Proportion and risk factors for death by euthanasia in dogs in the UK

The loss of a pet can be particularly distressing for owners, whether the method of death is euthanasia or is unassisted. Using primary-care clinical data, this study aimed to report the demographic and clinical factors associated with euthanasia, relative to unassisted death, in dogs. Method of death (euthanasia or unassisted) and clinical cause of death were extracted from a random sample of 29,865 dogs within the VetCompass Programme from a sampling frame of 905,544 dogs under UK veterinary care in 2016. Multivariable logistic regression modelling was used to evaluate associations between risk factors and method of death. Of the confirmed deaths, 26,676 (89.3%) were euthanased and 2,487 (8.3%) died unassisted. After accounting for confounding factors, 6 grouped-level disorders had higher odds in euthanased dogs (than dogs that died unassisted), using neoplasia as the baseline. The disorders with greatest odds included: poor quality of life (OR 16.28), undesirable behaviour (OR 11.36) and spinal cord disorder (OR 6.00). Breed, larger bodyweight and increasing age were additional risk factors for euthanasia. The results highlight that a large majority of owners will face euthanasia decisions and these findings can support veterinarians and owners to better prepare for such an eventuality.


Method of managing the remains.
Of the euthanased dogs, the method of managing the remains was not recorded in 3740 (14.0%) and the remains of 23 (0.09%) dogs were managed by means other than cremation or burial, such as those donated to science or used for taxidermy. Of the dogs that died unassisted, the method of managing the remains was not recorded in 674 (27.1%) dogs and the remains of 1 dog (0.04%) were managed by means other than cremation or burial. The aforementioned dogs were excluded from further analysis. Based on a chi-squared test, there was a significant difference between the proportions of dogs buried, individually cremated or communally cremated between dog that were euthanased and dogs that died unassisted (p < 0.001). (Fig. 1).

Risk factors for euthanasia compared with unassisted death.
Univariable logistic regression modelling for risk factors for euthanasia compared with unassisted death in dogs identified seven variables that were liberally significant and were carried forward for multivariable modelling: Breed, Bodyweight (kg), Age at death (years), Sex-Neuter status, Insurance status, Vet Group and Grouped-level disorder. The final multivariable model retained five variables: Breed, Bodyweight (kg), Age at death (years), Vet Group and Grouped-level disorder ( Table 2).
One breed showed increased odds of euthanasia compared with Labrador Retrievers:

Discussion
This is the first study to report proportional death rates for dogs by euthanasia versus unassisted in the population of animals under primary veterinary care in the UK. The study additionally reports on demographic and clinical factors associated with euthanasia relative to unassisted death. Much of the previous literature focused on risk factors for euthanasia in animals in shelters, particularly in the US, limiting the generalisability to the UK dog population under primary care [17][18][19][20] . Instead, the current study included deaths in 29,163 dogs under primary care with the aim of getting a deeper understanding of the processes around the deaths of dogs in the wider owned dog population. The results showed that the majority of deaths in dogs involved euthanasia, with 26,676 (91.5%) dogs euthanased compared with 2487 (8.5%) that died unassisted. It therefore provides novel insights in to the demographic and clinical factors associated with euthanasia relative to unassisted deaths, providing an evidence-base for veterinarians on the aspects influencing euthanasia and unassisted deaths in dogs. In turn, www.nature.com/scientificreports/ this will enable veterinarians to use benchmark data collected from this broad range of UK veterinary practice to support owners with an evidence-base to assist their decision-making process.
