Improving the health-related welfare of dogs requires identification and prioritisation of important health issues for targeted reform. Breed-related disorders are established as a significant issue affecting UK dog welfare, but identifying specific disorders on which to focus available resources for maximum canine welfare improvement is challenging [8, 14, 18]. This paper describes the development and application of a VetCompass data-driven strategy for evidence-based canine disorder prioritisation based on comparing relative health-related welfare impact. Novel welfare metrics, developed to assess annual frequency, severity and duration across disorders, were generated for eight major disorders with evidence of breed associations, using EHR data from randomly selected cases identified from VetCompass. Comparable, group-level disorder prevalence, severity and duration scores were combined into novel VetCompass Welfare Impact scores, reflecting relative disorder-specific, population-level welfare impact and highlighting areas of particular welfare concern for potential targeted reform [35, 36]. The described methods for data-driven, health-related welfare metric generation and comparison are applicable across the disorder spectrum. This offers opportunities for future studies to both expand the reported eight disorder comparison, or to objectively assess, compare and suggest priorities within other disorder groups.
Disorders assessed
Breed-related disorders given highest media exposure or cited as particularly concerning by expert panels on pedigree dog health are often those perceived as especially severe for affected individuals within certain high-risk breed groups [8, 32]. Some of these issues have extreme animal-level welfare implications, being present from birth to death, clinically unresolvable and difficult/impossible to manage through owner education and/or changes to practice. While important, such disorders may however be relatively uncommon within dog populations as a whole and priority should arguably be given to disorders with the greatest welfare burden across canine populations.
This study used prevalence estimates generated from VetCompass data [46] to identify the most common health issues of welfare impact with any breed associations, shifting the focus of prioritisation toward issues with greatest health-related welfare burden across UK dogs [35, 36]. All eight disorders assessed had sufficient existing evidence to score at least two on the SEHB scale [14, 18, 19], supporting likely breed-related as a primary disorder or secondary to disorders with breed-associations (e.g. dermatitis secondary to atopy [47]). Despite lower media profiles and historically poorer research focus [48] several of these (e.g. bodyweight/obesity and dental disorders) have been highlighted by recent UK surveys as health and welfare concerns of veterinary surgeons and other UK canine welfare stakeholders [29, 49,50,51].
Prevalence
Of the disorders studied, dental disorders (9.6%), overweight/obese (5.7%) and anal sac disorders (4.5%) had the highest prevalence and showed results that were broadly in line with previous reports [46, 52]. The prevalence values reported in the current study were generally lower than previous estimates that were often based on referral populations that may have been selectively biased towards sicker populations. Nonetheless, these estimates may also reflect under-reporting of certain conditions such as overweight/obese or osteoarthritis by primary-care veterinarians as well as differences in case definition between this and other studies [42, 46]. As applied here, prevalence represented a reproducible and comparable metric providing an annual ‘snapshot’ of relative disorder frequency across the study population as a whole and within breed groups. Whilst the absolute frequencies may be underestimates, the relative order of conditions was likely to be a reliable indicator of their relative disease burden. Building on previous disorder-specific work on data from VetCompass, it provided a systematic basis for disorder groupings which could realistically be investigated using these large-scale, primary care EHR datasets [46].
Duration
Duration of individual health problems is increasingly emphasised by animal welfare scientists as a key contributor to the overall welfare impact, as recognised in human medicine [23]. Disorders affecting dogs for extended periods pose greater risk of cumulative or lifelong pain and other welfare issues, even where daily levels of discomfort/pain, debilitation or distress appear relatively low. A recent report by Teng et al. proposed and applied the Welfare-Adjusted Life Year (WALY), adapted from the human Disability-Adjusted Life Year (DALY) concept, as a measure of cause-specific welfare impact on individual dogs [53]. This complex metric puts strong emphasis on disorder duration as a component of overall welfare compromise, taking into account both time lived with impaired welfare due to a given cause and years of life lost after premature death from that cause (with weighting by perceived cause-specific level of welfare compromise). Application to 10 common disorders demonstrated that those with the greatest and least adverse impact on dogs, according to magnitude of WALY, were atopic dermatitis (WALY 9.73, largely attributable to time lived impaired) and thoracolumbar intervertebral disc disease (WALY 2.83, mainly comprised of years of life lost). It is also worth noting that the course of many chronic, progressive disorders involves episodic “steps”, where issues increase to a threshold which precipitates revisits and additional intervention. Recognising and promoting better management of chronic conditions at these times provides opportunities to significantly reduce welfare burdens associated with chronic disease that may have been historically underestimated [54].
