Identification and reduction of racial and ethnic disparities in health has become an important strategy for decreasing the staggering toll of avoidable human morbidity and premature deaths.1–3 For example, in 2000, approximately half of all deaths in the United States could be attributed to a few behaviors and exposures that were likely preventable.4 These conditions become health disparities when their prevalence becomes defined by gender, race or ethnicity, or social class.1,3,5 As a result, people affected by these disparities are less healthy, have less access to medical care, and generally have worse outcomes with medical treatment when it is provided. Various measures of health disparities have been recommended.6,7 Of all these measures, premature death may be the most global indicator. In particular, investigators with the Public Health Disparities Geocoding Project at the Harvard School of Public Health8 have argued that the rate of premature deaths is a “highly informative, easily calculated, and easily understood single measure that captures social disparities in community health.”5
The concept of community-level disparities in health and welfare among companion animals has not been explored to the author's knowledge. However, a basic premise of the AVMA's One Health Initiative is that the welfare of people and other animals is closely linked.9 By extension, the spectrum of social ills and problems that result in health disparities and premature deaths in the human population would also likely create conditions that result in intakes into shelters of stray, unwanted, injured, or ill pets. Shelter-related euthanasia is believed to be the single leading cause of preventable death for dogs and cats in the United States10,11; thus, premature death should be an equally relevant measure of health disparities in animals other than humans. This outcome is known to vary considerably across communities and can be modified with concerted intervention. Communities that have been deemed most successful in applying effective interventions currently achieve per capita annual shelter-associated mortality rates of approximately 2 to 7 dog and cat deaths/1,000 persons.12 In those same communities before interventions, the reported per capita mortality rates were in the range of 20 to 30 dog and cat deaths/1,000 persons.13 In communities without the infrastructure to support development of progressive programs, the numbers are likely far higher. It is not yet known to what extent there might be geographic disparities in mortality rate among companion animals within the same community, in addition to those already determined across communities.
One of the problems that have plagued investigations of health disparities in human public health research is the difficulty of linking individual-level demographic and socioeconomic data to available outcomes databases. This has been approached through the use of area-based analyses, in which demographic data at the neighborhood level or census tract are linked to population-based data on health outcomes and mortality rates.14–20 This is justifiable because, as Messer et al6 have reported, multiple markers of economic disadvantage typically cluster at the neighborhood level. Indeed, it has been shown that ABSMs can provide estimates of health inequalities that are comparable to those based on individual-level data.21 Thus, neighborhood-level factors must be understood and addressed to reduce disparities in health status among individuals and groups.2 It makes sense, therefore, that efforts to address health disparities among underserved companion animals must be considered at the neighborhood level as well. Under the right circumstances (eg, reasonably complete reporting from all of the shelters serving an area) and with the appropriate tools (eg, geospatial analysis software to accurately and quantitatively connect shelter animal records to a specific human population), animal shelter data provide the opportunity to examine the possibility of health disparity among companion animals and its relationship to human health. The purpose of the study reported here was to determine whether disparities in health and welfare among cats are present within neighborhoods and across census tracts in Boston, and to compare results with area-level human data.
Materials and Methods
Study location and population—Routinely collected intake and outcome data from 2 large animal sheltering organizations in Boston—the ARL-B and the MSPCA—were used to examine health disparities in the cat population (by use of ABSMs) in the same neighborhoods and census tracts as previously reported for Boston's human residents.5 Three shelters from these 2 organizations contributed detailed information: the MSPCA and ARL-B headquarters shelters, which are both within the city limits, and the Dedham branch of the ARL-B, which is approximately 1 mile outside the southwest border of Boston. Each organization used the same computerized data management system for all shelter operations,a and client addresses were routinely captured at intake.
Cat data were chosen for the study, rather than dog data, for several reasons. First, the ARL-B and MSPCA handle the majority (> 90%) of cat intakes from the city and are therefore considered to provide representative data. Second, there would not have been a sufficient number of dog records, even over a 5-year period, for analysis at the level of the census tract. Finally, stray dogs are primarily handled by City of Boston Animal Control, and those data were not available at the level of detail required for this analysis.
Cat data—Records for all cat intakes at the ARL-B and MSPCA from Boston neighborhoods for a 5-year period (2004 through 2008) were downloaded into spreadsheets. Variables of interest were as follows: intake type (eg, owner surrender, stray, or confiscation), intake reason, intake date, cat's age and disposition (eg, adopted, sent to foster home, returned to owner, euthanized, or died), and street address for owner or location of pickup, city, and zip code.