The results support the study hypothesis that the odds of euthanasia relative to unassisted deaths in dogs rises with aging. Increasing age was associated with an increased risk of euthanasia, with dogs aged ≥ 15 years at 5.92 times the odds of euthanasia, relative to an unassisted death, compared with those < 6 years. Previous studies have cited "old age" as a major reason for euthanasia 21,22,30 , although to the authors' knowledge, this risk relative to unassisted death has not been quantified previously. Given that age is the greatest risk factor not only for the probability of death, but also for the majority of morbidities associated with mortality 31-34 , the increased risk of euthanasia relative to unassisted death identified in aging dogs in the current study is not unsurprising.   www.nature.com/scientificreports/ However, this finding does highlight the increasing importance of quality of life (QOL) and euthanasia decisionmaking discussions between veterinarians and owners as dogs age. Decision-making in older animals may be more likely to include euthanasia as an option, as has been reported in a previous study based on dogs and cats diagnosed with diabetes mellitus 35 . Conversely, veterinarians may be more reluctant to agree to euthanasia for younger, treatable patients 14 . This reflects the current study findings, given younger dogs were at greater risk of an unassisted death compared with euthanasia. Although euthanasia may be used to relieve animals from suffering, their health and welfare prior to this decision should still be considered and suffering minimised. It is of concern that in a previous study, 69% of dogs aged over 10 did not receive veterinary care during the 18 months leading up to euthanasia 36 . Promoting the welfare of geriatric patients is of high priority, including palliative veterinary care where appropriate 37 . Disorders with greatest probability of euthanasia compared with an unassisted death (using neoplasia as the baseline) included: poor quality of life, undesirable behaviour and spinal cord disorder. Improved nutrition and healthcare have contributed to extended canine lifespans, therefore maintaining QOL is a growing concern in companion animal practice 38 . QOL in pets has been defined as states of comfort or discomfort representing a combination of physical and non-physical factors 39,40 . However, QOL is often assessed by owners and is somewhat subjective 38 . More recently, QOL tools have been developed to help optimize and standardize euthanasia decisionmaking in pets 38,41 . Given that poor QOL was identified as a strong risk factor for euthanasia (OR 16.28), this suggests that veterinary professionals and owners are considering QOL as a determining factor when deciding on euthanasia. Although, the current study did not extract information on proportional usage of validated QOL tools in these considerations. This could be a useful area of future research. The use of validated QOL tools could assist veterinary professionals and owners to better recognise and manage functional decline as canine patients age 38 . Additionally, guidelines produced by the animal hospice movement 42 may prove useful in assessing QOL and, in turn, predicting quality of death 8 . The magnitude of the odds ratio for poor QOL is not surprising, given poor quality of life is associated with discomfort 39,40 , and therefore veterinarians and owners may be more likely to opt for euthanasia rather than allow the dog to die unassisted.
The current study identified undesirable behaviour as a significant risk factor for euthanasia relative to unassisted death (OR 11.36). Behavioural problems are considered an important factor in the euthanasia of dogs 2,43,44 , with previous reports suggesting behavioural problems account for 2-39% of canine euthanasia in veterinary practices 21,45,46 and 50-70% in animal shelters 43,47 . The current study identified undesirable behaviour as the fourth most common reason for euthanasia, accounting for 7.3% of euthanasia deaths. Both genetic and environmental factors influence behavioural development and there is growing evidence to suggest that educating owners about puppy-raising practices, and the provision of ongoing socialisation and habituation, can reduce the incidence of problem behaviour 48 . Steps to reduce the incidence of problem behaviours may not only improve the welfare of individual dogs (if problems are associated negative emotional states and/or inappropriate punishment from caregivers), but may also reduce the number of dogs euthanased due to behavioural issues. It might therefore be apt for veterinary professionals to prioritise behavioural discussions at initial puppy consultations. Veterinary practitioners often lack confidence in dealing with behavioural problems, which has been attributed to poor coverage of the subject in veterinary education 49 . The current study findings support the suggestion for improved provision of behaviour medicine in veterinary education 49 . It should be noted that very few dogs with an undesirable behaviour will die naturally, reflected in the finding that only 7 dogs in the current study died unassisted of an undesirable behaviour compared with 1679 dogs that were euthanased. Therefore, absolute values as well as the relative odds ratio should be considered when interpreting these results. Additionally, behaviours may be labelled as undesirable by the owners, and what is considered undesirable for one owner might be acceptable for another 44 . Therefore, owner perception of the desirability of any specific behaviour may affect the likelihood for euthanasia.
