Geographic transitions of domestic cats in urban areas through animal adoption centers and the implications for population dynamics

doi:10.21203/rs.3.rs-1766586/v1

Download PDF

Research Article

Geographic transitions of domestic cats in urban areas through animal adoption centers and the implications for population dynamics

https://doi.org/10.21203/rs.3.rs-1766586/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this older preprint version

Read the latest preprint version →

Animal shelters address animal welfare in communities through the intake and outcome of companion animals, but these efforts affect population dynamics of companion animals based on the distance animals are moved and the factors that underlie intake and outcome. Using data from an animal shelter in Washington, D.C., we analyzed cat intakes and outcomes based on geographic and socioeconomic factors. Most intakes were stray cats (59%) and cats relinquished by owners (38%) whereas most outcomes were adoptions (84%). The highest number of intakes were in high development, low-income neighborhoods whereas the lowest number of intakes were in low development, high-income neighborhoods. The highest number of outcomes were to high-income neighborhoods and there was a trend towards more outcomes in neighborhoods further from the shelter. Cats returned to the shelter were more likely to originate nearer the shelter compared to relinquishment. Stray intakes were less common, and return to shelter were more common, in high-income, high development areas. Seized cats originated from low-income neighborhoods. Relative to adoptions, the proportion of returned to owner outcomes was higher in low-income neighborhoods that were closer to the shelter as well as high-income neighborhoods that were distant the shelter. Our results highlight the factors underlying cat intakes and outcomes in shelters that ultimately determine where, when, and how animals are moved across urban areas; these factors must be considered when developing cat population management plans to reach animal welfare and societal goals.

animal shelter

animal welfare

companion animal

adoption

relinquishment

stray cat

Animal shelters play a pivotal role in addressing animal welfare in communities in a multitude of ways, including providing care to animals that need assistance (i.e., intake) and adopting those animals to new homes (i.e., outcome). Inherent to the intake and outcome of animals is both the movement of animals across some geographic service area and the actions taken by shelters to mitigate pet overpopulation, such as spay-neuter programs (White et al. 2010). The movement and mitigation procedures undertaken in shelters are akin to natural movement of wild populations (e.g. dispersal) and management actions taken in some wildlife populations that ultimately influence population dynamics (Diffendorfer 1998). Therefore, to model population dynamics of companion animals, we must understand their movement through the shelter system (Flockhart and Coe 2018).

Quantifying intakes and outcomes are standard practice in shelter operations (Morris and Gies 2014, Janke et al. 2018, Hawes et al. 2019), however, most research has focused on how intakes and outcomes result from the traits of individual animals (Weiss et al. 2012), such as coat color (Svoboda and Hoffman 2015, Jones and Hart 2020, Carini et al. 2020) and age (Brown and Morgan 2015). Understanding how shelters contribute to population dynamics requires investigating intake and outcome counts at two geographic scales. The first, larger scale is that of the companion animals moving in and out of a shelter’s service area over a period of time; do more animals leave (via outcomes) the service area than arrive (via intakes), or, conversely, do more animals arrive in the service area than leave? These movements are important for understanding regional population dynamics and the magnitude to which the service area has a net increase or decrease of animals over time.

The second scale is smaller and occurs within the service area itself, with the aim of identifying neighborhoods where the number of intakes is greater than the number of outcomes (here termed ‘source’ areas) and those where the number of outcomes is greater than the number of intakes (here termed ‘sink’ areas). This scale is directly applicable to individual shelters and their operations. Relating intake and outcome counts to underlaying geographic and socioeconomic conditions can identify relationships between intakes and outcomes and how they vary across geographic and socioeconomic gradients. Understanding the expected counts of intakes and outcomes are important for identifying limitations of the public to access shelter services as well as how shelters may allocate resources to reach their customers. Additionally, since decisions can be made at animal shelters with respect to where, when, and how animals are moved across urban areas, they serve an important role in developing optimal management plans to reach animal welfare and societal goals.

Geographic and socioeconomic factors commonly underlie different reasons for intake and outcome (Ly et al. 2021a,b). The geographic and socioeconomic factors can help explain differences in intake type and outcome type across neighborhoods in a shelter’s service, which is useful for allocating shelter resources effectively. For example, if stray animals brought to the shelter are the predominant intake type from one neighborhood, then the services needed in that neighborhood are likely different than neighborhoods where the predominant intake type is owner relinquishment. In addition to shelter practices, relating underlying geographic and socioeconomic conditions to intake and outcome types can identify differences in cat owner behavior across geographic and socioeconomic gradients. For example, if intakes shift from owner relinquishment to stray pickup with increasing distance to the shelter, it may imply that cat owners abandon their animal instead of relinquishing their animal had they lived closer to the shelter. If so, then a shelter may investigate how to mitigate constraints of distance and improve accessibility of the shelter for some residents.