Basic data cleanup was performed to eliminate records for cats that originated from locations outside of Boston and remove duplicate entries, such as cats that were readmitted to the shelter after a period in a foster home. Cats relinquished by their owners specifically for euthanasia were also excluded for several reasons. First, it was the author's strong belief that these deaths over-whelmingly reflected natural, nonpreventable deaths attributable to serious health problems, advanced age, or both, rather than trivial reasons. This is a position consistent with findings from a national shelter study.22 Also, reporting guidelines established in the Asilomar Accords document23 recommend excluding owner-requested euthanasia from the calculation of live-release rates when assessing shelter data statistically. Finally, in the author's experience in Boston, clients select a shelter for this service because of either the cost of euthanasia and disposal at a private veterinary hospital or the reluctance of some veterinarians in private practice to perform euthanasia for a new client. Personnel at the ARL-B or MSPCA will not perform convenience euthanasia of healthy potentially adoptable pets and will request that those animals be signed over to the organization for evaluation and potential placement.
In a study of Boston neighborhoods, Chen et al5 used a broad definition of human premature deaths (ie, deaths before age of 75 years, without regard to cause). To use a similarly broad definition of cat deaths for purposes of this study, given the fact that accurate age data are difficult to verify in a shelter setting, all shelter-associated cat deaths (with the exception of owner-requested euthanasia) were included in the count of cat deaths. This included cats that were known to be nonadoptable at intake due to disease or injury or that were deemed untreatable after evaluation, as well as animals that died from injuries or illness or were brought in dead on arrival as strays.
Geocoding of addresses—Geocoding is the process of converting street address information to latitude and longitude coordinates by use of standard reference data files of road centerlines and address ranges. To optimize the success of address matching, a list of all street names in the study records was printed, and each street name was first checked for spelling accuracy by use of 2 separate databases of Boston street addresses.b,c Corrections were made when errors were identified. Extraneous information, such as apartment numbers, was manually removed so that only key address information (ie, house number, street name, and street type) needed for address matching remained in a single address field. Following this verification process, a total of 15,566 records were suitable for geocoding. An additional 1,763 records did not contain sufficient information to attempt to assign a street address. For example, these may have been records for stray cats, cats abandoned at the shelter, or cats brought in by animal control personnel or records with incomplete street data. Those records were retained for city-wide per capita outcomes analysis.
Initial geocoding of 15,566 records was performed by use of commercial geospatial softwared and an address locator created from a commercially available street database.e On the basis of the street address and zip code, 13,474 (86.6%) records matched on this first pass. All unmatched or tied records were then geocoded by use of a second address locator created from a street databasec maintained by the Boston Water and Sewer Commission, which contained more up-to-date street addresses but without zip code information. A total of 1,385 additional addresses were matched, and the remaining 707 unmatched or tied locations were manually reviewed and rematched if possible. Overall, 15,285 of 17,587 (86.9%) cat intake records were ultimately geocoded to a Boston street address.
Linkage to neighborhood data—Boundaries for 16 neighborhood polygons were constructed by assigning each of 156 Boston census tracts to neighborhoods as defined by the Boston Public Health Commission.5 Census tract demographic data were downloaded from US 2000 Census Summary File 3.24 Geospatial softwared was used to join geocoded addresses to neighborhoods and census tracts, and a third-party extension for the geospatial software was used to count the number of cat intakes and deaths in each geographic entity.f Cat intake and outcome data, now summarized by neighborhood and census tract and linked to community-level demographic variables, were exported into a spreadsheet in preparation for statistical analysis.g
Statistical analysis and mapping—By use of the projected 2007 human population as the denominator, mean annual per capita rates of cat intakes and deaths over the 5-year period were calculated by neighborhood and by census tract.25 These were the primary outcome variables. As previously described,5 a Bayesian approach was used to account for instability associated with calculation of small area rates. Smoothed estimates of cat mortality rates by census tract were obtained by fitting a Poisson mixed-effects model by use of computer softwareh,i for the Bayesian analysis of complex statistical models and a Markov chain Monte Carlo algorithm.5 Univariate relationships between cat deaths and census tract socioeconomic variables were explored via correlation analyses.26 These variables were selected primarily because they have either previously been associated with neighborhood deprivation and disadvantage6 or premature death5 in human studies. Relative risks for death and 95% credible intervals (ie, 95% Bayesian confidence intervals) were calculated by fitting Poisson mixed-effects models as described.
