Abstract
Objectives
The objective was to evaluate factors associated with euthanasia in an animal shelter in Kitchener-Waterloo, Ontario, Canada.
Methods
Data from 3737 cats admitted to the shelter between January and December 2011 were evaluated.
Results
Overall, 1989/3737 (53%) of admitted cats were euthanized. Male cats had greater odds of being euthanized than females (odds ratio [OR] 1.63, 95% confidence interval [CI] 1.29–2.05; P <0.001) and surrendered cats were more likely to be euthanized than strays (OR 38.0, 95% CI 14.8–97.69; P <0.001). Black cats were more likely to be euthanized than cats of another color (OR 1.45, 95% CI 1.16–1.80; P <0.001). Cats that spent >5 days in the shelter were more likely to be euthanized than those that spent <5 days in the shelter (OR 1.57, 95% CI 1.25–1.97; P <0.001). Cats that spent >20 days in the shelter were less likely to be euthanized than those that spent <5 days in the shelter (OR 0.26, 95% CI 0.19–0.34; P <0.001). Age, an age quadratic term, neuter status and interactions among these variables were statistically significant; the odds of unneutered animals being euthanized was high and relatively stable across age groups, but in neutered animals the odds of being euthanized increased with age before plateauing in older cats.
Conclusions and relevance
With >50% of the cats admitted to the shelter in 2011 euthanized, it is important to understand the contributing risk factors that predispose shelter cats to euthanasia and what changes can be made to the shelter system and in owner education to lower the incidence of euthanasia.
Introduction
Approximately 6–8 million dogs and cats are brought into animal shelters in the USA every year.1,2 Animals that enter shelters are either reclaimed by their owners, adopted, die as a result of illness or are euthanized. 1 Other animals may be transferred to other shelters or rescue organizations. In the USA alone, almost $1 billion are spent annually by animal shelters to care for and subsequently euthanize animals that are unable to be adopted. 3
Every year in the USA, 10–25% of the companion animal population admitted to shelters is euthanized. 1 One study noted that as many as 57% of the cats admitted to shelters in Michigan in 2003 were euthanized. 4 Because of the large number of admitted animals, combined with finite numbers of potential homes and finances, shelters often must assess animals when they are admitted for physical characteristics, behavior and health status, in part to determine whether the animal should be made available for adoption. 5 While many animals are deemed unadoptable and euthanized, a large number of potentially adoptable animals are also ultimately euthanized. A variety of factors may influence this outcome, including shelter capacity, animal health and preferences of prospective owners for certain pet characteristics, such as color, age, breed and size.1–6 Studies by Lepper et al 1 and Fantuzzi et al 6 investigated what characteristics were desirable to adopters, to understand the probability of an animal being adopted or euthanized and to give shelters strategies to promote all different types of animals as companions. It was found that younger cats, neutered males, white- and gray-colored cats and rare breeds, such as Siamese cats, were more likely to be adopted than euthanized. More research and information on the preferences of adoptive owners may help shelters understand and improve their strategies to adopt animals that are currently perceived to be less desirable.
Despite extensive research and exploration, information about factors that influence adoption and euthanasia is still limited, hampering efforts to increase adoption and decrease relinquishment and euthanasia, while optimizing shelter resources. There is a gap in the literature concerning how characteristics such as age, sex, breed, source, total time in the shelter, color, altered status, jurisdiction, time of year and number of times in the shelter contribute to the probability of an animal being euthanized. The objective of this study was to evaluate associations between euthanasia of cats admitted to an animal shelter and these characteristics.
Materials and methods
Cats admitted to the Kitchener-Waterloo Humane Society in Kitchener, ON, Canada between January and December 2011 were studied through retrospective evaluation of data from the shelter’s electronic records system. All cats admitted to the shelter during this period were followed. This shelter has open admission and foster programs are in place to increase the amount of animals admitted. Factors such as sex, altered status (castrated/spayed, intact), source, jurisdiction of origin, age, total time in the shelter, coat color, the last season the cat was kept by the shelter, number of times in the shelter and breed were recorded. The variable ‘times in shelter’ was dichotimized as ‘re-admission’ and categorized as either ‘yes’ or ‘no’. The variable ‘last season the cat was kept by the shelter’ was defined as the last reported season when the animal was present in the shelter prior to re-admission. Seasons were defined as follows: winter (December–March), spring (April–May), summer (June–August) and fall (September–November).
