stay can influence the type of welfare threat experienced by the animal. For example, in a large survey study, Stephen and Ledger (2005) assessed the prevalence and onset of 15 behavioral indicators of welfare (positive and negative) in shelter‐housed dogs in the UK. They found a higher prevalence of fear‐related behaviors (e.g., escape, hiding, inappetence) at the study outset, but these decreased during the six‐week period, while frustration‐related behaviors (e.g., wall bouncing, pacing) increased. Which behavioral indicators of welfare are relevant may differ between assessments for shorter‐ or longer‐stay facilities. Similarly, it may not be clear that clinical measures pose a long‐ or short‐term welfare threat unless medical records are included in the welfare assessment (Webster et al. 2004). To date, no one has reported implementing the Shelter Quality assessment protocol in a “normal” stay shelter in the United States; for now, it is unclear how directly relevant the Shelter Quality assessment is for shorter‐stay facilities.

Shelter Quality has the benefit of using outputs from an extremely large and concerted research initiative as a starting point to validate measures for on‐site animal welfare assessment (Veissier et al. 2009). Additionally, because it uses a detailed protocol it can be applied across multiple shelters as a tool to help pinpoint risk factors for poor welfare (Arena et al. 2019). However, because the aim is an “average” facility‐level welfare audit, most measures are averaged across the entire shelter or a subset of the shelter’s housing units. That means fewer measures recorded at the individual‐animal level, including those meant to determine emotional state or abnormal behavior (both measured at “pen” level) (Berteselli et al. 2019). In contrast, an assessment developed by Kiddie and Collins (2014) focused exclusively on animal‐based measures of individual dogs in 13 different shelters. Most indicators were aimed at assessing emotional state (positive and negative), with three health‐related indicators. An equal number of newly admitted and long‐stay animals were included. Dogs within each duration group were assigned to either standard husbandry or additional enrichment (composed of human interaction) treatment groups. Animals’ behavior was observed from a distance (i.e., without human disturbance) and during in‐kennel interactions. The authors report that the animal‐based assessment distinguished between the treatment groups, though most of the variability was explained by the shelter (which was included as a random effect in the statistical analyses) rather than treatment group (Kiddie and Collins 2014). Although a conceptually promising study, the biological relevance of the small difference in scores between the treatment groups is unclear—especially given the large range of scores within groups (see Botreau et al. 2007a). Further, a presence‐absence approach to behavioral indicators may not accurately reflect differences in the duration or intensity of emotional states.

An alternative (though compatible) approach to documenting what an animal does and/or how frequently they do it is observing how an animal expresses behavior. This can be done quantitatively, through kinematic analyses. Kinematic analyses are most familiar in the context of disease or injury assessment (Tashman et al. 2004) but are conceptually linked to an animal’s subjective experience and may be useful for assessing welfare (Guesgen and Bench 2017). More research is needed before kinematics is brought to bear on shelter animal behavior assessment.

The most common way to look at the “how” of behavior is an approach called Qualitative Behavior Assessment, or QBA (Wemelsfelder 2007). Unlike typical welfare assessments, QBA does not attempt to measure various welfare parameters and then integrate them (Spoolder et al. 2003). Instead, QBA asks raters to assess the “expressivity of the whole animal” as the animal engages with their environment. The approach is based on the premise that such qualitative assessment better captures the animal’s welfare state compared to breaking the whole into parts and then attempting to reassemble them to arrive at a welfare score (Wemelsfelder and Mullan 2014). Multiple raters watch a video clip of the animal in their environment and use their own words to describe the behavior and affect of the animal (a process called “free choice profiling”). Raters then use their own list of descriptors to score other video clips (Wemelsfelder and Lawrence 2001).

QBA was first used as an on‐farm tool for assessing welfare of production animals but was subsequently extended to a variety of other farm and companion animal species, including working dogs (Walker et al. 2010) and dogs in animal shelter and home environments (Walker et al. 2016). Qualitative ratings of dogs’ overall subjective state were found to generate high agreement between raters (Walker et al. 2010), suggesting QBA may be a useful approach to assess dog welfare. Agreement does not guarantee accuracy; two raters could consistently misinterpret the animals’ welfare state and still show high inter‐rater agreement. However, a later study comparing dogs in shelters and home environments found a correlation between the QBA ratings and behaviors coded from video using an ethogram, suggesting raters are accurately assessing the dogs’ demeanors (Walker et al. 2016). It is worth noting, however, that the correlations between QBA and quantitative behavior coding appeared to be most robust for relatively straightforward states like “relaxed” and “anxious.” How well QBA can parse “depressed” from “bored,” and what physiological measures they could be validated against, is still unclear. This question matters because people differ in their interpretation of canine behavior—including experts in animal behavior—regardless of their experience working with dogs (Tami and Gallagher 2009).