In this study we apply principles of wildlife ecology to quantify the movement of cats across Washington, D.C. by modeling domestic cat intake and outcome records at an animal shelter. We first determine if the movement of cats out of Washington, D.C. is higher or lower than the number of cats coming into the District. Next, we use intake and outcome counts of cats that move within the District to identify source and sink areas and use simulations to identify which locations are consistently source and sink areas over time. Finally, we develop models to explain intake (e.g. relinquishment) and outcome (e.g. adoption) types and visualize these geographically to identify which areas have high and low levels of different intake/outcome types. These visualizations serve to target appropriate resources in space and time to maximize cat welfare through service provisions to cats and their owners.

Shelter data

The Humane Rescue Alliance (HRA) is the only animal shelter in Washington, D.C. and therefore represents a complete dataset to estimate the shelter-facilitated movement of cats through Washington, D.C. homes and neighbourhoods. The Humane Rescue Alliance has been operating since 1870 and serves about 3,500 cats annually. Data standardization was enhanced starting in 2016 with the use of electronic data collection records for all cats entering the shelter. We used records that were collected between May 2016 and April 2020 that were composed of a complete intake and outcome record for each cat.

Upon entry to the shelter, cats were screened by shelter staff and a data record was initiated and completed with the best available information. The record for each cat includes an identification number, intake time and date, source location, and intake type. Intake types were classified as stray, owner/guardian relinquishment (hereafter “relinquishment”), seized/custody (hereafter “seized”), returned by owners/guardians within 30 days (hereafter “returned to shelter”), service in, or transfer in. We excluded service in records since this intake type was specific to cats that were born in the shelter or were submitted for specific trap-neuter-return programs, such as targeted vaccine or sterilization clinics. Transfer in records were also excluded as this intake type was used for adoptable cats that were being transferred from other organizations in different geographic regions and thus were coming from outside the Washington, D.C. population. For cats that were relinquished or returned to shelter, the intake location was recorded as the owner/guardian’s home address. Cats that were classified as seized intake were cats who were confiscated by animal service personnel, police, or fire services and the intake location was recorded as the address where the cat originated. For cats submitted as strays, the intake location was recorded as the location the cat was found and so was recorded as the nearest house, intersection, or landscape feature.

When cats left the shelter, the outcome type was recorded as adoption, euthanasia, dead on arrival, service out, return to owner/guardian (hereafter “returned to owner”), or died. We only considered live outcomes in our analysis and therefore excluded records where the outcome was euthanasia, dead on arrival, or died. For cats that were adopted or returned to owner the exact outcome address of cats was known, while cats that were serviced out (e.g. trap-neuter-return programs) the location where the cat was returned was recorded as the nearest house, intersection, or landscape feature and thus was always the same as the intake location.

We used a geocoding tool in ArcMap 10.8 to assign the latitude and longitude of the intake and outcome locations (ESRI, 380 New York Street, Redlands, CA 92373). An address locator was created around D.C. and the surrounding area to limit the search area for the tool to match addresses within. The locator was specifically helpful for matching addresses that were lacking zip code, city, or state, which was not unusual for stray intakes and service out outcomes. Minimum match score was set to a 95% confidence. All addresses that matched below the threshold, were tied between two possible points, or were unable to match were manually reviewed to ensure accuracy. Upon export, the geocoding tool standardized matched addresses and provided the coordinates associated with each address.

Spatial data

We obtained spatial data of distance to shelter locations, land use, household income, and housing locations to extract values for modeling cat intake/outcome. We overlaid a 400m-by-400m grid over Washington, D.C. over several covariate spatial data (polygons: household income; points: shelter locations, address locations; raster: land use) and summarized each grid square for land use, sum of residential units, and median household income. For each cat record we then extracted the covariate values for the intake and outcome grid square (Figure S1).

Distance to shelter for each grid square was calculated as the minimum distance to the nearest Humane Rescue Alliance shelter location, either 1201 New York Avenue NE (latitude: 38.91ºN; longitude: -76.99ºW) or 71 Oglethorpe Street NW (latitude: 38.96ºN; longitude: -77.01ºW).

We determined the location and number of households (residences) in two steps. First, we obtained a shapefile of the address points of residences in Washington D.C. from the Open D.C. Data portal (https://opendata.dc.gov/datasets/DCGIS::address-points/about). These address points indicate a residence type but could range from a single home to a high-rise apartment building. To obtain a total number of households, we spatially joined the address points to a table that indicated the number of residential units at each point from the Open D.C. data portal (https://opendata.dc.gov/datasets/DCGIS::address-residential-units/about). The result was a count of residences at each location which we then overlayed with our sample grid and summed the total number of residences in each grid square (Figure S1).

Median household income at the grid level was calculated in two steps. In the first step, we acquired median household income from the census tract-level for 2020 (Environmental Systems Research Institute 2020) which uses the same methodology as the American Community Survey (ACS) 5-Year Estimates from the U.S. Census Bureau (U.S. Census Bureau 2020). Secondly, we assigned the median household income to each residential unit from the joined shapefile mentioned above and then overlayed the 400m-by-400m grid to calculate the mean income for each grid square for each year (Figure S1).