Spinal cord disorder was identified as a significant risk factor for euthanasia relative to unassisted death in the current study (OR 6.00). Disorders of the spinal cord include congenital defects, degenerative diseases, inflammatory and infectious diseases, tumours, injury and trauma and vascular diseases 50 . Typical age at onset and presentation vary according to aetiology, but diagnosis of spinal cord disorders often involves advanced imaging such as computed tomography (CT) or magnetic resonance imaging (MRI) 50 . CT and MRI scanning are complex and expensive procedures 51 , therefore may not be accessible to all owners. If advanced imaging and possible subsequent surgery are required to treat a spinal cord disorder, it might be that an owner opts for euthanasia if the diagnostic and treatment options are not financially viable and there is not a realistic alternative to protect their dog's QOL. Such data were not available in the current study, but suggest an important area for future research.
The disorders with lowest odds of euthanasia (and therefore the highest odds of unassisted death) were traumatic injury, disorder not recorded, complication associated with clinical care and heart disease. Traumatic injuries are a common emergency presentation 52 , with a US study reporting prevalence of traumatic injury at two large university veterinary hospitals as approximately 13% 53 . The most common causes of trauma in the US study were motor vehicle accident (53.2%), unknown cause (12.2%) and animal interaction (11.1%) 53 . The current study included deaths that occurred within or external to a veterinary practice, which may in part account for why euthanasia risk was lower in these dogs (and hence risk of unassisted death higher) as many dogs may have died at the time of injury, during transport or on arrival to a veterinary practice or at home.
Dogs without a cause of death recorded had reduced odds of euthanasia (OR 0.08) compared with neoplasia. Since euthanasia must be performed by a veterinary surgeon, or by a person who has been authorised to do so by a veterinary surgeon 54 , it follows that euthanased dogs are more likely to have a disorder diagnosis prior to or on presentation for euthanasia. Bereaved pet owners report that they wanted to explore all possible treatment options before considering euthanasia 55 . Dogs that died unassisted may have died at home, at an emergency out of hours clinic or may not have visited a veterinary clinic for a specific problem prior to death and hence might not have a disorder diagnosis. www.nature.com/scientificreports/ Complications associated with clinical care in dogs resulted in reduced odds of euthanasia (OR 0.08) compared with neoplasia. The prevalence, however, was relatively low with 0.2% dogs euthanased and 2.1% dogs dying unassisted due to complications. This highlights the value of considering both the absolute (prevalence) as well as the relative (odds) values when interpreting epidemiological results. "Complications" may have encompassed a range of specific factors, therefore future studies may help evaluate this finding further to determine if there are specific complications more likely to result in euthanasia or unassisted death and whether there are preventative measures that could be implemented, such as those designed to reduce postoperative complications in elderly human patients 56,57 .
Heart disease as a cause of death showed reduced odds of death by euthanasia (OR 0.14). There were 4.6% (1047) euthanasia cases attributed to heart disease compared with 19.7% (251) unassisted deaths. Cardiac disease is often subclinical, non-fatal or chronic 58 , therefore it may be that a proportion of the euthanased dogs died with heart disease (whether diagnosed or undiagnosed) rather than this being the definitive reason for euthanasia. Heart disease was the most common cause of unassisted death in dogs, highlighting management of this condition as important for veterinarians to discuss with owners of affected dogs. For example, one important area for such discussions may be the impact of breathlessness on quality of life; this clinical symptom is reported as having a profound impact on quality of life in human patients 59 .
After accounting for other confounding factors such as grouped-level disorder and age, Rottweilers were more likely to be euthanased (OR 1.76) than to die an unassisted death, compared with Labrador Retrievers, whilst seven breeds were more likely to die from unassisted deaths including Bulldog, Pug, Cavalier King Charles Spaniel and Yorkshire Terrier. The reasons for the breed risks identified are likely multifactorial, with possible differences including the risk of sudden death between breeds, that would preclude euthanasia. Heart disease has previously been identified as the most common cause of sudden and unexpected death in dogs 60 . In a recent VetCompass study, Yorkshire Terriers and Cavalier King Charles Spaniels had a significantly increased odds of degenerative mitral valve disease diagnosis compared with crossbred dogs 61 , a cardiac disorder where 50% of disorder-related deaths are considered 'sudden' 62 .
