Prioritization of Farm Animal Welfare Issues Using Expert Consensus

Prioritization of animal welfare issues can help identify which areas most require research funding and raise awareness of best practices. A modified Delphi method was used to obtain expert opinion on the highest priority welfare issues for UK farmed livestock. Fifty-eight UK-based experts were recruited onto the study, with a minimum of 3 years experience of working with either cattle, pigs, poultry, or small ruminants (12–16 experts per group). Experts were chosen to represent a broad range of opinions. Two rounds of surveys were conducted online using Online Survey, and the final round was an in-person workshop with 21 experts. In the first survey, experts were provided with a comprehensive list of species-specific welfare issues derived from the literature. Participants were asked to score each welfare issue, for (i) severity, (ii) duration, and (iii) prevalence on a 6-point Likert scale. The results of the first survey were reviewed and the welfare issues which scored a neutral-to-high response (scores 3–6) were carried forward. In round 2, participants were asked whether they agreed or disagreed with the rankings that were made from the results of round 1. The final stage of the process was a workshop, which consisted of a combination of group exercises and discussions, to reach the final consensus. Welfare priority lists were divided into two categories: severity/duration, and prevalence, to identify the priority welfare issues affecting individual animals and the population, respectively. Across all farmed species common concerns were inadequate or inappropriate nutrition, inability of stockpeople to recognize or treat welfare issues (such as pain or behavioral problems), lameness, chronic or endemic health issues, euthanasia and mortality and morbidity of neonates. Specific concerns related to behavioral restriction and damaging or abnormal behavior in pigs, poultry and dairy animals, inadequate housing for pigs and poultry, consequences of breeding decisions in pigs and poultry, and lack of access to veterinary care in sheep and beef. This Delphi process resulted in consensus on the most significant welfare challenges faced by UK livestock species and can help to guide future research and education priority decisions.

Prioritization of animal welfare issues can help identify which areas most require research funding and raise awareness of best practices. A modified Delphi method was used to obtain expert opinion on the highest priority welfare issues for UK farmed livestock. Fifty-eight UK-based experts were recruited onto the study, with a minimum of 3 years experience of working with either cattle, pigs, poultry, or small ruminants (12-16 experts per group). Experts were chosen to represent a broad range of opinions. Two rounds of surveys were conducted online using Online Survey, and the final round was an in-person workshop with 21 experts. In the first survey, experts were provided with a comprehensive list of species-specific welfare issues derived from the literature. Participants were asked to score each welfare issue, for (i) severity, (ii) duration, and (iii) prevalence on a 6-point Likert scale. The results of the first survey were reviewed and the welfare issues which scored a neutral-to-high response (scores 3-6) were carried forward. In round 2, participants were asked whether they agreed or disagreed with the rankings that were made from the results of round 1. The final stage of the process was a workshop, which consisted of a combination of group exercises and discussions, to reach the final consensus. Welfare priority lists were divided into two categories: severity/duration, and prevalence, to identify the priority welfare issues affecting individual animals and the population, respectively. Across all farmed species common concerns were inadequate or inappropriate nutrition, inability of stockpeople to recognize or treat welfare issues (such as pain or behavioral problems), lameness, chronic or endemic health issues, euthanasia and mortality and morbidity of neonates. Specific concerns related to behavioral restriction and damaging or abnormal behavior in pigs, poultry and dairy animals, inadequate housing for pigs and poultry, consequences of breeding decisions in pigs and poultry, and lack of access to veterinary care in sheep and beef. This Delphi process resulted in consensus on the most significant welfare challenges faced by UK livestock species and can help to guide future research and education priority decisions.