Based on the median age at earliest disorder-associated record, dermatitis, anal sac disorders and conjunctivitis appeared to affect a relatively younger demographic than osteoarthritis and lipoma. Although late-onset (geriatric) health issues are often cited as of particular concern [55, 56], it could also be argued that chronic or progressive disorders routinely affecting dogs from an early age constitute a greater priority concern due to their potentially lifelong welfare effects.
Osteoarthritis, dental disorder, lipoma and overweight/obese had median annual duration estimates that exceeded 50, 60, 70 and 80% respectively with over a quarter of cases affected for the entire annual period. In contrast, otitis externa, dermatitis, anal sac disorder, conjunctivitis had much lower median durations (< 4%). In the latter group of conditions, the number of episodes per case could generally be more reliably determined than individual episode durations. Ambiguous or absent revisit data introduced degrees of uncertainty and subjective interpretation when time delimiting distinct episodes, with median duration often based on only small numbers of episodes. Even where feasible episode ‘start’ and ‘end’ dates were available, these likely underestimated true duration as disorders were likely clinically present before the diagnostic visit. In addition, the median duration metric itself potentially underestimated the longer term, cumulative importance of seasonally recurring episodic disorders. Annual recurrence patterns seen in seasonally triggered allergic skin disorders can manifest as a single, annual ‘flare up’ of dermatitis, otitis externa, conjunctivitis and/or anal sac irritation [47, 57]. This single, seasonal episode per study year generates a low median duration which may not properly reflect cumulative lifetime welfare burden for affected individuals. In the context of the present retrospective study design and data, it was not possible to completely address these limitations. However, median duration could theoretically be assessed with greater accuracy within prospective clinical studies, gathering reliable episode duration through planned progress assessments and welfare impact metrics could be generated using longer defined study periods to explore potential limitations for seasonally linked or infrequently recurrent disorders.
Objective and quantitative methods are required to compare duration across the spectrum of canine disease. The current study estimated the annual duration for each disorder but given that this was across a sample of 250 cases, it also reflected the varying age structures of these cases. Annual disorder duration could also be useful to estimate the lifetime duration for each disorder although this remains challenging until large-scale birth-to-death health data on dogs become available. In the meantime, the novel median duration metric developed in the current study offers a reproducible, comparable alternative to true lifetime duration measures (or WALYs), by restricting retrospective review of EHR data to a single study year and using information typically recorded in these datasets across a wide spectrum of disorders.
Severity
Overall severity scores were highest for osteoarthritis (score 13), otitis externa [11] and dermatitis [10]. Over 50% of osteoarthritis cases presented primarily for the condition at least once, over 50% received over 14 consecutive days of analgesic/anti-inflammatory therapy and over 3% were referred for osteoarthritis management in the study year. This suggests that osteoarthritis is recognised and perceived by both veterinary surgeons and owners as important enough for significant clinical care, often involving long term maintenance prescription only analgesia, multiple additional medical interventions, frequent clinic visits and relatively high levels of referral uptake. Severity scores were marginally lower but showed similar patterns in otitis externa and dermatitis. However, the particularly high sub-score assigned to otitis externa for non-analgesic/non-anti-inflammatory therapy prescribed was influenced by the formulation of many aural medication products, which typically contain anti-septic, anti-microbial, anti-fungal and physical cleansing elements in a single preparation.
Overweight/obese and dental disorders had comparatively low overall severity scores (3 and 7), which appears contradictory to established veterinary consensus on the critical importance of these issues to health-related canine welfare [29, 49,50,51]. Diagnosis was usually incidental, with few associated veterinary visits and minimal interventions received, suggesting that owners may fail to recognise these problems or are not sufficiently motivated or concerned to seek veterinary attention or take up recommended interventions. A 2011 online survey reported that only 52.3% of owners would present a pet primarily for halitosis, supporting low owner recognition or appreciation of the serious clinical consequences associated with dental disorders [55]. Alternatively, owners may be unaware of treatment possibilities, unconvinced of the need for dental care/weight loss in their pets or confused by conflicting advice from different areas of the pet health care sector. Owner choices may be based on external and non-veterinary advice, or financial restrictions or priorities, particularly if treatment or care is excluded by pet insurance policies. Some may be unwilling or unable to invest the time or find recommended day-to-day interventions like tooth brushing and dietary restriction difficult to implement [58,59,60].