Age-adjusted premature death rates among humans in Boston neighborhoods for the period 1999 through 2001 were obtained from a report5 published by the Public Health Disparities Geocoding Project. The relationship between age-adjusted human premature death rate and per capita cat death rate was assessed by use of linear regression analysis26 performed with commercially available statistical software.g Commercially available geospatial software was used for choropleth mapping.d For all analyses, a value of P < 0.05 was considered significant.
Results
Geographic disparities in cat welfare—Mean annual rates of cat intakes and deaths each varied dramatically over the 16 neighborhoods of Boston (intake rates, 0.85 to 10.30 cats/1,000 persons; mortality rates, 0.27 to 3.90 cats/1,000 persons; Figure 1). The mean ± SD number of people in the census tracts was 3,701 ± 1,509. Over the 156 census tracts, mean annual intake rates ranged from 0.10 to 22.1 cats/1,000 persons and mortality rates ranged from 0.15 to 6.47 cats/1,000 persons, with considerable variation among tracts within a given neighborhood. Census tracts with the highest cat mortality rates (4 to 7 cats/1,000 persons) were often no further than 1 to 2 miles from tracts with the lowest rates (< 1 cat/1,000 persons). These data and outcomes from 2,032 records (accumulated over a 5-year period) that either were not suitable for geocoding or could not be matched to a street address combined with projected summary intakes and outcomes from the City of Boston Animal Control shelter indicated that the city-wide mean annual intake and mortality rates for cats were approximately 6.60 cats/1,000 persons and approximately 2.60 cats/1,000 persons, respectively.
Mean annual cat intake (A and B) and mortality (C and D) rates (number of cats/1,000 persons) by neighborhood (A and C) and census tract (B and D) determined from data for a 5-year period (2004 through 2008) provided by 2 animal shelters serving Boston. The scale in panel D applies to all panels. Neighborhoods are designated as follows: AB = Allston-Brighton, BB = Back Bay, CT = Charlestown, EB = East Boston, FW = Fenway, HP = Hyde Park, JP = Jamaica Plain, MP = Mattapan, ND = North Dorchester, NE = North End, RD = Roslindale, RX = Roxbury, SB = South Boston, SD = South Dorchester, SE = South End, and WR = West Roxbury.
Citation: American Journal of Veterinary Research 71, 2; 10.2460/ajvr.71.2.161
Mean annual cat intake (A and B) and mortality (C and D) rates (number of cats/1,000 persons) by neighborhood (A and C) and census tract (B and D) determined from data for a 5-year period (2004 through 2008) provided by 2 animal shelters serving Boston. The scale in panel D applies to all panels. Neighborhoods are designated as follows: AB = Allston-Brighton, BB = Back Bay, CT = Charlestown, EB = East Boston, FW = Fenway, HP = Hyde Park, JP = Jamaica Plain, MP = Mattapan, ND = North Dorchester, NE = North End, RD = Roslindale, RX = Roxbury, SB = South Boston, SD = South Dorchester, SE = South End, and WR = West Roxbury.
Citation: American Journal of Veterinary Research 71, 2; 10.2460/ajvr.71.2.161
Mean annual cat intake (A and B) and mortality (C and D) rates (number of cats/1,000 persons) by neighborhood (A and C) and census tract (B and D) determined from data for a 5-year period (2004 through 2008) provided by 2 animal shelters serving Boston. The scale in panel D applies to all panels. Neighborhoods are designated as follows: AB = Allston-Brighton, BB = Back Bay, CT = Charlestown, EB = East Boston, FW = Fenway, HP = Hyde Park, JP = Jamaica Plain, MP = Mattapan, ND = North Dorchester, NE = North End, RD = Roslindale, RX = Roxbury, SB = South Boston, SD = South Dorchester, SE = South End, and WR = West Roxbury.
Citation: American Journal of Veterinary Research 71, 2; 10.2460/ajvr.71.2.161
Demographic variables versus cat mortality rates—Census tract socioeconomic variables that are known to be indicative of neighborhood deprivation5,6 were all significantly associated with cat mortality rates in the present study (Table 1). The association between census tract poverty and cat mortality rate was further explored because of its documented association with human premature mortality rate in other studies.5,16 Indeed, when examined categorically by use of the same cut points reported by Chen et al,5 annual per capita cat mortality rate associated with the 70 poorest tracts (ie, ≤ 20% of persons below the US Census threshold for poverty) was 2.36 cats/1,000 persons; this value was more than 3.5 times as high as the rate associated with the 7 wealthiest census tracts (ie, < 5% of persons below the US Census threshold for poverty), for which the rate was 0.63 cats/1,000 persons. It is important to note that mortality rates in these most deprived tracts varied by a factor of more than 40 (0.15 to 6.47 cats/1,000 persons/y). Not surprisingly, the univariate relationships between overall poverty and cat intakes (r = 0.27; P = 0.001) and deaths (r = 0.22; P = 0.007) were weak. Family poverty (defined as families that have an annual income less than the US Census threshold for poverty and include children < 18 years old) was examined and had stronger relationships with respect to both cat intakes (r = 0.53; P < 0.001) and deaths (r = 0.46; P < 0.001). These findings were similar to those for the relationship between public assistance recipients and cat intakes or deaths (relative risk, 2.65; 95% credible interval, 2.06 to 3.19).