INTRODUCTION
Animal welfare remains an area of consistent public concern, with acceptance of animal sentience enshrined in the legislation of many countries. Historically, animal welfare concerns have been directed primarily at farmed livestock, with increasing awareness and unease at the intensification and confinement of animals in the 1960s [e.g., following the publication of "Animal Machines" by Ruth Harrison in 1964 (1)]. In the United Kingdom, there is strong animal welfare legislation with good compliance generally (2). However, there are differing views of what is important for animal welfare, from different stakeholders, which may result in conflicting welfare outcomes (3,4), and decisions on which aspects of welfare to prioritize.
Traditionally, farm animal welfare has referred to minimal acceptable standards of farm animals. However, development of animal welfare science over the last decade have started to shift from providing acceptable welfare standards that minimize animal "harms" or suffering toward providing animals with "positive" welfare (2,5). Acceptable standards of animal welfare can be determined by legislation, which may set the minimal standard that a society is prepared to accept in food production, and by retailer and other accreditation schemes, labels or standards that provide higher levels of animal care. The inclusion of positive animal-based welfare measures, such as the ability to move freely and a positive human-animal relationship, among the proposed guiding principles for OIE animal welfare standards (6) reflects that positive welfare is now an active topic of discussion on the world stage (7). Assessing farm animal welfare concerns and improving standards is an ongoing process and one that requires significant scientific research and financial support.
The relative priority given to specific welfare issues can be determined by public concern, political interests, such as international trade or legislation, or by weighing the scientific evidence. Animal welfare science is a large field, since it embraces everything that may affect the physical and emotional state of the animal, its ability to cope and its overall quality of life (8). Governmental and funder resources for tackling animal welfare issues are often limited (9), therefore the prioritization of farm animal welfare issues can help identify which areas most require research funding and raise awareness of best practices.
There are many possible methodologies to try to rank the welfare priorities of animals. However, eliciting expert consensus has been shown to be reliable and very helpful in animal welfare science (10)(11)(12). In this study a modified Delphi method was used. The Delphi method is based on the assumption that group consensus is more representative than individual opinion. The method relies on a group of experts participating in multiple rounds of surveys in an attempt to reach consensus on an issue. The Delphi method uses "rounds" of surveys, with each round becoming more specific/focussed. The classical Delphi method is anonymous, liberating participants to speak their mind without influence of others' views or opinions.
The objective of this study was to prioritize welfare issues of cattle, pigs, poultry, and small ruminants in the UK using expert consensus. To achieve this we brought together expertise from multiple key stakeholders in animal welfare (e.g., veterinarians, academics, practitioners, farmers, voluntary, and charitable bodies, etc.) to generate consensus on the most important animal welfare issues for farmed livestock.

METHODS
All research generated from this study was approved by the University of Edinburgh's Human Ethics Review Committee (HERC). The full description of the methods can be found in Rioja-Lang et al. (13). The results presented here are part of a larger study that identified priority welfare issues for a number of managed animal species in the UK.

Recruitment of Experts
Experts were defined as having worked in their field for more than 3 years, were based in the UK and thus assumed to be knowledgeable about UK farming systems, and were practicing veterinarians, academics, charity sector employees, farmer representatives, and policy officials with responsibility for farmed livestock.
We recruited 58 experts to the project (pigs = 15; poultry = 12; cattle = 16; small ruminants = 15). The recruitment process began with building a list of contacts of well-known experts in their field and contacting them via email and describing the study, the aims, and the Delphi procedure. Additionally, we employed a "snowball-sampling method" whereby these initial contacts were also asked to refer us to other experts in their field who would be a good addition to the study. When an expert agreed via email to participate in the study, they were then sent a consent form to sign in accordance with the HERC guidelines. The consent form also contained a more detailed description of the study objectives, protocol, and expected timeline. It was also explained to each expert that participation was anonymous (except for the workshop) and voluntary, and they were able to leave the process at any time.

Literature Review
A review of the scientific and industry literature was conducted to construct a comprehensive list of documented welfare issues for each species. However, at every stage experts were also given an opportunity to add any other issues that they felt might be missing. These original lists generated from the literature consisted of welfare issues covering a broad spectrum of animal-based welfare outcomes, management, and resourcebased risk factors (cattle = 72; pigs = 80; poultry = 81; small ruminants = 76; see Supplementary Material for full lists). For the purposes of the online surveys, the welfare issues of beef and dairy cattle were considered together, small ruminants consisted of meat and dairy goats and sheep, and the poultry group consisted of laying and spent hens, broiler birds, turkeys and ducks. During the workshop the experts re-categorized the ruminant welfare issues to "Dairy animals (cows and goats)" and considered "Sheep and Beef cattle" together.