Cause of death terms were taken in to account in the multivariable modelling, therefore the differences in breed risks identified may result from other factors that the model could not account for. Such factors might include owner characteristics e.g. attitudes towards pet death, and the dog-owner relationship. The finding that brachycephalic breeds including the Bulldog and Pug were at increased risk of unassisted death is of interest, given their current boom in popularity 63 . Although further research is needed, results of recent studies exploring the ownership behaviours and beliefs of owners of brachycephalic breeds may indicate that owning this breed type has the potential to influence when, and if, a dog is euthanased. Owners of brachycephalic breeds (specifically Bulldogs, Pugs and French Bulldogs) have been reported to form particularly strong dog-owner relationships 64 . However, the aforementioned study did not compare brachycephalic dogs to non-brachycephalic dogs, therefore the strength of the relationship cannot be quantified. Previous studies have found strength of the dog-owner bond is related to health-seeking behaviours in dog owners (e.g. owners with strong bonds seeking higher levels of veterinary care and being more likely to follow veterinary recommendations regardless of cost) 65 , research on brachycephalic dog owners reveals disparities in their perceptions of dog health versus those of veterinary professionals; these differences may disrupt tendency to follow veterinary advice that is shown by owners of dogs in general. Owners of brachycephalic breeds tend to 'normalise' poor health in their breed. Although they may be aware that their dog is showing signs of respiratory disease, for example, they may not consciously accept that this is a real problem for their dog, and instead attribute it as a 'normal' feature of their breed 66 . This phenomenon has been found to extend beyond wakeful respiratory dysfunction, and to also cover dysfunctional sleeping, thermoregulation and eating that are normalised in these breeds, such that dogs need to reach a critical level of clinical severity before owners consciously acknowledge their dog has a 'problem' 64 . These normalisation and thresholding phenomena may affect euthanasia-decision making, whereby owners may fail to perceive their dog as 'unwell' and do not consider their dog's quality of life as sufficiently impaired to justify euthanasia. Consequently, some severely affected dogs are more likely to die an unassisted death.
Bodyweight was a significant predictor of death by euthanasia, with dogs 20 to < 30 kg at 1.24 times the odds of euthanasia, relative to an unassisted death, compared with dogs < 10 kg. A previous study evaluating euthanasia or rehoming in dogs with behavioural issues reported an association between heavier bodyweight and increased risk of rehoming or euthanasia and/or the owners considering rehoming or euthanasia 67 . The reasons behind this association were not discussed in detail, however there is a disproportionate risk of injury associated with larger and/or more physically powerful breeds, as well as the existence of breed stereotypes, which may have contributed to this previous association 68 . Similarly, the precise reasons for differential euthanasia across weight categories in the current study are not clear. However, management considerations, such as the financial cost of treating larger dogs and difficulties in assisting with end of life care, might contribute to the association identified, although this has not been previously explored.
A greater proportion of euthanased dogs were communally cremated compared with dogs that died unassisted, whilst a greater proportion of dogs that died unassisted were individually cremated compared with euthanased dogs. In a previous study, Chur-Hansen et al. (2011) found that owners needed memorials, such as an animal's ashes, to help them move through the grieving process, with some owners stating it felt as if their animal was not completely gone 69 . It is possible that owners whose dogs died unassisted may have desired to keep their dog's remains (either through individual cremation or home burial) to aid in processing the grief of their loss. Conversely, owners of dogs who were euthanased may have benefited from a longer time processing the loss of their dog prior to their death, compared with those whose dog had an unassisted death, and therefore might not have felt as inclined to keep their dog in close proximity through home burial or retaining of ashes. Financial considerations may also play a part. Dogs that were euthanased may have been receiving prior treatment at a veterinary www.nature.com/scientificreports/ clinic, therefore owners may have opted for a less expensive method of managing remains following the cost of treatment. Conversely, dogs that died unassisted may not have been receiving costly treatment prior to death. Decision-making about euthanasia involves the veterinarian and the animal owner(s). The veterinarian must shift focus from trying to cure the animal to admitting that there is nothing further that can be done, then convincing the client that the time has come for euthanasia 70 . Animal owners, in turn, appreciate the support of a veterinarian when making the decision for euthanasia in cases where treatment options have been exhausted 55 . The current study focuses on animal-related factors for euthanasia. However, client-related factors such as caregiver burden and financial constraints may lead to euthanasia decisions being made on grounds other than the animal's quality of life and which could be the subject of future studies 71 .