For land use data, we used the National Landcover database (Dewitz 2019) and reclassified the land uses to 3 levels. We categorized areas as “high development” if they were classified as either high development, moderate development, or mixed/high development. Additionally, we categorized “low development” areas if they were areas classified as any of low development/open development, low development, mixed low development, or open development. Lastly, our “forested” land use category included areas classified as forest, forest/mixed low, forest/open development, or water (Figure S1).

The response variable in our first set of models was the count of intake and outcomes by grid square. However, we observed the number of intakes and outcomes to vary seasonally, likely based on the seasonal breeding of cats in Washington, D.C. (Fig. 1). Therefore, we summarized the intake and outcome counts by season for each grid square and included season as a predictor variable in our linear models (see below). We defined spring as March through May, summer as June through August, autumn as September through November, and winter as December through February.

Analysis and Model selection

We first summarized intake and outcome locations from within and outside Washington, D.C. to identify if the District operates as a source (more intakes than outcomes) or sink (more outcomes than intakes). Next, we excluded records where the intake location, outcome location, or both was outside Washington, D.C. and subsequently summarized intake and outcome frequencies. We also plotted the number of intakes and outcomes across the District which showed the spatial distribution of the records but ignored the underlying spatial covariates to explain the number of intakes and outcomes.

We used two sets of regression analyses to understand intakes and outcomes to the animal shelter. First, we used generalized linear models with a negative binomial distribution to model the counts of cat intakes and outcomes by several predictor variables: season, household income, distance to shelter, and land use; we conducted separate analyses for intakes and outcomes. The patterns of both intake and outcome counts were very similar to housing density in the District (Fig. 2; Figure S1). Therefore, since the number of intakes and outcomes occurs at the household level, and housing density varies widely across the District, all regression models included the number of residences as an offset. As such, the model parameters represent the count of intakes/outcomes per household. Our global model was a four-way interaction among all predictor variables. We then used AICc to rank a candidate list of models of all reduced models to explain the counts of intakes and outcomes (Table S1; Table S2). We considered competing models to be those within 10 AICc units (delta AICc) of the top model. We used the parameter estimates from the top model to plot the predicted relationships among the predictor variables to understand how intake and outcome per household varied among the variables of interest.

The second set of regressions were multinomial regressions to explain the intake and outcome types. We used the top model to explain intake and outcome counts as the global model and used the dredge function in the MuMIn package (Barton 2020) to build a candidate list of models. We again evaluated the support of each model relative to the best supported model using delta AICc values. Due to model uncertainty (Table S3; Table S5), we used a model-averaged technique to estimate the mean and confidence intervals of the parameter estimates of all variables. We then plotted the predicted relationships among the predictor variables to understand how intake and outcome types varied among the variables of interest.

Finally, we modeled the difference in intakes and outcomes per household in each grid square and the proportion of intake and outcome types in each square. The difference between intake and outcomes in each square identifies which squares consistently had higher counts of cat intakes to the shelter (which we termed source areas) or higher counts of cat outcomes from the shelter (which we termed sink areas) among seasons. We resampled, with replacement, 66% of the original datasets and conducted 4000 iterations of each analysis and used the predict function to estimate the number of intakes and outcomes, as well as the proportion of each intake and outcome type per square. In each iteration we subtracted the outcome count from the intake count of each square, where a positive number represented a higher outcome than intake (a sink area) and a negative number indicated a higher intake than outcome (a source area). If the 95% confidence interval of the distribution did not overlap zero, we considered this to indicate statistical significance that the grid square was either a source or sink square. As we accounted for season in the modeling, we plotted the results for each season to identify which squares varied seasonally between source and sinks. Additionally, we plotted the source/sink status over all seasons to identify which squares were consistent in their status over the entire year. Similarly, we plotted the proportion of each intake and outcome type for each square to understand how intake and outcome types varied across the District.

Over the study, 14,499 cats with complete records entered the Humane Rescue Alliance. The movement of cats out of the District was greater than the movement of cats into the District. A total of 3,015 more cats left Washington, D.C. compared to entered (total number moved out of D.C.: 4724; total number moved into D.C.: 1709), indicating Washington, D.C. operates as a source of cats (Table 1). In total, 5,145 cats had both an intake and outcome within Washington, D.C. (Table 1). The vast majority of shelter intakes were stray cats and cats relinquished by owners (Table 2). Intakes occurred from across the District but were most prominent in the central, south, and east portions of the District (Fig. 2). Outcomes were more geographically dispersed than intakes and were highest in the central portion of the District (Fig. 2).