Questionnaire Design
Two rounds of surveys were conducted online using the Online Surveys tool (formerly Bristol Online Survey). Each round of surveys was completed anonymously at a time suitable to each expert within a 2-week window. In both rounds of online surveys, demographic data were also collected from the participants. This included: year of birth, gender, profession, highest level of education, and number of years since graduating highest level of education.
The experts were asked via email for the criteria they considered important to rank/score welfare issues. Experts agreed that welfare issues should be ranked according to three different criteria: (i) severity (defined as the likely severity associated with the welfare issue in the experts opinion), (ii) duration (defined as the likely period of time that the animal would be affected by the welfare issue in the experts opinion), and (iii) prevalence (defined as the experts perceived proportion of the population affected by the welfare issue). However, all criteria were considered equally important in prioritizing animal welfare, and therefore were all used in determining welfare ranking. During round 1 of the online surveys, experts were asked to score each welfare issue for each of the three criteria, on a 6-point Likert scale, where 1 = never/none, and 6 = always/high. An even numbered scale was chosen as this forced experts to make a decision (at least somewhat important to unimportant).
The results of the first survey were reviewed and the welfare issues which scored a median neutral-high response (scores 3-6 on the Likert scale) on any of the criteria were carried forward into round 2 as these were those that the experts ranked as being at least somewhat important. These welfare issues were then presented to the experts according to their ranking from round 1, with the highest-ranking issues (score 6) at the top of the list, and the lowest ranked issues (score 3) at the bottom. This process was completed for all issues scoring at least 3 for each criteria. Experts were asked whether they agreed or disagreed with the rankings. In instances where experts did not agree with the relative ranking, they were asked to indicate if they would like to see that issue higher or lower on the ranked lists. Agreement between experts was determined by calculating Fleiss' kappa statistics, and assessing the percentage of experts who agreed with the ranking.

Workshop
The final stage of the Delphi process was the workshop, held at Edinburgh University in early Sept, 2018. Twenty-one animal welfare experts participated in the workshop, including nine specializing in farm animal welfare (academics = 3, veterinarian = 3, farmer representative = 3). However, the other experts in attendance were general animal welfare experts (academics = 3, vets = 2, industry representative = 1, NGO = 6) therefore were able to competently give an expert opinion on farm welfare issues also. Over the 2 days there was a series of small group (species specific) and large group (to identify cross-cutting issues) exercises and discussions in order to finalize the priority welfare lists for each species group and to rank them.

Demographic Information
Experts had a similar age structure across the different species groups and were female-biased, except for small ruminants

Round 1, Online Surveys
Of the initial list of welfare issues prepared from the literature 35, 37, 42, and 45 issues scored a mean of 3.0 or above for pigs, poultry, cattle and small ruminants, respectively, from Round 1 for either prevalence, severity, or duration of the welfare issue (Tables 2a-d). Standard deviations were relatively high for some issues ( Table 2), suggesting there were considerable differences of opinion between experts.

Round 2, Online Surveys
The level of agreement between experts on the average ranking for each welfare issue was relatively low (Fleiss' κ, Table 3), with particularly low levels of agreement between cattle experts. For pigs 86% of experts agreed that delayed euthanasia and tail biting were the most severe challenges to welfare, but did not agree on other high scoring issues. For poultry good agreement was achieved for the prevalence of keel bone fractures (75%), the severity of poor foot and leg health (78%), and duration of impact of artificial lighting regimes and inappropriate housing conditions (both 75%). Cattle experts had good agreement (77-92% of experts agreed on the ranking position) on the issues ranked most highly for prevalence, severity and duration (poor foot health, high calf mortality, common production diseases,       poor general health, and delayed euthanasia), but agreement elsewhere was poor. For small ruminants, only the severity of lameness (86%) and lack of suitable analgesics (79%) had high agreement amongst the most highly ranked welfare issues.