The proportion of animals that were euthanized and the proportion of animals among categories of the following independent variables was estimated: sex (male, female, unknown), breed (purebred, non-purebred), source (stray, surrender, euthanasia request, shelter offspring), total time in shelter (<5 days, 5–20 days, >20 days), color (black, other), altered status (yes, no, unknown), jurisdiction (urban [Kitchener-Waterloo area], out of area/unknown, small town/rural [areas around the Kitchener-Waterloo area]), the last season the cat was kept by the shelter (winter, spring, summer, fall) and number of times admitted to the shelter (re-admitted, not re-admitted). The proportion of animals euthanized among the categories of these independent variables and their respective exact 95% confidence intervals (CIs) were also estimated.
Univariable logistic regression models were constructed to estimate the association between a cat being euthanized and the independent variables listed above. Continuous independent variables were assessed as to whether the variable had a linear relationship with the log odds of the outcome using a lowess curve (ie, locally weighted regression). If the lowess curve indicated there was a quadratic relationship, the square term of the variable was included with its main effect in subsequent models. However, if the relationship was non-linear, but not quadratic, a transformation to linearize the relationship was performed or the variable was categorized. Prior to multivariable logistic regression modeling, the correlation was assessed among independent variables using various correlation coefficients to avoid issues associated with collinearity. Correlations >80% (ie, r >|0.8|) were considered problematic and only the variable with the most complete information was included in subsequent multivariable modeling.
All significant variables based on a liberal P value (ie, P <0.20) in the univariable analysis were considered for inclusion in the multivariable model. A multivariable model was then fitted using a manual backwards step-wise approach to create a main effects model using a significance level of alpha (α) = 0.05 while retaining confounding variables regardless of statistical significance. Confounding was assessed by examining the change in the coefficients for the remaining statistically significant variables in the model, once the potential confounding variable was removed. If the coefficient for one of these variables changed >20%, the removed variable was considered a confounder, assuming it did not appear to be an intervening variable based on a logical causal pathway, and retained in the main effects model. Next, interactions between the following variables were then assessed based on their biological plausibility: sex and breed, breed and age, sex and age, total time in shelter and age, season and age, and reproductive status and age. Variables were retained in the final multivariable model if they were statistically significant (α = 0.05), acted as a confounding variable, or were part of a statistically significant interaction term. The statistical significance of categorical variables, including interaction effects involving categorical variables, was assessed using likelihood ratio tests. Predicted curves were generated for any significant interaction terms involving continuous variables to interpret the relationship between the independent variables and the probability of euthanasia. For any categorical variables that had multiple levels (eg, season), or interactions involving categorical variables, contrasts were generated to assess pairwise differences among categories of a variable or different covariate patterns among interacting variables.
Model fit was assessed using a Hosmer–Lemeshow goodness-of-fit test. Pearson residuals, leverage, delta χ2, delta-beta and delta deviance graphs were assessed visually to identify any potential outliers or highly influential observations. If there were observations that fit the model poorly or had extreme influence on the model, recording errors were investigated and the model was refit without the observation to determine its impact on the interpretation of the model. All statistical analyses were performed using Stata 13.0 (StataCorp LP, College Station, TX, USA).
Results
In total 5439 animals were admitted to the shelter in 2011; 3737 (69%) were cats. Only animals that had sufficient data were included in the statistical analyses. Animals were dropped from the final analyses if they did not have sufficient data with respect to the variables that were included in the final multivariable model. In total, 3435 cats were included in the final multivariable model. Approximately, 48% of cats were female, 45% were male and the remaining 8% were of an unknown or unrecorded sex. At the time of admission 39% of cats were altered, 18% were not and the status of 43% were unknown. The cats admitted to the shelter came from various sources: 78% were strays, 11% were surrendered to the shelter, 8% were euthanasia requests and 3% were shelter offspring. This shelter received admissions from various jurisdictions, and 88% of admitted cats were from urban/suburban areas in Kitchener-Waterloo, 3% came from out of the area or were from an unknown jurisdiction and 9% came from smaller towns or rural areas in and around the Kitchener-Waterloo area. Of the cats admitted to the shelter, 28% were black and 72% were a variety of other colors. Various breeds of cat were admitted into the shelter, but 97% were non-purebreds such as domestic short-, medium- and longhair cats. Twenty-three percent of cats left the shelter (adoption, reclamation or euthanasia) in winter, 21% in the spring, 24% in the summer and 31% in the fall. Ninety-eight percent of cats that were adopted or claimed were not re-admitted during the study period, while 2% were re-admitted to the shelter. Overall, 1989 (53%) of these cats were euthanized. The incidence of euthanasia by each independent variable is presented in Table 1. No disease outbreaks were identified during the study period.