Table 1

Number of cats (proportion in parentheses) of all cats whose geographic location on intake and outcome at the Humane Rescue Alliance was within or outside Washington, D.C. between May 2016 and April 2020.
	Intake
Outcome	Washington, D.C.	Outside Washington, D.C.
Washington, D.C.	5145 (35%)	1709 (12%)
Outside Washington, D.C.	4724 (33%)	2921 (20%)

Table 2

Number of cats (proportion in parentheses) of intake sources and outcome results of the cats coming to the Humane Rescue Alliance between May 2016 and April 2020. The total number of cats was 5145 and represents the sample sized used in the regression analyses.
	Intake
Outcome	Relinquishment	Return to shelter	Seized	Stray
Adoption	1768 (34.4%)	76 (1.5%)	59 (1.1%)	2441 (47.4%)
Return to owner	79 (1.5%)	4 (0.08%)	28 (0.5%)	159 (3.1%)
Service out	90 (1.7%)	4 (0.08%)	2 (0.04%)	435 (8.5%)

Intakes per household were best described by an effect of season and an interaction among income, distance to shelter, and land use (Table 3). No other model held support to explain intakes per household (Table S1). Intakes per household were highest in summer and autumn and lowest in winter (Table 4; Fig. 3). The highest intake per household occurred in high development areas and was lowest in low development areas (Table 4). There was a consistently negative relationship between intake per household and income, meaning the highest number of intakes were in low-income neighborhoods (Table 4; Fig. 3). The distance to shelter was negatively related to intake per household but did not, on its own, effect intake (Table 4). Furthermore, in high development areas the modelled lowest intake per household were in low-income neighborhoods and a long distance from the shelter whereas the highest intake per household were from high income neighborhoods a long distance from the shelter (Fig. 3).

Table 3. Comparison of negative binomial models (ΔAICc < 30) to explain intake counts in Washington. D.C. All models have offset of the number of households.

Model	K	AICc	ΔAICc	*L_i*	*W_i*	LL	∑ W_i
Season + Income x Distance x Land use	16	7052.1	0	1	1	-3509.86	1
Season + Income x Distance + Land use	10	7064.0	11.96	0	0	-3521.94	1
Season + Income x Land use	10	7065.1	13.02	0	0	-3522.48	1
Season x Income x Distance + Land use	19	7073.7	21.63	0	0	-3517.61	1
Season + Income x Distance	8	7076.5	24.41	0	0	-3530.2	1
Income x Land use x Distance	13	7080.8	28.71	0	0	-3527.28	1

Table 4

Parameter estimates for the best supported model of intake count. Since the model had an offset of the number of households the response is the number of intakes per household basis. The intercept represented spring and forested land use.
Parameter	Estimate	SE	Z	P
Intercept	-3.087	0.369	8.4	< 0.001
Season
Summer	0.223	0.056	4.0	< 0.001
Autumn	0.172	0.056	3.1	0.002
Winter	-0.073	0.060	1.2	0.23
Income	-0.000013	0.000006294	2.0	0.04
Distance	-0.028	0.058	0.5	0.62
Land use
High development	0.420	0.414	1.0	0.31
Low development	-0.526	0.398	1.3	0.19
Income x Distance	0.0000005214	0.0000009796	0.5	0.59
Income x Land use
Income x High development	-0.0000112	0.0000067395	1.7	0.10
Income x Low development	0.00000260	0.0000066325	0.4	0.69
Distance x Land use
Distance x High development	-0.175	0.083	2.1	0.03
Distance x Low development	0.021	0.063	0.3	0.74
Income x Distance x Land use
Income x Distance x High development	0.000002580	0.000001217	2.1	0.03
Income x Distance x Low development	0.00000002707	0.00000104544	0.03	0.98
Theta	3.71	0.27

Like intake, outcomes per household were also best described by an effect of season and an interaction among income, distance to shelter, and land use (Table 5). Again, no other model held support (Table S2). Outcomes per household were highest in summer, autumn, and winter and lowest in spring (Table 6). A higher number of outcomes occurred in neighborhoods with higher household income and there was a trend towards higher outcome counts in neighborhoods further from the shelter (Table 6). There was a negative relationship between outcomes per household and household income in high development areas with outcome counts increasing in low-income neighborhoods the further the neighborhood was from the shelter (Fig. 4). In contrast, the relationship between forest, low development land use areas, and outcome counts changed from negative to positive based on household income and distance to the shelter (Fig. 4). In low development land use areas, there was a negative relationship with outcome counts in neighborhoods nearest the shelter in contrast to a positive relationship with outcome counts in neighborhoods further from the shelter (Fig. 4). In forested land use areas, the pattern was reverse: there was a positive relationship with outcome counts in forested neighborhoods nearest the shelter and a negative relationship with outcome counts in neighborhoods further from the shelter (Fig. 4). Low development areas generally had lower outcomes per household compared to forest and high development land uses (Table 6; Fig. 4).

Table 5

Comparison of negative binomial models (ΔAICc < 50) to explain outcome counts in Washington, D.C. All models have offset of the number of households.
Model	K	AICc	ΔAICc	L_i	W_i	LL	∑ W_i
Season + Income x Distance x Land use	16	7568.2	0	1	1	-3767.9	1
Income x Land use x Distance	13	7614.8	46.57	0	0	-3794.3	1
Season x Income x Distance x Land use	49	7615.7	47.5	0	0	-3757.4	1
Season + Income x Land use	10	7618.3	50.08	0	0	-3799.1	1