Round 3, Workshop
Final rankings of welfare issues for each species were achieved during the workshop ( Table 4). During the course of the Workshop experts considered the severity and duration of welfare issues together (considering the welfare of individual animals), and prevalence of a welfare issue was considered alone (considering the average welfare impact on a population of animals). The experts were asked to create a list of approximately 10 welfare issues; however, this was not strictly upheld and the lists contain between 8 and 11 issues. During the workshop discussion process the cattle and small ruminant experts decided to combine the welfare issues faced by dairy goats and dairy cattle, and also sheep and beef cattle as the issues facing each group were more similar. Therefore, the results from the workshop have been presented in this manner.

DISCUSSION
The study was able to achieve consensus among experts of what was considered the most important welfare issues for farmed species, with the workshop outcomes including all the highly ranked issues from the first online survey, which had good agreement in the second. However, the opportunity to discuss each issue in more detail, and in some cases to combine a number of issues into a single category (for example, combining different aspects of housing design or diseases relating to production), meant that experts had better agreement and reported more confidence in the final rankings produced in the workshop. Overall, there were a number of common issues that arose in all or nearly all species including: inadequate or inappropriate nutrition; inability of stockpeople to recognize and/or treat welfare issues (such as pain or behavioral problems); foot and leg health resulting in lameness; chronic or endemic health issues; euthanasia delay and methods (particularly those used on farm for killing surplus or unwanted male animals), and neonatal mortality and morbidity. A number of specific welfare issues were identified including abnormal or damaging behaviors in pigs, poultry, and dairy animals; inadequate or poor housing and environments for pigs and poultry; consequence of breeding decisions and genetic selection strategies in pigs and poultry; lack of access to veterinary care for beef and sheep, and issues with handling and transport in sheep, beef and poultry.
The study focused on welfare issues experienced by UK farmed livestock, and the experts who took part in the study were based in the UK. However, similar farming systems and methods are used across Europe and other industrialized countries and thus the outcomes are likely to be relevant to the livestock production of other countries with similar systems to those used in UK. A proviso might be that some management and husbandry practices are not legally permitted in the UK that may be allowed elsewhere (e.g., battery cages and gestation stalls are banned within the EU), and some practices are permitted in UK that may be illegal elsewhere (e.g., tail docking of piglets and lambs). Thus, consideration of these issues may be more or less relevant in some countries.

Inappropriate Nutrition
Inappropriate nutrition was identified as a source of welfare concern for individual animals (as indicated by the severity and duration of suffering), and also as perceived to be prevalent or widespread for dairy animals and pigs. This reflected a broad category of concern related to high metabolic rate and consequences for feeding dairy cows (14), low body condition score in sheep and beef (15,16), and methods of feeding in pigs, which may lead to gastric ulceration (17). These concerns were chiefly those experienced by the breeding stock, which may remain on the farm for several years, rather than the young animals destined for prime meat production. However, this does expose that feeding and nutritional management of pregnant and breeding animals may be sub-optimal, either through deliberate food restriction [e.g., of broiler breeders and pregnant sows; (18)] or through difficulties in managing forage resources at certain times of the year (19,20).

Lack of Stockperson Skills or Knowledge
For all species lack of knowledge, training, skills, or expertise in animal behavior and welfare, and/or access to knowledgeable veterinary care, were highlighted as both a source of individual welfare issues in all species, and as prevalent as a welfare issue within each industry, in the experts' opinion. This may be because behavior, pain, and welfare issues are not routinely included in training for stockpeople or livestock advisors, and farm animal behavior and welfare teaching is often limited in veterinary education. Delayed access to veterinary care may be due to economic concerns (veterinary care of individual animals may be considered economically unviable), an inability to provide individualized care to extensively kept species or poultry (leading to individual neglect where pain or disease may go undiagnosed or untreated), incorrect diagnosis and inappropriate therapies from farmers or lack of access to specialist vets in isolated areas. Veterinary care is rarely sought by sheep/goat farmers (21), and Lovatt (22) suggests that the veterinary services available for sheep are generally the general farm veterinary practitioner, and not a specialist (unlike the pig and poultry sector). However, veterinarians can have the biggest influence on farmer behavior [e.g., in uptake of strategies to reduce pig aggression (23)]. Therefore, improved access or use of veterinary advice for behavior or welfare problems would have benefit in improving animal welfare. These are complex issues but may be partially addressed through improved staff training programmes, consideration of veterinary Continuing Professional Development (CPD) in these specific areas and better on-farm protocols.