Association of demographic variables in 156 census tracts in Boston (as reported in US 2000 Census Summary File 324) with mean annual cat mortality rate determined from data for a 5-year period (2004 through 2008) provided by 2 animal shelters serving that city.
| Census table | Demographic variable description | Pearson correlation (r) | Mean annual cat mortality rate (No. of cats/1,000 persons)* | ||||
|---|---|---|---|---|---|---|---|
| P value | Intercept | Relative risk of death† | Lower 95% CI | Upper 95% CI | |||
| P64 | Receive public assistancea | 0.557 | < 0.001 | 1.16 | 2.56 | 2.06 | 3.19 |
| P43 | Unemploymentb | 0.243 | 0.002 | 1.53 | 1.61 | 1.14 | 2.26 |
| H20 | Crowded housingc | 0.281 | < 0.001 | 1.45 | 1.38 | 1.16 | 1.65 |
| P90 | Families with children in povertyd | 0.457 | < 0.001 | 1.27 | 1.38 | 1.25 | 1.51 |
| P15 | Female-headed households with childrene | 0.675 | < 0.001 | 0.87 | 1.45 | 1.36 | 1.54 |
| P37 | Lack of high school diplomaf | 0.456 | < 0.001 | 0.99 | 1.31 | 1.21 | 1.41 |
| P87 | Overall povertyg | 0.217 | 0.007 | 1.56 | 1.10 | 1.00 | 1.21 |
| P50 | Males in management or professionh | −0.580 | < 0.001 | 5.10 | 0.74 | 0.70 | 0.79 |
Smoothed mortality rate was calculated by use of a Poisson mixed-effects model.
Relative risk was calculated for each 10% absolute increase in corresponding census variable (eg, an increase in families with children in poverty from 10% to 20% would be associated with a 38% increase in cat mortality rate).
CI = Credible interval.
aPercentage of households receiving public assistance.
bPercentage of population ≥ 16 years old and unemployed.
cPercentage occupied housing units having > 1 person/room.
dPercentage of families in poverty with children < 18 years old.
ePercentage of families headed by a female with children.
fPercentage of population ≥ 25 years old that have not graduated from high school.
gPercentage of persons with status below Federal definition of poverty.
hPercentage of males in management and professional occupations in the employed population ≥ 16 years old.
Human premature mortality rates versus cat mortality rates—There was a strong linear relationship (r = 0.71) between mean annual number of cat deaths/1,000 persons (for the years 2004 through 2008) and previously reported age-standardized human premature deaths across 16 Boston neighborhoods (for the years 1999 through 2001), which explained half of the variance in cat mortality rates (R2 = 0.50; P = 0.002; Figure 2). Two neighborhoods (one of which was also regarded as anomalous in the human data) had less than half the expected number of deaths in the cat population. When the relationship was examined without those 2 datum points, the linear relationship became even stronger (r = 0.88), explaining more than three-fourths of the variance in cat mortality rates (R2 = 0.77; P < 0.001).
Comparison of age-standardized human premature (< 75 years old) deaths/100,000 persons (data for years 1999 through 2001 reported by Chen et al5) and mean annual cat mortality rates (number of cats/1,000 persons for years 2004 through 2008) in 16 Boston neighborhoods. Datum point A (in dotted circle) represents the Fenway neighborhood, and datum point B (in dotted circle) represents the South Boston neighborhood. The solid line represents the regression line through all points, and the dashed line represents the regression line with the 2 circled points omitted.
Citation: American Journal of Veterinary Research 71, 2; 10.2460/ajvr.71.2.161
Comparison of age-standardized human premature (< 75 years old) deaths/100,000 persons (data for years 1999 through 2001 reported by Chen et al5) and mean annual cat mortality rates (number of cats/1,000 persons for years 2004 through 2008) in 16 Boston neighborhoods. Datum point A (in dotted circle) represents the Fenway neighborhood, and datum point B (in dotted circle) represents the South Boston neighborhood. The solid line represents the regression line through all points, and the dashed line represents the regression line with the 2 circled points omitted.