Table 6

Parameter estimates for the best supported model of outcome count. Since the model had an offset of the number of households the response is the number of outcomes per household. The intercept represented autumn and forested land use.
Parameter	Estimate	SE	Z	P
Intercept	-4.75	0.324	-14.6	< 0.001
Season
Summer	0.37	0.055	6.84	< 0.001
Autumn	0.33	0.055	5.97	< 0.001
Winter	0.18	0.058	3.06	0.002
Income	0.00001092	0.000004157094	2.63	0.009
Distance	0.098	0.054	1.81	0.071
Land use
High development	0.38	0.366	1.05	0.294
Low development	0.22	0.355	0.62	0.534
Income x Distance	-0.00000212	0.000000680010	-3.12	0.002
Income x Land use
Income x High development	-0.00001528	0.0000045429	-3.36	0.001
Income x Low development	-0.000013803	0.00000453897	-3.04	0.002
Distance x Land use
Distance x High development	0.15	0.073	2.06	0.039
Distance x Low development	-0.12	0.059	-2.08	0.038
Income x Distance x Land use
Income x Distance x High development	0.000000915319	0.000000843453	1.09	0.278
Income x Distance x Low development	0.000002831050	0.000000741201	3.82	< 0.001
Theta	4.17	0.32

When considering all portions of the District across all seasons, the south and east portions of the District consistently operate as sources of cats whereas the central portion of the District and those in the northwest consistently operate as sinks (Fig. 5). The highest proportion of the District acted as a source in the spring, whereas the highest proportion of the District operated like a sink in winter (Figure S2). The strong seasonal effect resulted in some areas, especially in north-central Washington, D.C., shifting seasonally from sources or sinks to neutral within a year (Figure S2).

The best model to explain the intake type included distance to the shelter, household income, land use, and season as well as an interaction between distance and land use (Table S3). Several competing models held some support to explain intake type including some evidence of an interaction between distance and income on intake type (Table S3). Model averaged parameter estimates among the different intake type categories showed that stray intakes were significantly more common than relinquishments in summer and autumn (Table S4). Cats returned to shelter were less likely to originate from farther distances from the shelter, whereas seized cats were more likely to originate from further from the shelter compared to relinquishment (Table S4; Figure S3). Stray intakes were less common and return to shelter were more common in high-income, high development areas compared to other land uses (Fig. 6). Seized cats predominately originated from low-income neighborhoods and were more common in low development areas compared to high development areas (Fig. 6). The probability of any intake type varied geographically across the District (Fig. 8). Areas in the southeast and extreme west of the District were dominated by intakes of surrendered cats. Stray cats were more likely to come from areas nearest the shelter which correspond to the central and northeastern portions of the District. Intakes of returned cats were predominately from the high-income neighborhoods in northern Washington, D.C. In contrast, seized cats originated from low-income neighborhoods in southern Washington, D.C. (Fig. 8).

The best model to explain outcome type included interactions among all predictor variables (Table S5). Competing models did not hold support (Table S5). Adoption outcomes were the most common under all combinations of household income, season, land use, and distance to the shelter (Table S6). Relative to adoptions, the proportion of returned to owner outcomes was higher in low-income neighborhoods that were closer to the shelter as well as high-income neighborhoods that were distant the shelter (Fig. 7; Figure S4). Relative to adoption, the proportion of service out outcomes was more common in high development areas with high household income whereas service out outcomes were more common in low-income neighborhoods in forest and low development areas (Fig. 7). The probability of outcome type varied geographically among central, northern, and southern neighborhoods (Fig. 9). Adoption outcomes were the most common in central, highly developed neighborhoods and northern, forested, and low development neighborhoods (Fig. 9). In contrast, return to owner outcomes were the highest in southern and eastern neighborhoods (Fig. 9). Service out outcomes occurred in southern neighborhoods and areas surrounding the New York Avenue shelter location (Fig. 9).

Animal shelters play an important role for moving animals across short and long distances. We found that Washington, D.C. operates as a source of cats, with more cats departing than entering the District. More than three thousand left the city compared to entered, which suggests, under the assumption that the cat population is at equilibrium, that almost one thousand cats may need to be moved out of the District annually to maintain a stable cat population. Although we found four consecutive years of net export of cats from Washington, D.C., annual variation in shelter intake of companion animals is common for a variety of reasons (Janke et al. 2018, Rowan and Kartal 2018, Powell et al. 2021), so it is possible that a net import of cats may have been observed in some years outside of our study. However, over a relatively stable period, we documented a net export of cats. This is likely driven by HRA being open-admission for only D.C. residents, but also influenced by D.C.’s connectivity to Maryland and Virginia, HRA’s consistent availability of animals, and its open-adoption process for in-state and out-of-state adopters alike. The result makes out-of-state adoptions both accessible and easy, which promotes geographic transition of cats out of D.C. On the other hand, accessibility of HRA’s two shelter locations may limit access to services for D.C. residents in certain areas and income-levels who may rely heavily on public transportation (Serulle and Cirillo 2016).