Pain Recognition and Management
A prominent issue in the area of caretaker knowledge, as identified by the experts, concerns the lack of pain recognition, and pain management. The source of the pain can be from a variety of common diseases, physical trauma, and through the imposition of painful management procedures. Stockpeople may not recognize the subtle behavioral indicators of pain [e.g., as found by (24)], and pain following management procedures may be unobserved or considered relatively trivial, despite considerable research suggesting otherwise [e.g., (25,26)]. In some instances, stockpeople may recognize that an animal is in pain, but lack access to suitable medication to administer, feel unable to deal with the issues [as shown by O'Kane et al. (27)] or have concerns regarding residues. This is particularly problematic for small ruminants, where few analgesic and anesthesia options are available (28), which can limit the ability of veterinarians to effectively manage animal pain. This is an obvious area of concern that requires further research or legislation.

Lameness and Mobility
Lameness and mobility issues were of concern for all species, arising from selection for rapid growth in pigs and broilers, footpad dermatitis in poultry, infectious agents in sheep and poor claw health in cattle. A number of studies have provided good evidence for the suffering caused by these issues (29)(30)(31)(32), the impact on production losses (33)(34)(35) and have defined the prevalence of these conditions in the UK. For example, a study of 80 farms over 2.5 years suggested dairy cow lameness varied from 16.2 to 19.3% over the year (36). A more recent study by Vee Randall et al. (37) reported that across 43 farms lameness was 30.1% (7.3-60.6%) from a total of 5,620 dairy cows. In broiler chickens 27.6% of 40 weeks old birds were lame (38), despite on farm culling for poor leg health. Lameness has also been reported to affect nearly 5% of sheep in England (39) and 4.5% of sows (40).
Given the extensive knowledge about the impact and treatment of lameness in different species, development of effective education, training and intervention programmes are required to reduce this area of farm animal welfare concern.

Chronic and Endemic Health Issues
Experts raised the issue of untreated or unresolved chronic disease issues, and poor health issues, across all the farmed mammalian species. Some of these were specifically linked by experts to production, such as mastitis in dairy animals, lameness or dystocia, whereas others may reflect housing or hygiene practices, failure to maintain biosecurity and the presence of endemic disease. Production diseases are considered those that arise through a complex interaction of pathogen presence, with contributory factors such as housing, feeding or management that facilitate disease spread or persistence. In addition to causing sickness, pain and malaise, poor animal health can lead to an increase in the use of antibiotics on farm, and limit animal production [as described for Johnes disease: (41,42)]. Although larger and more intensive farms may be better at maintaining biosecurity than smaller or more extensive farms, intensification and crowding can increase stress and disease pressure. Consumers generally are more supportive of changes to housing and management that reduce intensification as a way of managing disease, compared to other approaches, and are particularly concerned about antibiotic use, human and animal health and food safety in these systems (43). In addition to various disease control strategies, such as quarantine, vaccination or testing and culling, McAloon et al. (44) has highlighted improved communication with farmers about best practice to limit disease transmission as a requirement to deal with these issues.