Citation: American Journal of Veterinary Research 71, 2; 10.2460/ajvr.71.2.161
Comparison of age-standardized human premature (< 75 years old) deaths/100,000 persons (data for years 1999 through 2001 reported by Chen et al5) and mean annual cat mortality rates (number of cats/1,000 persons for years 2004 through 2008) in 16 Boston neighborhoods. Datum point A (in dotted circle) represents the Fenway neighborhood, and datum point B (in dotted circle) represents the South Boston neighborhood. The solid line represents the regression line through all points, and the dashed line represents the regression line with the 2 circled points omitted.
Citation: American Journal of Veterinary Research 71, 2; 10.2460/ajvr.71.2.161
Discussion
The primary goal of the present study was to determine whether it was possible to detect disparities in welfare of cats, as indicated by variations in rates of cat intake and death at animal shelters, within neighborhoods, and across census tracts in a large US city (Boston). Results of the study indicated that cat mortality rates varied by a factor of approximately 14 by neighborhood and by a factor of > 40 across census tracts. A second goal of the present study was to determine whether cat mortality rates had any relationship with disparities in human health, as reflected by premature mortality rates. Cat mortality rate did indeed have a strong, positive, linear relationship with human premature mortality rate, with human premature deaths explaining 77% of the variation in cat deaths. Other than a study27 of US school children involving data collected from 1986 through 1988 in which a significant association between highest level of parental education achieved and reported death of a pet was identified, this is the first study in which a relationship between human premature deaths and cat deaths has been demonstrated to the author's knowledge and reaffirms the close connection between the well-being of humans and other animals.
In the present study, disparities in cat mortality rate were clustered. For example, mean annual mortality rate was < 1.0 cat/1,000 persons in 6 of the 16 neighborhoods and < 2.0 cats/1,000 persons in another 4 neighborhoods. The highest cat mortality rates were associated with the remaining 6 neighborhoods, albeit with considerable variation among census tracts within those neighborhoods. Seventeen of the 19 census tracts with the highest mortality rates (4 to 7 cats/1,000 persons) were geographically contiguous. Census tracts (n = 37) in which the annual cat mortality rates were ≥ 3 cats/1,000 persons accounted for only 21.5% of the city population, but for half of the cat deaths. Clustering notwithstanding, census tracts in the lowest mortality rate category were located within a 1-mile radius of the tracts in the high–mortality rate categories.
The percentage of persons living below the US Census Bureau threshold for poverty has been reported to estimate expected socioeconomic gradients in health across a wide range of outcomes in the population overall as well as within racial and ethnic subgroups.16,28 Therefore, the correlation between overall census tract poverty and cat mortality rate (r = 0.22; P = 0.007) determined in the present study was less than might be expected. This could reflect Boston's large student population; low income associated with student status likely has fundamentally different implications with respect to being disadvantaged than has family poverty. As Chen et al5 have stressed in their report, statistical associations of demographic variables with health outcomes must be interpreted in light of knowledge of the social geography of neighborhoods. This may identify potential important statistical confounders. The 2 outlier datum points identified in the present study illustrate the merit of exploring unmeasured confounding variables. The Fenway neighborhood, which is known to have a particularly high concentration of students, had much lower than expected annual per capita rates of cat intake and death (1.42 and 0.49 cats/1,000 persons, respectively). It is plausible that students have low rates of cat ownership or that the social geography of student communities is more favorable to the welfare of cats (eg, greater retention and fewer strays). With respect to the lower than expected cat mortality rate in South Boston, that neighborhood is served by an established, small, adoption-guarantee shelter that provides low-cost veterinary services to the community and supports cat trap-neuter-return programs. Indeed, the fact that cat mortality rate was approximately 50% lower than expected in South Boston could be considered evidence of the potential impact of increased access to veterinary care for underserved populations. These proposed explanations are of course conjectural and remain to be proved.
A strength of the present study is that geospatial methods allowed accurate connection of cat records to actual human population data, even to the level of the census tract (mean ± SD number of people/tract, 3,701 ± 1,509). Both the ARL-B and MSPCA receive cats from communities throughout the greater Boston area, including much of eastern Massachusetts and beyond; hence, accurate determination of which cats came from within Boston city limits was essential for putting citywide intake and mortality rates in proper perspective. A second important consideration was the availability of data from most of the shelters serving the population. Had data from only a single organization been used or too broad a population counted in the denominator, these per capita estimates would have been considerably underestimated.