This study has many similar limitations to previous studies based on retrospective primary-care data 2 . A large proportion of dogs that died unassisted did not have a cause of death reported (48.8%). Although a large proportion of these data were missing, dogs without a cause of death recorded were still included in the analysis to reduce bias. Some of the disorder groupings in the study might overlap, such as neoplasia and mass. The authors did not make any diagnostic assumptions i.e. disorders were recorded according to the attending veterinarian's most specific diagnosis. Misclassification bias is possible in this instance, but may have been even greater if diagnoses are assumed. Euthanasia is a decision-making process involving owners, however the same cannot always be said for unassisted death and so this distinction should be accounted for when interpreting the results.

Conclusions
The demographic and clinical factors associated with euthanasia and unassisted deaths were identified, which could be used by veterinarians to better understand the aspects influencing the euthanasia decision-making process in dogs. Poor quality of life and undesirable behaviour were major risk factors for euthanasia in dogs. Conversely, dogs with traumatic injury, complication associated with clinical care and heart disease showed increased risk of unassisted death. Older and heavier dogs also had higher odds of euthanasia. These findings could inform future studies, particularly qualitative research studies that might be designed to evaluate quality of life discussions, to develop measures to consider the appropriateness of palliative care for specific conditions and to assess the impact of euthanasia decision-making on owners and veterinary professionals.

Methods
The study included all available dogs under primary veterinary care at clinics participating in the VetCompass Programme during 2016. Dogs under veterinary care were defined as those with either a) at least one electronic patient record (EPR) (VeNom diagnosis term, free-text clinical note, treatment or bodyweight) recorded during 2016 or b) at least one EPR recorded during both 2015 and 2017. VetCompass collates de-identified EPR data from primary-care veterinary practices in the UK for epidemiological research 29 . Data fields available to VetCompass researchers include a unique animal identifier along with species, breed, date of birth, sex, neuter status, insurance status and bodyweight, and also clinical information from free-form text clinical notes, summary diagnosis terms 72 and treatment with relevant dates.
A cohort study design was used to estimate proportional euthanasia among deaths in dogs under primary veterinary care in the UK during 2016 and to report on demographic and clinical risk factors associated with euthanasia (relative to unassisted death). Sample size calculations in Epi info (CDC) estimated that approximately 185 dogs that were euthanased and 185 dogs that died unassisted would be required, to identify if dogs ≥ 6 years had at least twice the odds of euthanasia (relative to unassisted death) compared to dogs < 6 years, assuming 80% of dogs ≥ 6 years that die are euthanased, 80% power and 95% confidence 73 . Ethics approval was obtained from the RVC Ethics and Welfare Committee (reference number SR2018-1652).
Candidate death cases were identified using search terms appropriate to euthanasia and unassisted death in the clinical notes (euth*, pts*, crem* ashes, pento*, casket, beech, decease*, death, "put to sleep", doa, died, killed, "home bury" ~ 1, "bury" and "home") and in the treatment fields (euth*, pento*, crem*, casket, scatter, beech). The search findings were merged and a random selection of these was then manually reviewed in detail to identify dogs that died during 2016 74 . Death did not have to take place at the attending veterinary clinic, because VetCompass also captures information reported by owners to their veterinary clinics on deaths that occurred outside the clinic setting.