Delayed Euthanasia
Delayed euthanasia, i.e., not killing animals that are suffering in a timely manner, and the methods used to kill animals, particularly those for unwanted male animals in the dairy and egg industries were considered a significant source of individual suffering (delayed euthanasia) and perceived to be prevalent (methods of killing). Delayed euthanasia can occur for several reasons, including extending an animal's time for recovery from an illness or injury "to give them a chance, " inexperience of the stockperson in either assessing an animal's prognosis or carrying out the procedure of euthanasia, or waiting for the animal to complete a stage of production before finally being euthanized. For many good stockpeople who care for the animals in their charge it can be a very difficult decision to euthanize an animal. Blackwell (45) has found that it was easier for farm stockpeople to euthanize a sick or injured pig if the farm had a written policy that clearly stated the conditions when an animal should be euthanized. Euthanasia protocols and sufficient staff training in procedures provide staff with reassurance and can reduce the suffering of animals. A particular issue for the dairy and poultry industries is dealing with unwanted or surplus male animals, and this was highlighted for dairy goat kids in particular. As low value animals in a production system that requires female animals, male goat kids are sometimes killed on farm by a variety of methods that include manual killing. However, these methods may have low reproducibility and can impact on welfare, thus methods such as use of non-penetrating percussive devices may improve kid euthanasia practices on farm (46).

Neonatal Mortality and Morbidity
Mortality and morbidity of young mammalian livestock was ranked highly in ruminant livestock, for both the impact on individual suffering and prevalence, and as a consequence of breeding for larger litters in pigs [as discussed by Rutherford et al. (47)]. Concern for the early separation of young livestock from their mothers, particularly in dairy animals and poultry, but also early weaning in piglets, additionally contributed to the priority given to this issue. High mortality can also be considered as a production disease, as described above, with factors such as housing and repeat breeding affecting piglet survival (48), nutrition, hygiene, and litter size contributing to lamb mortality (49) and genetics influencing calf mortality (50). In addition, the failure of passive transfer of immunity, particularly in dairy calves, is a significant risk factor for poor neonatal outcomes. As with other health issues, these problems are complex and multi-factorial, but there has been a significant body of work to improve rates of neonatal mortality in different farmed species. Opportunities to implement these changes would require improved communication to farmers, with solutions tailored to farm characteristics (49), where veterinarians and others can play a role.

Species-Specific Welfare Issues
Housing and Environment For farmed species that are often managed intensively (pigs and poultry), inadequacy of the housing environment and an inability of the management system to meet behavioral needs, leading to damaging behaviors such as tail-biting or featherpecking, were amongst the most important concerns, as rated by the experts. In modern commercial production systems, pigs and poultry are often confined within simple, invariant, housing systems that offer little potential to accommodate their highly motivated species-specific behaviors (51,52). There is substantial evidence that animals suffer when they are unable to show motivated behaviors. For example, when sows are prevented from expressing nest-building behaviors by the restrictive nature of the farrowing crate (53), and insufficient substrate is given for pigs to show rooting behavior (54). In poultry, chickens may be unable to show dust-bathing or nesting behavior (52), because the housing does not provide sufficient space or a suitable substrate. Under UK legislation, pigs must be provided with environmental enrichment, and laying birds and pullets must be housed in "enriched" cages, as barren "battery" cages have been banned in the EU since 2012. Enriched cages do provide hens the opportunity to nest, perch and dustbathe, however, the provisions of these resources can often fall short of ideal and still supress the ability of birds to express some natural behaviors. In addition, often only a minimal amount of environmental enrichment is provided for pigs which, although it might provide some welfare improvement (55), falls well short of truly meeting the animals behavioral needs. A consequence of this is that tailbiting, when pigs bite and chew the tails of pen-mates (56), remains a significant welfare problem, which is often dealt with by docking tails to limit the ability of pigs to carry out this damaging behavior. Although tail-docking is regulated in the EU and should not be a routinely performed procedure, it is still carried out on 80% of farms in UK (57) and occurs without provision of analgesia or anesthesia. Tail-docking as a welfare solution to tail-biting, therefore, involves the acute pain associated with docking, and does not address the underlying causes of tail-biting. As with many of the welfare issues for farmed animals identified and prioritized in this study, tail-biting is influenced by a number of factors, including barren housing, suboptimum nutrition (including misdirected foraging, lack of fiber, restricted feeding regimes), poor health, environmental stressors, stocking density and group size (58). Thus, it can be a difficult problem to mitigate, although recent studies have demonstrated that pigs can be successfully managed with long tails by environmental improvements (59, 60) although these may be financially less profitable than tail-docking (57).