To derive the most conservative estimates of cat mortality rates in the present study, cats that had untreatable injuries or terminal disease at intake and those that were dead on arrival at the shelter were included; thus, the analyzed data were derived from a combination of savable, manageable, and untreatable cats. Nevertheless, the estimated total annual city-wide shelter-associated mortality rate for cats (approx 2.6 cats/1,000 persons), in conjunction with a projection for annual city-wide mortality rate for dogs (approx 1.0 dog/1,000 persons), suggests that the combined mortality rate for cats and dogs in the city of Boston (< 4.0 cats and dogs/1,000 persons) is well within the range reported for communities deemed as having the most successful sheltering and animal control programs (eg, approx 2 to 7 dog and cat deaths at shelters/1,000 persons/y).12,13 However, the small-area analysis reported here revealed that even within a city that has a substantial infrastructure to assist unowned animals, areas with considerable health disparities can exist. These findings would have been obscured if mortality rate data were averaged for the entire city. Similarly, improvements in these small areas would be more difficult to detect if only city-wide data were being monitored.
The area-based analysis used in the study of this report here should not be confused with an individuallevel analysis. However, area and individual measures of wealth, education, and occupation are moderately to highly correlated.29 It has been argued, for the human population, that aptly chosen ABSMs provide estimates of health inequalities that are comparable with those based on individual-based socioeconomic measures, and that the results provide a conservative estimate of socioeconomic inequalities in health.21 Furthermore, disadvantaged neighborhoods themselves are associated with increased human mortality rates, independent of individual-level effects.20
The socioeconomic heterogeneity of census tracts within neighborhoods typically biases any estimates of association toward the null. Therefore, the correlations determined in the present study are likely underestimates of the true association of those factors with cat mortality rate. It might seem intuitively simpler to rely on zip codes instead of census tracts when trying to assess area-level effects because zip codes are easily captured in shelter databases as part of client addresses. However, zip code groupings have drawbacks.30,31 Standard 5-digit zip codes are no longer tied to the human population in the US Census, and boundaries of zip code areas are not well documented and continue to change over time.31 Zip code tabulation areas introduced in the 2000 US Census to attempt to align populations with zip codes may also be markedly misaligned.31,32 Geospatial analysis offers a much more precise approach.
Because criteria for recording outcomes within the ARL-B and MSPCA databases changed over time and there were differences in how cat deaths were coded between the 2 organizations, it was not possible to investigate cause-specific mortality rates for cats in the present study. Thus, these mortality data are best thought of as reflective of a spectrum of societal ills that result in an animal being relinquished, producing an unwanted litter, sustaining an injury or being hit by a vehicle, or developing a vaccine-preventable illness, all of which could result in premature death. A broad view of death when looking at underserved populations has relevance. Indeed, human deaths attributable to “signs, symptoms, and ill-defined conditions” (International Classification of Diseases codes 780 to 79933) are more prevalent in minority populations, compared with findings in white populations.34
Use of records from a 5-year period (2004 through 2008) provided a sufficient number of cat intakes and deaths to make inferences at the level of the census tract, but averaging data collected over this period may have masked decreases that have occurred in more recent years. It is unknown whether inability to geocode data from the City of Boston Animal Control and some data from ARL-B and MSPCA resulted in any bias with respect to comparison of disparities in intake or death at either the neighborhood or census tract level. If any bias existed, it would most likely result in underestimation of the health disparities reported here because the City of Boston Animal Control shelter is adjacent to some of the most disadvantaged areas of the city and, if anything, would likely serve those areas disproportionately. Also, there were no data available from 1 small shelter serving South Boston. That agency was an adoption-guarantee shelter with a strong focus on providing low-cost veterinary care, so the effect of this exclusion would solely be to underestimate per capita intakes and not mortality rates.
In the present study, many factors known to indicate neighborhood deprivation were associated with an increased relative risk of cat death in Boston. These findings further underscore the close relationship between the well-being of humans and other animals. Although causal pathways remain to be established, it is likely that disparities in cat health and welfare result from the same combination of inequities (eg, violence, lack of education, poverty, lack of opportunity, and reduced access to health care) that act in concert at both the community and individual level to foster health disparities among people. Recognition of marked disparities in cat intake and mortality rates within neighborhoods and across census tracts suggests that targeted, small-area intervention should be explored as a means of reducing disparities in cat welfare.