Method of death and cause of death were defined as two distinct entities. Method of death was categorized as euthanasia or unassisted, with missing data recorded as "Not recorded" and excluded from the statistical analysis. Cause of death described the main or first reported biomedical cause for the death regardless of the method of death. Cause of death terms were mapped to a grouped-level of diagnostic precision. Grouped-level terms mapped the original diagnosis terms to a general level of diagnostic precision (e.g. inflammatory bowel disease would map to gastro-intestinal) as previously described in the literature 74 . In order to maintain sufficient power for analysis, a combined disorder list from all of the 15 most common grouped disorders from dogs that were euthanased and dogs that died unassisted was generated and compared in the analysis. Dogs without a specific cause of death recorded were categorised and included in the analyses as "Not recorded". In addition, method of dealing with the remains was reported and categorised as "Communal cremation", "Individual cremation", "Burial", "Other" or "Not recorded".
Breed information entered by the participating practices was cleaned and mapped to a VetCompass breed list derived and extended from the VeNom Coding breed list (The VeNom Coding Group, 2019). In order to maintain sufficient power for analysis, the breed variable included specific breeds with at least 20 dogs in either the euthanasia or unassisted death grouping respectively. Remaining dogs were grouped in to "Purebred-Other" and "Crossbred". Neuter status was defined by the final available EPR neuter value and was combined with sex to create four categories: female entire, female neutered, male entire and male neutered. Bodyweight was defined www.nature.com/scientificreports/ as the bodyweight (kg) value closest to the date of death for each dog. Bodyweight (kg) was categorised: < 10, 10 to < 20, 20 to < 30 and ≥ 30. Age (years) was defined at the date of death and was categorised: < 6.0, 6.0 to < 9.0, 9.0 to < 12.0, 12.0 to < 15.0 and ≥ 15.0. Veterinary group attended was categorised as 1-5, based on 5 practice groups involved in the study. Insurance status was categorised as insured or not insured. Missing data were recorded as "Not recorded" and included in the analysis if this category accounted for > 10% of the study variable. Following data checking for internal validity and cleaning in Excel (Microsoft Office Excel 2013, Microsoft Corp.), analyses were conducted using SPSS version 24.0 (IBM Corp). The proportion of deaths by euthanasia and unassisted were reported. All continuous variables were nonnormally distributed and so were summarised using median, interquartile range (IQR) and range. Mann-Whitney U test, chi-square test and Fisher's exact test were used as appropriate for comparison of demographic data between euthanasia cases and unassisted deaths 75,76 . Binary logistic regression modelling was used to evaluate univariable associations between risk factors (grouped-level disorder, breed, bodyweight, age, sex-neuter and insurance) and outcome. Outcome was defined as either euthanasia or unassisted death. Dogs with missing data on the method of death were recorded as "Not recorded" and excluded from further analysis. Euthanasia was the event of primary interest, therefore the model reported the odds of euthanasia relative to unassisted death for each category within a variable. Neoplasia was used as the baseline in the grouped-level disorder variable, since neoplasia has been reported as the most common cause of death in dogs under primary veterinary care 2 . Labrador Retriever was used as the baseline in the breed variable, rather than crossbreed which has been common in previous studies 77 , since Labrador Retrievers are more standardised with regard to genetics, bodyweight, skull shape and conformation compared to crossbreeds 78 .
Risk factors with liberal associations in univariable modelling (P < 0.2) were taken forward for multivariable evaluation. Model development used manual backwards stepwise elimination. Vet Group attended was evaluated as a fixed effect. Potential confounders were assessed by checking for a marked change in the odds ratio (OR) after removal of the variable from the model. Collinearity was investigated by examining the variance inflation factor (VIF) and tolerance, with collinearity indicated if VIF > 10 and tolerance < 0.1 79,80 . The area under the ROC curve and the Hosmer-Lemeshow test were used to evaluate the quality of the model fit. Statistical significance was set at the 5% level.
Ethics approval. Ethics

Data availability
The datasets generated during and/or analysed during the current study are available at the RVC Research Online repository: https:// resea rchon line. rvc. ac. uk/ id/ eprint/ 13486.