Looking within the District, more than five thousand cats geographically transitioned over the study period. Contrasting patterns of more cat intakes coming from low development and low-income areas and more cat outcomes from areas with high development and income indicate a directional movement of cats across Washington, D.C. via the animal shelter. Geographically this was represented by cats moving from south and southeast portions of Washington, D.C. to central, northern, and northwestern portions of the city. However, many parts of the District fluctuated seasonally between acting as sources or sinks. This seasonal fluctuation is likely driven by the influx of cats into the shelter that we observed between June and November, which is almost certainly due to the summer/autumn breeding season for domestic cats (Kasten et al. 2017, Janke et al. 2018). Large numbers of intakes during the breeding season months may consequentially result in increased average length of stay and, as a result, some areas of the District appear to act like sinks during winter but not in other seasons. In this sense, shelters may act as sponges that absorb many cats during the summer and autumn and then adopt those cats out over the subsequent weeks and months back to the community. Grid squares in the District that fluctuate seasonally (i.e. switch from green to neutral among seasons in Figure S2) may, on average, indicate higher reproductive rates in the local free-roaming cats. If so, then our analysis reveals one way to a priori identify areas where free-roaming cat reproduction may be unabated, and thus where targeted management intervention may be most needed.

The pattern of intake and outcomes changed among household income, land use, and the distance to the shelter. In most instances, per household intake was negatively related to income except in high development neighborhoods distant to the shelter. Additionally, distance to the shelter differentially impacted per household intake from low and high-income areas. Distance to the nearest shelter was a limiting factor in per-household intakes in low-income areas while this was not the case for high-income areas. This provides support for the hypothesis that proximity to shelter services may limit the accessibility of these services for low-income residents. These effects may be more pronounced in residents that rely on transportation services where transporting pet animals is prohibited (e.g. ride-sharing) or where travel times to reach a destination can be prohibitively long (e.g. public transportation or taking extended time off from work). Further, the household income inflection points where distance was no longer a limiting factor on intakes changed between land uses and was lowest in low development areas and highest in high development areas. Given the predominance of owner relinquishment and the effects of distance on intake, one hypothesis may be that there is a marked benefit of enhanced efforts to address intakes in low-income neighborhoods distant to the shelter. If an alternative to relinquishing an unwanted cat to the shelter is to abandon the cat, then these efforts may concurrently serve to enhance services to residents and to reduce the number of cases of abandonment that influence the stray cat population (Flockhart et al. 2018). Furthermore, these efforts would be most beneficial in high development areas given the magnitude of intakes per household in these areas were two to three times more than intakes from forest and low development areas.

Per household outcomes were generally positively related to income in forest and low development areas but negatively related to income in high development areas. Under these conditions, high income neighborhoods, especially in forest and low development areas, are areas with a disproportionate number of cat outcomes; the vast majority of these outcomes are through adoptions. Compared to the results for intakes, distance to the shelter did not appear to be a limiting factor for cat outcomes. For example, the only evidence of distance limiting outcome rates was in low-income (< $50,000/year) neighborhoods in low development areas; however, this relationship seems relatively weak and almost all low-income neighborhoods are distant to the two animal shelter locations. Still, this effect may result from limitations such as access or financial constraints on adoption or providing care for adopted animals (Weiss and Gramann 2009).

There were strong socioeconomic and geographic determinants to intake type and outcome type that underlie the geographic patterns of intake and outcome. Cats relinquished to the animal shelter originated from across the District and there was little variation across different geographic and socioeconomic categories. These results are perhaps not surprising given that relinquished animals make up the majority of intakes and is considered by residents to be the primary service of animal shelters (Weng et al. 2010). Stray intakes were also common and occurred across the District and although there was a distinct geographic pattern of many stray cats coming from areas closer to the shelter, distance to the shelter was not a significant predictor for the probability of intakes being strays. Seized and returned cat intakes show distinct contrasting patterns: returns predominated from areas in the north and seized cats predominated from areas in the south. Adopted cats that are subsequently returned to the shelter within 30 days represent one possible outcome of a cat that has been adopted but can no longer be cared for. Alternative outcomes include relinquishing an adopted cats more than 30 days after adoption, gifting or selling the cat to someone else, or abandonment. Since data is lacking on all possible outcomes of adopted cats, it is possible that our data represent those that were returned while similar proportions of adopted cats from other areas could have been transfers to other owners or abandoned. Alternatively, there may be a consistent rate for no longer wanting an adopted cat across the city but relinquishing a cat takes longer than the 30-day threshold of what constitutes a return to shelter and instead these cats are recorded as a relinquishment; more detailed data that follows adopted cats for longer periods may provide insight (Hawes et al. 2020). Seized cats predominately occurred in southern areas of the District suggesting that allocation of certain resources or interventions may be necessary in these areas to minimize the situations where seizure is required. Identifying the drivers and geographic patterns of different intake types are necessary to further evaluate how shelters can best serve their communities because different resources are necessary to alleviate different reasons for intake.