19.3.1 Methodological Considerations

19.3.2 Daily Rounds

Daily rounds can serve as an informal and practical means of judging welfare and addressing welfare concerns in a shelter environment and can enhance collaborative care among different teams (medical, behavior, operations). A well‐known population management tool, daily rounds have been defined as “a systematic monitoring process to promptly identify any health or welfare problems and help keep the population management plan on track” (Newbury and Hurley 2013); however, they may best be used as a proactive approach to in‐shelter care that helps identify animal needs prior to problem development. When properly and regularly conducted, daily rounds can help create a sense of urgency in meeting animals’ needs and moving them through the shelter system, decrease length of stay, prevent crowding, maintain the population within the available housing capacity, and maintain operations within the available resources of staffing, time, and money. In this way, daily rounds are one of the most effective tools available to protect and promote good welfare in the shelter environment. Daily rounds are functionally distinct from clinical case rounds (i.e., assessing status of clinical conditions and effectiveness of treatment plans) or facility rounds (i.e., addressing maintenance, safety, and cleanliness), although it may be most efficient to conduct each of these processes simultaneously given the overlap of personnel involved.

For each animal in the shelter population, the specific goals of daily rounds are to (i) identify their unique physical and behavioral needs; (ii) create, enact, update, or confirm their anticipated disposition pathway as needed; and (iii) ensure accountability of personnel in addressing goals one and two. It is recommended that daily rounds be conducted by a team of staff members with decision‐making authority. Minimally, this team should include individuals to represent shelter operations, medical, and behavior perspectives, and attendees should have direct knowledge of the animals in the current shelter population. A population manager should be appointed to lead the daily rounds team and initiate the action steps identified during the process; this role may be designated to an existing member of the daily rounds team. Perspectives from adoption or animal placement staff as well as executive management team members should also be periodically included.

Active participation in daily rounds should be expected from those assigned to the daily rounds team, and time to engage in the process must be protected. Although there may be a benefit to occasionally conducting rounds at different times of the day, establishing and adhering to a consistent schedule will ensure it is a regular part of daily tasks and may facilitate detection of changes in physical or behavioral health. Conducting daily rounds prior to cleaning and feeding can aid in the identification of animals that are inappetant or otherwise cannot access food or water and can allow for the assessment of physical (e.g., diarrhea, increased urination) and behavioral (e.g., prior housetraining) habits. For group‐housed animals, consideration should be given to conducting rounds during feeding on a regular basis to ensure ready access to fresh food and water. Once an efficient system is established, daily rounds can generally be conducted within an hour, even in shelters with large animal populations.

With limited exceptions (e.g., animals in foster homes or infectious disease isolation), the rounds team should physically approach each animal’s enclosure to evaluate their status and have real‐time access to current animal records. Typically, variations on the following questions should be asked and answered for each animal during each rounds session:

1. Who are you? Do physical, cage‐front, and animal record identifiers match?
   
2. How are you—physically and behaviorally?
   
3. Are you where you should be? Is your housing location appropriate to meet your needs while also being suitable for the health and well‐being of the larger population? Does it match the location indicated in the animal inventory?
   
4. What resources or services can we provide today to better meet your physical and/or behavioral needs?
   
5. Where are you going (i.e., What is your anticipated outcome pathway?), and what resources or services can we prepare or schedule in advance to ensure you get there as efficiently as possible?

Discussing the answers to these questions also allows for assessment of the anticipated outcome pathway and the opportunity to follow‐up on action steps from the previous day’s assessment. Knowledge of the animal’s intake date/length of stay and stage in care is also critical context to inform decision‐making during daily rounds. Ensuring that these discussions are founded on a welfare‐based framework, such as the Five Freedoms, the Five Domains, or the four principles, can help guide decision‐making and prioritization of animal needs.

The daily rounds process, decisions made, and action steps identified should be recorded in real‐time on an animal inventory log or purpose‐made action list to document findings, aid in animal care task assignments, and allow for accountability and follow‐up. (A variety of sample logs are available online.) A key component of the daily rounds process also includes regular analysis of such documentation to identify population‐level trends and persistent bottlenecks to animal flow. Not only does an effective approach to daily rounds protect and promote animal welfare by optimizing the use of resources and preempting delays in animal care or flow‐through; it also ultimately results in increased numbers of animals served.

19.3.3 Monitoring Welfare in a Population of Dogs with Extreme Fear or Anxiety

The ASPCA Behavioral Rehabilitation Center (BRC) is a purpose‐built facility dedicated to providing behavioral rehabilitation for extremely fearful dogs and operates on the philosophy that “Everyone is on the Behavior Team.” This approach recognizes that, as they do when it comes to animals’ physical health, all staff have important insights on the behavior and psychological welfare of animals in the facility, not just the behavior experts on staff. Successful care and rehabilitation require all teams working as a single unit to continually monitor, document, and assess each animal’s progress and QoL (see Box 19.1). This collaboration can result in increased recognition of welfare threats and enhanced delivery of care, ultimately hastening the animal’s recovery and improving their chances for long‐term success. It can also result in timely outcome decision‐making when rehabilitative interventions are ineffective and humane euthanasia is appropriate.