Descriptive statistics of categorical variables associated with euthanasia among cats from a shelter in Kitchener-Waterloo, Ontario (2011) (n = 3737)
CI = confidence interval
Sex, age, age squared, source, total time in shelter, color, altered status, jurisdiction, the last season the cat was kept by the shelter and the number of times the cat had been in the shelter were significantly associated with euthanasia in the univariable analysis (Table 2).
Results of univariable logistic regression models concerning demographic characteristics and other factors that were associated with euthanasia among cats from a shelter in Kitchener-Waterloo, Ontario (2011)
OR = odds ratio; CI = confidence interval
The following variables and interaction terms were included in the final multivariable model: sex, age, age squared, source, total time in shelter, color, altered status, jurisdiction, the last season the cat was kept by the shelter, age × altered status and age squared × altered status (Table 3). The predicted curves generated for the interactions involving age and neuter status showed that if a cat was unaltered or of an unknown altered status, the likelihood of being euthanized was high regardless of age (Figure 1). However, among neutered animals, the odds of euthanasia were low for young animals and increased with age while eventually leveling off for older cats. Male cats were significantly more likely to be euthanized than females (Table 3). Also, male cats were significantly more likely to be euthanized than those of an unknown sex (Table 3). There was no significant difference in the odds of euthanasia between female cats and cats of unknown sex (Table 3). Cats that were surrendered to the shelter and admitted as euthanasia requests were significantly more likely to be euthanized than stray cats (Table 3). There was no significant difference in the odds of euthanasia between shelter offspring and stray cats (odds ratio [OR] 1.29, 95% CI 0.69–2.40; P = 0.430). Cats that came from out of the area or small town/rural jurisdictions were significantly more likely to be euthanized than cats that were admitted from local urban (Kitchener-Waterloo) areas (Table 3). Cats that came from out of the area or unknown areas were significantly more likely to be euthanized than those that came from a small town or rural areas (OR 3.13, 95% CI 1.39–7.05; P = 0.006). Black cats were significantly more likely to be euthanized than cats of any other color (Table 3). Cats that were last in the shelter in the spring, summer and fall months were significantly more likely to be euthanized than those cats that were last in the shelter during the winter (Table 3). Cats that were last in the shelter in the spring were at significantly lower odds of being euthanized than those that were in the shelter in the summer (OR 0.73, 95% CI 0.55–0.98; P = 0.03). There was no significant difference in the odds of being euthanized for those cats that were last in the shelter in the summer (OR 1.18, 95% CI 0.91–1.54; P = 0.20) and spring (OR 0.87, 95% CI 0.66–1.13; P = 0.30) compared with those that were last in the shelter in the fall months. Cats that spent between 5 and 20 days in the shelter were significantly more likely to be euthanized than those cats that spent <5 days in the shelter (Table 3) and cats that spent >20 days (OR 6.12, 95% CI 4.64–8.06; P <0.001). Cats that spent >20 days in the shelter were at significantly lower odds of being euthanized than those that spent <5 days in the shelter (Table 3).
Results of multivariable logistic regression analysis (n = 3435) concerning demographic characteristics and other factors that were associated with euthanasia among cats from a shelter in Kitchener-Waterloo, Ontario (2011)
OR = odds ratio; CI = confidence interval
Predictive curves, based on the multivariable logistic regression model on risk factors for cat euthanasia, illustrating the interaction between neuter status and age
The Hosmer–Lemeshow test was not significant (χ2 = 11.11; df = 9; P = 0.20) indicating the model fit the data. A small number of observations appeared to have a large influence on the model or fit the model poorly. However, their removal did not change the interpretation of the model and no recording errors justified their removal.
Discussion
This study highlights the characteristics that have contributed to the increasing odds of euthanasia in shelter cats. Study has been limited and is becoming dated, 1 as time passes and the approaches to sheltering evolve. The high risk of euthanasia noted here (53%) is consistent with another study in North America, 4 and highlights the importance of this topic. As is evidenced by this study, euthanasia is common and certain cat characteristics, and regional, seasonal and demographic variables impact the probability of being euthanized in this shelter.
Most cats that entered the shelter were strays, which was not surprising. Most of the animals came from urban areas, likely owing to the larger human population in these areas, therefore increasing the number of cats owned by the people that reside there and the greater likelihood someone would identify and report a stray cat. Further, community cats, feral cats that may be informally cared for by various members of the community, are more common in urban settings and provide another pool of cats to be captured for entry into the shelter system. More non-purebred cats came into the shelter than purebreds, something that may simply reflect the breed distribution of cats in the region, the rarity of feral or community purebred cats or be associated with increased care of purebred cats. Also, this may be related to the fact that purebred cats are not as common in the pet population.