Adoption was the most common outcome type in our study and occurred across the District. The highest rates of adoption occurred in high income areas in the northern and central portions of the District. The high per-household adoption rate in these areas is the major driver of these regions being sinks for cats in Washington, D.C. While it may be tempting to classify these areas as promising targets for future adoptions, these trends may instead be revealing systemic barriers to pet ownership in areas where the proportion of adoption outcomes is lower. The highest probability of return to owners of lost cats and service out was in southern portions of the District. While it may seem counter-intuitive for return to owner rates to be highest in the low-income areas of the south, these patterns may arise from the fact that few lost animals are returned to owners (Weiss et al. 2012) and these areas of Washington, D.C. have few outcomes. However, identifying areas where, on average, higher return to owner rates occur may also indicate areas that have higher rates of pet loss. Pet loss may occur through free, unrestrained outdoor access of pet cats and if so, these areas serve as candidates to both understand how cats may be lost in these areas, methods for finding pets, and information to cat owners about the risks to their pet from unrestrained outdoor access. Furthermore, the need for outdoor cats to have visible signs of ownership (a collar or tag) may negate the good intentions of residents assuming an outdoor cat is lost.

Our results highlight the important relationship between the role of shelters and the underlying demographics of residents and the cats they interact with in cities. The implications are that any cat management program must consider the directional movements of cats among neighborhoods through shelter operations and how the number of cats moved may negate impacts from other cat management actions. By identifying the locations that consistently operate as sources and sinks, shelters could prioritize resources to certain neighborhoods in specific seasons to reduce the number of intakes and maximize positive outcomes. In this sense, we are interested in improving the efficiency and return on investment of shelter services. For example, if cats are more likely to enter the shelter as lost from certain portions of the city, then it suggests efforts by the shelter to enhance returning cats to owners in these areas may provide a high return on investment by the shelter. If stray cats come into the shelter from specific areas, then resources might be invested to reduce the number of stray animals in these areas – assistance to keep cats confined or inside to reduce wandering, supplemental food to help keep pets in their homes despite monetary constraints, enhanced services to reduce reproduction of free-roaming cats, and recourses to reduce abandonment. These types of targeted programs in specific areas based on data collected at the shelter provide one way to enhance the return on investment of shelters providing services to their communities. At the same time, neighborhood characteristics that explain how cats move through the shelter system may work in contrast against neighborhood-level management actions that target free-roaming cats and therefore reduce chances of reaching mutually agreeable objectives. Given the large numbers of cats that move through shelter systems, their dynamics are important when considering all the components of the cat populations and their optimal management in cities to reaching our collective animal welfare objectives.

Acknowledgments

We wish to thank Lauren Lipsey for assistance in data collection. Major financial support was provided by: PetSmart Charities, American Society for the Prevention of Cruelty to Animals, The Humane Society of the United States, Humane Rescue Alliance, Winn Feline Foundation, Maddie’s Fund, Cat Depot, and B. Von Gontard.

Funding – Major financial support was provided by: PetSmart Charities, American Society for the Prevention of Cruelty to Animals, The Humane Society of the United States, Humane Rescue Alliance, Winn Feline Foundation, Maddie’s Fund, Cat Depot, and B. Von Gontard.

Conflicts of interest/Competing interests – The authors have no relevant financial or non-financial interests to disclose.

Availability of data and material – Data to reproduce the results of the paper will be provided upon paper acceptance

Code availability – Code to reproduce the results of the paper will be provided upon paper acceptance

Authors contributions – Tyler Flockhart and Julia Kilgour designed the research, all authors conducted the data collection, Tyler Flockhart and Julia Kilgour conducted the data analysis, wrote the main manuscript, and prepared all figures. All authors read and approved the final manuscript.