Besides improving cat welfare, elimination of social disparities in cat health may help address disparities in human health. A large, albeit still inconclusive, body of literature has accumulated in which companion animal interaction and caregiving are linked with a protective effect against a variety of adverse human health events.35 Given the implications of these claims, efforts to strengthen, preserve, and promote beneficial human-animal relationships should be taken seriously for disadvantaged as well as more-privileged communities.
An emerging understanding within the animal welfare community is that as the number of unwanted animals becomes more controlled within a community, a greater proportion of animals handled by shelters will be more difficult to place and require considerable rehabilitation prior to adoption. Therefore, historical shelter outcome measures such as so-called save rates or live-release rates will have less relevance as indicators of performance, and per capita intake and mortality measures, which are benchmarked to the human population served, will become the most useful. A geospatial approach provides a way with which these shelter events can be accurately tied to the appropriate underlying human population, which is a necessary prerequisite to calculating accurate per capita rates. As more and more communities across the United States come closer to eliminating preventable animal deaths in shelters, detection of small-area gradients in health may be necessary to identify the areas of most need as well as for monitoring progress. Both may be obscured if outcome data are simply averaged across an entire city or community. The precision of geospatial tools enables monitoring animal welfare outcomes at even small geographic levels, such as the census tract, where the effects of interventions such as increasing access to veterinary care (eg, spay-neuter, vaccinations, and preventive health care) may be most pronounced or immediately visible.
ABBREVIATIONS
| ABSM | Area-based socioeconomic measure |
| ARL-B | Animal Rescue League of Boston |
| MSPCA | Massachusetts Society for the Prevention of Cruelty to Animals |
Chameleon/CMS integrated shelter management software, release 4.636, Chameleon Software Products. Available at: www.chameleonbeach.com. Accessed Aug 13, 2009.
City of Boston Public Works Department's Boston street book. Available at: www.cityofboston.gov/publicworks/streetbook. Accessed Aug 13, 2009.
Boston street list, provided by City of Boston Water and Sewer Commission, Boston, Mass.
ESRI ArcMap 9.3, ESRI Inc, Redlands, Calif. Available at: www.esri.com. Accessed Aug 13, 2009.
ArcGIS StreetMap USA, ESRI Inc, Redlands, Calif. Available at: www.esri.com. Accessed Aug 13, 2009.
Beyer HL. Hawth's Analysis Tools for ArcGIS, version 3.2. Available at: www.spatialecology.com/htools/overview.php. Accessed Aug 13, 2009.
SPSS for Windows, version 15.0.0, SPSS Inc, Chicago, Ill.
The BUGS Project: WinBUGS. Bayesian inference using Gibbs sampling. MRC Biostatics Unit, Cambridge, England. Available at: www.mrc-bsu.cam.ac.uk/bugs. Accessed Aug 13, 2009.
R, version 2.8.1, The R Foundation for Statistical Computing, Vienna, Austria. Available at: www.r-project.org. Accessed Aug 13, 2009.
References
- 1.
Adler NE, Rehkopf DH. US disparities in health: descriptions, causes, and mechanisms. Annu Rev Public Health 2008;29:235–252.
- 2.↑
PolicyLink Health Disparities Team. Reducing health disparities through a focus on communities. November, 2002. Available at: www.policylink.org/atf/cf/%7B97c6d565-bb43-406d-a6d5eca3bbf35af0%7D/REDUCINGHEALTHDISPARITIES_FINAL.pdf. Accessed Aug 13, 2009.
- 3.
Boston Public Health Commission. The Disparities Project. Data report: a presentation and analysis of disparities in Boston. Available at: www.bphc.org/programs/healthequitysocialjustice/toolsandreports/Forms%20%20Documents/Center%20Reports%20and%20Tools/BPHCOHEdatareport.pdf. Accessed Aug 13, 2009.
- 4.↑
Mokdad AH, Marks JS, Stroup DF, et al. Actual causes of death in the United States, 2000 (Erratum published in JAMA 2005;293:293–294 and JAMA 2005;293:298). JAMA 2004;291:1238–1245.
- 5.↑
Chen JT, Rehkopf DH, Waterman PD, et al. Mapping and measuring social disparities in premature mortality: the impact of census tract poverty within and across Boston neighborhoods, 1999–2001. J Urban Health 2006;83:1063–1084.
- 6.↑
Messer LC, Laraia BA, Kaufman JS, et al. The development of a standardized neighborhood deprivation index. J Urban Health 2006;83:1041–1062.
- 7.