Breeding and Genetic Issues
The most important welfare issue for poultry, at both the individual and the population level, was considered to be the negative consequences that can arise from breeding decisions. In broiler chickens, selection for rapid growth over the last 50 years has had a dramatic impact on the time taken for birds to reach slaughter weight (e.g., increased growth rates from 25 to 100 g per day), has led to exaggerated conformation through selection for breast muscle and has enormously improved feed conversion rates. The consequences of this rapid growth has led to a number of welfare issues including a significant increase in leg weakness and an inability to walk, cardiovascular disease, musculo-skeletal disorders and reproductive problems in the parent birds unless subjected to severe food restriction (18,61). For example in a survey of broiler leg health, nearly a third of birds showed poor locomotion, and 3.3% were non-ambulatory (38). Lame birds will preferentially select a feed containing an analgesic (62), which then improved their walking ability, suggesting that altered gait was associated with pain. Poor leg health also contributes to other welfare issues, such as breast blisters, as the lame birds spend most of the time lying, which exposes them to contaminated litter for long periods of time. The welfare issues experienced by broiler chickens have been shown to be related to increased growth rate (38), which, although partly achieved through changes in management practice, has been largely due to genetic selection (63,64). A number of producers are now acknowledging this issue, but alterations in breeding goals to include aspects of welfare would be required to deal with this issue (63). Concerns were also raised for the welfare consequences of genetic selection of pigs, specifically selection for larger litter sizes. In UK and Denmark, litter sizes have increased from an average of 11.7 and 12.1 piglets per litter in 1996 to 12.0 and 16.6 piglets per litter in 2011, respectively (47). This has led to a parallel increase in prenatal and pre-weaning mortality, from 7.1-7.4% to 17.9-18.2%, respectively, to 10.8 and 23.5% in 2011. Piglet morbidity and mortality is a significant concern for these larger litters, as well as the associated management practices to deal with large numbers of piglets, as the number of viable piglets may exceed the number of available teats (65). These can include use of nurse-sow systems, early and split weaning and use of artificial rearing systems. These systems impose welfare challenges on the sow, such as prolonged confinement during lactation, as well as on the piglet. Although not included in the final lists of most important welfare issues for other species, the consequences of breeding decisions also featured in the important welfare concerns for all species. Because of their long term (often whole life) and enduring impact on animal welfare, these are significant welfare costs, which require longer term approaches to designing appropriate breeding goals (64) to be mitigated.

Handling and Transport
Welfare issues associated with animal handling and transport were highlighted particularly for poultry, and for extensively managed animals (sheep and beef) with little experience of human contact before handling. In poultry, different risk factors are important depending on the type of bird in transit, for example the risk of bone breakages is high in end-of-lay hens, which require gently handling and catching before transport. In general, however, the collection, handling and loading of poultry for transport is stressful for birds [e.g., (66)(67)(68)]. Ventilation and environmental control are also very important for transport of poultry, which have a relatively narrow thermoneutral range and are susceptible to high and low temperatures in transit (which can occur simultaneously in the same load at different locations). Under EU directives (EC1/2005), poultry can be transported for up to 12 h without provision of feed and water. This also acts as an effective limit on duration of transport as providing food and water in transit for poultry is largely impossible with current transport methods. Petracci et al. (69) suggest an estimate of 0.35% mortality in broilers in transit, more recent studies suggest broiler mortality ranges from 0.1 to 0.6% and a higher mortality is seen in end-of-lay hens (0.4-1.0%) compared to meat chickens (70). Risk factors for increased mortality were length of journeys, stocking density in transit and ambient temperature, although farm factors, such as catching practices are also important [reviewed by (70,71)]. Thus, despite some regulation of animal transport, welfare was still considered an issue for animals in transit, and implementation of best practice (e.g., the EU Animal Transport Guides: Consortium of the Animal Transport Guides Project (2017) "Good practices for animal transport in the EU": http://animaltransportguides.eu/) may help to reduce the welfare costs of transport to slaughter.