Ethics approval – Not applicable

Consent to participate – Not applicable

Consent for publication – Not applicable

Barton, K (2020). MuMIn: Multi-Model Inference. R package version 1.43.17. https://CRAN.R-project.org/package=MuMIn
Brown, W. P., & Morgan, K. T. (2015). Age, breed designation, coat color, and coat pattern influenced the length of stay of cats at a no-kill shelter. Journal of Applied Animal Welfare Science, 18(2), 169-180.
Carini, R. M., Sinski, J., & Weber, J. D. (2020). Coat Color and Cat Outcomes in an Urban US Shelter. Animals, 10(10), 1720.
Dewitz, J., 2019, National Land Cover Database (NLCD) 2016 Products: U.S. Geological Survey data release, https://doi.org/10.5066/P96HHBIE.
Diffendorfer, J. E. (1998). Testing models of source-sink dynamics and balanced dispersal. Oikos, 417-433.
Environmental Systems Research Institute. 2020. 2020 USA Median Household Income. https://demographics5.arcgis.com/arcgis/rest/services/USA_Demographics_and_Boundaries_2020/MapServer. Retrieved August 18. 2020.
Flockhart, D. T., & Coe, J. B. (2018). Multistate matrix population model to assess the contributions and impacts on population abundance of domestic cats in urban areas including owned cats, unowned cats, and cats in shelters. PloS one, 13(2), e0192139.
Hawes, S. M., Camacho, B. A., Tedeschi, P., & Morris, K. N. (2019). Temporal trends in intake and outcome data for animal shelter and rescue facilities in Colorado from 2000 through 2015. Journal of the American Veterinary Medical Association, 254(3), 363-372.
Hawes, S. M., Kerrigan, J. M., Hupe, T., & Morris, K. N. (2020). Factors informing the return of adopted dogs and cats to an animal shelter. Animals, 10(9), 1573.
Janke, N., O. Berke, E. Klement, D.T.T. Flockhart, J. Coe & S. Bateman (2018) Effect of Capacity for Care on cat admission trends at the Guelph Humane Society, 2011–2015, Journal of Applied Animal Welfare Science, 21:3, 283-294, DOI: 10.1080/10888705.2017.1414607
Jones, H. D., & Hart, C. L. (2020). Black cat bias: Prevalence and predictors. Psychological reports, 123(4), 1198-1206.
Karsten, C. L., Wagner, D. C., Kass, P. H., & Hurley, K. F. (2017). An observational study of the relationship between Capacity for Care as an animal shelter management model and cat health, adoption and death in three animal shelters. The Veterinary Journal, 227, 15-22.
Ly, L. H., Gordon, E., & Protopopova, A. (2021a). Inequitable Flow of Animals in and Out of Shelters: Comparison of Community-Level Vulnerability for Owner-Surrendered and Subsequently Adopted Animals. Frontiers in veterinary science, 8.
Ly, L. H., Gordon, E., & Protopopova, A. (2021b). Exploring the relationship between human social deprivation and animal surrender to shelters in British Columbia, Canada. Frontiers in veterinary science, 8, 213.
Morris, K. N., & Gies, D. L. (2014). Trends in intake and outcome data for animal shelters in a large US metropolitan area, 1989 to 2010. Journal of applied animal welfare science, 17(1), 59-72.
Powell, L.; Houlihan, C.; Stone, M.; Gitlin, I.; Ji, X.; Reinhard, C.L.; Watson, B. Animal Shelters’ Response to the COVID-19 Pandemic: A Pilot Survey of 14 Shelters in the Northeastern United States. Animals 2021, 11, 2669. https://doi.org/10.3390/ani11092669
Rowan, A.; Kartal, T. Dog Population & Dog Sheltering Trends in the United States of America. Animals 2018, 8, 68. https://doi.org/10.3390/ani8050068
Serulle, N.U., Cirillo, C. Transportation needs of low income population: a policy analysis for the Washington D.C. metropolitan region. Public Transp 8, 103–123 (2016). https://doi.org/10.1007/s12469-015-0119-2
Svoboda, H. J., & Hoffman, C. L. (2015). Investigating the role of coat colour, age, sex, and breed on outcomes for dogs at two animal shelters in the United States. Animal Welfare, 24(4), 497-506.
Weiss, E., & Gramann, S. (2009). A comparison of attachment levels of adopters of cats: Fee-based adoptions versus free adoptions. Journal of Applied Animal Welfare Science, 12(4), 360-370.
Weiss, E., Slater, M., & Lord, L. (2012). Frequency of lost dogs and cats in the United States and the methods used to locate them. Animals, 2(2), 301-315.
Weiss, E., Miller, K., Mohan-Gibbons, H., & Vela, C. (2012). Why did you choose this pet?: Adopters and pet selection preferences in five animal shelters in the United States. Animals, 2(2), 144-159.
White, S. C., Jefferson, E., & Levy, J. K. (2010). Impact of publicly sponsored neutering programs on animal population dynamics at animal shelters: The New Hampshire and Austin experiences. Journal of Applied Animal Welfare Science, 13(3), 191-212.
U.S. Census Bureau. (2020). 2016-2019 American Community Survey 5-year Public Use data. https://www.census.gov/programs-surveys/acs. Retrieved from August 18. 2020.
Weng, H. Y., Kass, P. H., Hart, L. A., & Chomel, B. B. (2006). Animal protection measures in Taiwan: Taiwanese attitudes toward the animal protection law and animal shelters. Journal of Applied Animal Welfare Science, 9(4), 315-326.

No competing interests reported.

floatimageS1.png
Figure S1. Spatial covariates of Washington, D.C. used to model the intake counts, outcome counts, intake probability, and outcome probability. Values are summarized at 400m by 400m grid.
floatimageS2.png
Figure S2. Balance of per household shelter intakes and outcomes in spring (top left), summer (top right), autumn (bottom left), and winter (bottom right). Red areas represent a significantly higher number of outcomes per household (sink areas) and green areas represent a significantly higher number of intakes per household (source areas) across all seasons. White areas are those that have an equal number of intakes and outcomes per household.
floatimageS3.png
Figure S3. Predicted intake type by season, land use, household income, and distance to shelter. The distance to shelter is located at the top left of each panel.
floatimageS4.png
Figure S4. Predicted outcome type by season, land use, household income, and distance to shelter. The distance to shelter is located at the top left of each panel.
SupplementaryTable1to6.docx

Download PDF

Version 1

posted

You are reading this older preprint version

Read the latest preprint version →

Geographic transitions of domestic cats in urban areas through animal adoption centers and the implications for population dynamics

Status:

Version 1

Abstract

Figures

Introduction

Methods

Shelter data

Spatial data

Analysis and Model selection

Results

Discussion

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1