Owner uptake (not just veterinary recommendation) of interventions, re-examination or referral were key when assigning disorder-specific sub-scores for individual severity metrics. This helps reflect owner perceptions of disorder importance but could mask veterinary-perceived severity in disorders where owners frequently declined recommended interventions or failed to return for follow-up care, perhaps due to lack of concern or limited resources. As applied here, the severity scoring system did not consider the welfare consequences of common secondary conditions or exacerbation of co-morbidities when assigning severity scores for assessed disorders. For example, no overweight/obese cases received directly therapeutic referral, analgesia/anti-inflammatories or procedures under GA/sedation, but potentially important detrimental consequences of sequelae/co-morbidities such as osteoarthritis [61,62,63,64] also remained unaccounted for in cases. Specific co-morbidity scenarios could theoretically be evaluated and compared in future studies, providing transparent case definitions were clearly defined. Overall, comparison of sub-scores for the seven severity metrics gave useful insights into owner-perceived importance of different disorders, by reflecting differential levels of drive to present affected dogs, take up recommended interventions and attend for ongoing or specialist care.
Overall welfare impact and suggested priority areas
Dental disorder, osteoarthritis, overweight/obese and lipoma ranked highest among the 8 studied disorders based on VetCompass Welfare Impact scores. Top-ranking duration and severity scores, interpreted alongside individual severity metric sub-scores characterised osteoarthritis as a relatively frequent primary reason for veterinary presentation, with multiple related visits per year, frequent association with chronic pain management and a notable level of associated referral. This evidence supports osteoarthritis as a priority disorder for targeted reform [42].
Recent surveys by the British Veterinary Association (BVA) and Peoples Dispensary for Sick Animals (PDSA) reported that UK veterinary surgeons consider obesity/excessive bodyweight [29, 49, 51] and dental disorders [50] to be the most important health related welfare issues currently affecting UK dogs. Both scored highly for prevalence and duration, but severity sub-scores appeared to indicate mismatches in veterinarian-owner perceptions of their individual welfare importance. This may also highlight dental disorders and overweight/obese as areas where increasing owner awareness of clinical signs, primary consequences and secondary risks of associated sequelae could improve health-related welfare in a significant proportion of UK dogs [65, 66].
Numerical over-representation of individual breeds within disorder case groups
A number of breeds were numerically over-represented in every disorder group. Assessing breed over-representation within a disorder by comparing breed-specific annual period prevalences with overall period prevalence was intended to avoid highlighting spurious potential disorder-associations in generally popular breeds (i.e. some breeds may present more with dental disorders simply as a function of their popularity within the population, but are not numerically over-represented within the randomly selected case group studied). The authors acknowledge that these comparisons were only undertaken for the most commonly represented breeds in each disorder (represented by > 5 individuals), thus less likely to flag-up potential disorder associations in less common breeds. However it is important to stress that these studies were based on relatively small samples and were likely underpowered to identify all true breed associations regardless of this approach. Exploration of individual breed-disorder predispositions was not a primary objective of this study but rather to establish evidence for breed associations in general, in support of disorders that may have an inherited element. It remains interesting to note that the majority of breed-disorder over-representations seen were consistent with previously reported breed-disorder associations, including otitis externa in Cocker spaniels [67, 68], osteoarthritis in Labrador and Golden retrievers [69, 70] and dermatitis in Bulldogs, Boxers, Pugs and German shepherd dogs [47, 57]. As such the findings of the present study appear to support existing evidence for established breed-disorder associations and could direct future predisposition investigations using larger sample sizes [71].