Length of stay in the shelter is a crucial variable to understand. Specifically, the length of time in days, the frequency at which they return to the shelter and the season that they are last in the shelter are important variables to examine. The largest portion of cats spent <5 days in the shelter, followed by 5–20 days and then >20 days. Those that spent between 5 and 20 days in the shelter were more likely to be euthanized than those that spent <5 days in the shelter. This could reflect that those animals that spent <5 days in the shelter may have been more likely to be picked up by their owners during this window of time or that these animals are more desirable and therefore adoptable, so they leave the shelter more quickly. This could also indicate that owing to a hold period of 5 days once cats are admitted, the time period of 5–20 days in the shelter would increase the likelihood of euthanasia, if these cats were not considered adoptable. In contrast, cats that spent >20 days in the shelter were less likely to be euthanized than those that spent <5 days. This is interesting and perhaps somewhat counter-intuitive, as those cats could perhaps be considered less adoptable since they had not been adopted during the first 20 day period. This may have resulted from these cats being placed in foster care and still remaining listed as cats in the shelter system. It could also reflect the tendency for some shelters to keep cats that have remained for this long, as they may already have been deemed adoptable and may need treatment for an illness or they may require additional training and socialization in order to be adopted successfully. This additional training could include placing these cats in foster care to allow for this further socialization and training to occur. There are also programs in place at the shelter to allow for this training to occur.
Most cats were only admitted to the shelter once during the study period. Some animals may be re-admitted if they are not contained by their owners (and are re-captured as strays) or if they are surrendered because of problems in the post-adoption period. Re-admission had no significant effect on euthanasia in our multivariable model. It is also possible that this population consisted of cats both at higher risk (eg, unsuccessful adoption surrenders) and lower risk (eg, indoor/outdoor cats that get recaptured as strays but are cared for pets that are identifiable and retrieved by their owners) for euthanasia. Re-admission had a significant sparing effect in the univariable model, but this effect may have been confounded by age and source, which was controlled for in our multivariable model. The source of an animal entering a shelter has not been examined previously in the literature; however, this may be an important element to explore in future studies. This could be undertaken by not only citing the sources of the animals entering the shelter, but specifying their reason for admittance in detail, if this information is known.
The distribution of the number of cats that were last kept in the shelter over the four seasons was relatively equal; however, those that left the shelter in the spring, summer and fall were more likely to be euthanized than those cats that were last in the shelter during the winter season. Caution should be used in interpreting these seasonal effects, because this study was only 1 year in length and there was no replication of seasons. However, it has been established that the peak of feline pregnancy is during the spring and late summer. 7 The consequence of seasonal estrus and the rise in the kitten population during the spring to fall months, fits with the finding in this study that cats that left in the spring, summer and fall were more likely to be euthanized than those that were last in the shelter in the winter.
Male cats were more likely to be euthanized than females, suggesting a preference towards adopting females. This may be due to the belief that male cats tend to be more aggressive than females; 8 however, some studies have found an increase in male adoptions when looking at sex neuter interaction terms.1,6 Aggressiveness concerns may relate mainly to intact males, and there may be a perception that neutered males especially are less aggressive than females. 9 The statistical interaction term between sex and neuter status was tested in our study, and it was not significant, so this perception may not be valid. All cats that enter the shelter are neutered prior to adoption if they are deemed adoptable. The altered status variable was at the time of admission not discharge. Future studies should further explore the relationship between sex and adoptability in other jurisdictions, with more consideration of the impact of altered status, particularly as this is a modifiable variable (ie, male cats can, and typically are, neutered prior to adoption).
Cats surrendered to the shelter and those that were euthanasia requests were more likely to be euthanized than those that came in as strays. The euthanasia requests have an increased likelihood of euthanasia than strays, as these animals specifically came in to be euthanized, as specified by their owners. This is consistent with another study that found that stray cats were more likely to be adopted. 1 Those that came in as surrenders may have not been wanted for a variety of reasons such as problematic behaviors, illness or old age, financial concerns of the owners or allergies, so this may be why they would be more likely to be euthanized.
Cats that came from out of the area from unknown regions and small towns were more likely to be euthanized than those from urban areas. This could be due to their poorer health status or possible feral-like behavior making these animals poorly socialized and possibly not even put up for adoption in the first place. This could be due to a lack of proper veterinary services in these smaller towns and areas, as well as the more free-range type lifestyle that these cats may be accustomed to on farms.