Limitations of the Approach
Although Delphi studies are popular and can address problems that would otherwise be contentious or intractable, it is important to acknowledge their limitations. Outcomes from a Delphi are based on the agreement between experts, rather than empirical objective data. However, this can be valuable in scoping the issues when these data are not readily available. The number of experts in this study was relatively small [this was close to the numbers recommended in some papers e.g., (72) but lower than others e.g., (73)], although these were drawn from a larger pool of 145 experts, and ideally the reliability of the study should be tested with other groups of experts. There is no statistical test for reliability in Delphi studies. However, the outcomes of this study met with the suggested criteria for credibility (54) as participants were interested and knowledgeable about the field, reasoned discussion and debate was part of the process in the workshop and supporting data from the literature have been presented here, supporting the validity of the outcomes. A previous (anonymous) study of farm animal welfare using a Delphi approach concluded that problems arise when animals are kept in environments that do not meet behavioral needs (74). This was broadly corroborated in our paper, particularly for pigs, poultry, and dairy. In UK, sheep and beef cattle are usually managed extensively, and their environments are more likely to meet behavioral needs, and thus the prioritized welfare issues focused on issues with the treatment of disease and injury. We took a broad approach to welfare issues and included both risk factors (such as housing or management practices) as well as welfare outcomes (such as tailbiting or diseases). In practice this was sometimes problematic for the experts and did require them to consider the importance of issues that were not necessarily on the same scale. In future work separation of these issues may be beneficial.
The Delphi method does not allow us to analyze why particular issues were prioritized and to explore the rationale behind some issues being prioritized over others. Welfare is often seen as being related to biological functioning of the animal, the naturalness of the environment and/or the feelings or subjective experiences of the animal (3). However, a particular preference for or bias toward only one of these areas can lead to different decisions about welfare. The Delphi method attempts to remove these potential differences of opinion by seeking an overall consensus, which may involve weighing the relative importance of these arguments. This may explain the relatively low agreement with the online surveys, where discussion was not possible, and the improvement in agreement in the workshop resulting from discussion. We cannot, however, rule out the possibility that the workshop, by not being anonymous, may have allowed particularly dominant individuals to influence the apparent consensus.

CONCLUSIONS
Overall, the modified Delphi methodology was effective in allowing animal welfare experts to reach consensus on the current most important welfare issues for farmed animals. There is always the possibility that different results might ensue depending on the panel of experts participating (75). However, in studies such as this one, where empirical evidence is unavailable, the Delphi method does provide a framework with which to work. Also, by inviting stakeholders from a range of professional and academic disciplines to participate, there is more likely to be a balance of inevitable discipline-specific biases (9).
The phrasing of survey questions, and focus of the study, was on "welfare issues, " which means that the study only considered negative aspects of welfare. However, as outlined earlier, positive animal welfare is becoming more important and was not addressed in this study, but could be included in further work to provide a more complete view of the welfare of farmed animals. The final priority welfare issues contained a mix of animal-, resource-, and management-based factors, which can be addressed by a mixture of further research, education, communication, and policy-change strategies to implement existing knowledge, or to understand the issues and how to mitigate them in more detail, and to achieve an improvement in the welfare of farmed animals.

DATA AVAILABILITY STATEMENT
The datasets generated for this study are available on request to the corresponding author.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by Human Ethical review Committee (HERC), Royal (Dick) School of Veterinary Studies, University of Edinburgh. The patients/participants provided their written informed consent to participate in this study.

AUTHOR CONTRIBUTIONS
CD designed the project and obtained funding with the assistance of HB, MC, and AL. FR-L was employed to work on the project and conducted the work under supervision of CD, MC, and HB, and produced the first draft of the manuscript. CD, HB, MC, and AL edited and completed the final draft of the paper.

FUNDING
This study was funded by the Animal Welfare Foundation (AWF). AWF is a fundraising and grant giving charity (charity number 287118) directed by veterinary professions, which uses veterinary knowledge to improve the welfare of animals through science, education, and debate. More information can be found at: www.animalwelfarefoundation.org.uk.