Limitations
This study aimed to develop and apply a standardised methodology for disorder assessment to maximise comparability of prevalence, duration and severity across disorders, but individual disorder assessments and comparisons remain sensitive to certain ‘critical control points’. As previously discussed, these include selection criteria for disorders to be assessed, specificity of case definitions, inclusion criteria, subjective interpretation of available clinical free text, availability of all required data for parameter estimation, criteria for assigning severity metric sub-scores and methodology for annual duration estimation. Nonetheless, the primary comparisons aimed to establish relative measures and scores rather than absolute numerical indices of health-related welfare. Defining SEHB level 2 in at least one breed as sufficient evidence to consider a disorder ‘breed-related’ did not constitute strong evidence for genetic inheritance/hereditability. However, in this study the interest was in comparing disorders with a tendency for association with certain breeds for various, not necessarily solely genetic, reasons. For example, environmental, epigenetic and socioeconomic factors may also be relevant if certain breeds tend to be owned by less experienced or affluent dog owners, those whose lifestyle/housing offers fewer opportunities for exercise, or are inclined to feed their pets in particular ways. Hence, it was not critical to seek the higher levels of evidence for genetic inheritance which remain relatively scarce within the available literature on canine genetics [14, 18, 19].
Broad diagnostic specificities applied within some disorder groupings may fall short of the fine-grained level achievable in referral or smaller-scale prospective studies. Some case definitions reflected common approaches in primary care practice, rather than confirmed diagnoses, e.g. lipoma diagnosis without pathological supporting evidence or osteoarthritis diagnosis without radiographic imaging. However, these wider case definitions better reflect the typically varied level of diagnostic certainty available to clinicians in the primary care setting and allowed researchers to harness the unique benefits of analytical scale offered by the VetCompass database.
Using data available within VetCompass in 2013 introduced certain potential limitations to the generalisability of study findings to all UK primary care practices. The study included a mix of private and corporate but with a predominance of corporate practices. This may have skewed the results towards the standards and policies of clinical management and care that are typical of these corporate practices. Additionally, the study represents the demographic breed structure in dogs in 2013. This structure of the UK veterinary dog population may change over time with consequent changes in the frequency and nature of the most common disorders. The age of the dataset could also mean that certain topical/relevant breed-disorder associations change, e.g. brachycephalic breeds are becoming increasingly popular in the UK. These breed types have strong, well-documented disorder associations that would therefore be expected to elevate the frequency of these disorders in the overall dog population over time. Future studies analysing more recent VetCompass datasets can provide valuable scope to reflect and compare current caseloads within both private and corporate veterinary settings.
All welfare metric scoring in this study was designed for disorder (rather than individual) level assessment, thus could not be used to reflect ranges in severity between individual cases of the same disorder. The authors recognise that the range of individual case severity within-disorder is an important area of research, and could provide opportunities to examine relative severity of a disorder between defined sub-groups of affected dogs (e.g. individual breeds). Appropriate re-framing of the proposed severity scoring system, or modification of published severity scoring systems such as the GISID [14] for practical application to VetCompass EHR data at individual dog level would expand the scope of future VetCompass studies to describe welfare impact at both individual, group and overall disorder levels. It should also be acknowledged that the current work focused on health related welfare impacts only and did not attempt to evaluate other elements of the animals’ welfare. Further, though the study identified seven specific metrics for assessment, it is acknowledged that not all of these might apply equally to all conditions. As such, the aim of the study was to offer transparent and consistent severity assessments. By providing the individual metric scores in addition to their combination, there remains the scope for those interpreting these scores to focus on one or more individual metrics, rather than all measures as deemed appropriate.
Generating an overall welfare impact score by summing selected indicators of prevalence, severity and duration may seem to contradict the published concept of population welfare impact as the product of these three key concepts [35, 36]. Taking a summative approach was considered a mathematically valid alternative, and allowed the inclusion of a zero score within the severity scoring system (to maintain distinction between none and few cases contributing data of relevance within individual severity metrics). While differential weighting of individual contributor factors was avoided here, a summative approach to welfare impact scoring still offers scope for mathematical weighting should future users wish to emphasise the importance of particular contributing factors on welfare impact score.
On balance, this large-scale epidemiological analysis of primary-care EHRs provides valuable evidence, complementary to that from more detailed, smaller scale studies undertaken prospectively or within referral/specialist caseloads. Specific controversies over disorder definitions or researcher judgements should not detract from the applicability of the disorder assessment approach across a range of conditions. Transparency of case definitions, data availability and criteria for individual metric calculation are key to the valid interpretation of findings from future applications of these methods.