Black cats were more likely to be euthanized than cats of other colors and other studies have also noted this finding as well, but explanations have varied. 1 To potential adopters, the appearance of a black cat may make it hard to distinguish certain facial features that would be evident with a lighter color, making them less desirable. One study looked at how potential adopters perceive personality based on coat color, and one significant finding was that orange cats were considered friendlier than cats of other colors. 10 This has been noted by various shelter organizations that have examined adoption patterns for black dogs and cats, such as the American Society for the Prevention of Cruelty to Animals. 11 This is especially true with the growing trend of using the internet to market animals available for adoption, as black cats may not photograph well. The Royal Society for the Prevention of Cruelty to Animals (RSPCA) in the UK has examined this further. Based on an article published in July of 2014, 70% of the 1000 abandoned cats that were in the care of the RSPCA were black in color. 12 The hypothesis was that these cats are not as popular and are not adopted as frequently because their black coloration makes them more difficult to photograph and it is also more difficult to recognize them as individual cats, owing to their lack of visible markings.
The effect of age on the odds of being euthanized significantly varied with the neuter status of the cats in this shelter. For animals that were unaltered or of unknown neuter status, the odds of being euthanized appeared to be high across age categories, while for neutered animals the odds of euthanasia increased with age with the effect of age leveling off for senior cats. This is consistent with the findings of another study that demonstrated that the likelihood of adoption decreased gradually with increasing age.1,6 These interactions may reflect that neutered animals are less of a hassle, avoiding the cost and inconvenience of the surgical procedure, eliminating the possibility of spraying behavior or that they may be less expensive to adopt. The change in the odds of euthanasia for neutered animals with age may simply reflect the preference of people to adopt younger animals. 1
Conclusions
There are many factors that influence whether or not a cat will be euthanized in the shelter. It is apparent that many people who adopt animals from shelters have discernible preferences for animals of certain color, age, sex and breed, 1 factors that would correspondingly influence the risk of euthanasia in the shelter. These could be changed with proper education and other interventions that could be carried out at the shelter level to optimize the adoptability of cats that are identified as being more difficult to adopt. Possible interventions that could be undertaken include using the information collected from this study to focus the shelter’s efforts and resources on those animals that are considered adoptable but harder to home. For instance, increased education or marketing on the benefits of adopting older cats and black cats might help reduce rates of euthanasia in these animals. Other interventions such as the use of foster programs, which keep the animals out of the shelter and allow for proper socialization and training in a household environment, which may be less stressful and more comforting to these cats, would also be helpful. This relates to the finding in this study of animals being euthanized owing to their length of stay in the shelter, so advocating the use of foster homes would be beneficial to reduce the length of stay in the shelter. This is especially important during the 5–20 day window, when the likelihood of euthanasia is increased. This could be improved with better record keeping by shelter staff with regards to fostering, to understand why animals that remain for >20 days are less likely to be euthanized.
This study only examined data from one shelter, so it is hard to discern whether or not these results can be generalized to other shelters in Ontario, or in Canada and the USA. Given the location of this shelter in a smaller city, this may also affect the likelihood of extrapolating these results to larger facilities in major cities across the country. In addition, a limitation of this study was that there was no specific indication as to why a cat was euthanized. This would be useful information to obtain from shelters for future studies. Also, information on stray returns to owners would be helpful.
While these data provide important information from an epidemiological context, evaluation of factors that impact adoption and euthanasia and development of effective interventions requires a broad, multidisciplinary approach. Assessment of areas such as social and psychological aspects that drive behaviors and decision-making of potential adopters, and information technology and social media approaches that can better link prospective adopters and appropriate pets, are important yet lacking. While studies of knowledge, attitudes and practices of shelters and shelter personnel are important, they cannot be the sole approach and studies more akin to ‘consumer-level research’ are needed to better understand and influence prospective adopters.
Footnotes
Acknowledgements
We would like to acknowledge the participation and assistance of the staff of the Kitchener-Waterloo Humane Society in collecting and providing access to shelter data. We would also like to acknowledge Dr Cheryl Yuill, for her help in designing this study during her tenure as the Veterinary Director of the Kitchener-Waterloo Humane Society.
Conflict of interest
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
Rachael Mozes was supported by an Ontario Veterinary College Graduate Scholarship, while this study was supported by Dr Weese’s Canada Research Chair in zoonotic diseases.