Phelan JC, Link BG, Diez-Roux A, et al. “Fundamental causes” of social inequalities in mortality: a test of the theory. J Health Soc Behav 2004;45:265–285.
- 8.↑
The Public Health Disparities Geocoding Project Monograph. Geocoding and monitoring US socioeconomic inequalities in health: an introduction to using area-based socioeconomic measures. Available at: www.hsph.harvard.edu/thegeocodingproject. Accessed Aug 13, 2009.
- 9.↑
AVMA. One Health: a new professional imperative. 2008. Available at: www.avma.org/onehealth/onehealth_final.pdf. Accessed Aug 13, 2009.
- 10.
Olson PN, Moulton C, Nett TM, et al. Pet overpopulation: a challenge for companion animal veterinarians in the 1990s. J Am Vet Med Assoc 1991;198:1151–1152.
- 11.
Patronek GJ, Rowan AN. Determining dog and cat numbers and population dynamics. Anthrozoos 1995;8:199–205.
- 12.↑
Maddie's Fund Statistical Reports. Available at: www.maddiesfund.org/No-Kill_Progress/Statistical_Reports.html. Accessed Aug 13, 2009.
- 13.↑
Maddie's Fund. Saving all of our healthy and treatable shelter dogs and cats by 2015 is not only possible, it's probable. Available at: www.maddiesfund.org/Documents/No%20Kill%20Progress/Getting%20to%20No%20Kill%20by%202015.pdf. Accessed Aug 13, 2009.
- 14.
Cromley RK, McLafferty SL. Analyzing access to health services. In: GIS and public health. New York: The Guilford Press, 2002;233–258.
- 15.
Krieger N. Overcoming the absence of socioeconomic data in medical records: validation and application of a census-based methodology. Am J Public Health 1992;82:703–710.
- 16.
Krieger N, Chen JT, Waterman PD, et al. Race/ethnicity, gender, and monitoring socioeconomic gradients in health: a comparison of area-based socioeconomic measures—the public health disparities geocoding project. Am J Public Health 2003;93:1655–1671.
- 17.
Krieger N, Chen JT, Waterman PD, et al. Choosing area based socioeconomic measures to monitor social inequalities in low birth weight and childhood lead poisoning: The Public Health Disparities Geocoding Project (US). J Epidemiol Community Health 2003;57:186–199.
- 18.
Whitehead SJ, Cui KX, De AK, et al. Identifying risk factors for underimmunization by using geocoding matched to census tracts: a statewide assessment of children in Hawaii. Pediatrics 2007;120:e535–e542.
- 19.
Pomerantz WJ, Dowd MD, Buncher CR. Relationship between socioeconomic factors and severe childhood injuries. J Urban Health 2001;78:141–151.
- 20.↑
Mujahid MS, Diez-Roux AV, Morenoff JD, et al. Assessing the measurement properties of neighborhood scales: from psychometrics to ecometrics. Am J Epidemiol 2007;165:858–867.
- 21.↑
Subramanian SV, Chen JT, Rehkopf DH, et al. Comparing individual- and area-based socioeconomic measures for the surveillance of health disparities: a multilevel analysis of Massachusetts births, 1989–1991. Am J Epidemiol 2006;164:823–834.
- 22.↑
Kass PH, New JC, Scarlett JM, et al. Understanding animal companion surplus in the United States: relinquishment of nonadoptables to animal shelters for euthanasia. J Appl Anim Welf Sci 2001;4:237–248.
- 24.↑
US Census Bureau. American FactFinder. Available at: factfinder.census.gov/home/saff/main.html?_lang=en. Accessed Aug 13, 2009.
- 25.↑
ESRI. ESRI® demographic update methodology: 2007/2012. Redlands, Calif: ESRI, 2007. Available at: www.esri.com/library/whitepapers/pdfs/demographic-update-methodology-2007.pdf. Accessed Aug 13, 2009.
- 27.↑
Moloo J, Jackson KL, Waller JL, et al. Xenotransmission of the socioeconomic gradient in health? A population based study. BMJ 1998;317:1686.
- 28.
Subramanian SV, Chen JT, Rehkopf DH, et al. Racial disparities in context: a multilevel analysis of neighborhood variations in poverty and excess mortality among black populations in Massachusetts. Am J Public Health 2005;95:260–265.
- 29.↑
Diez-Roux AV, Kiefe CI, Jacobs DR Jr, et al. Area characteristics and individual-level socioeconomic position indicators in three population-based epidemiologic studies. Ann Epidemiol 2001;11:395–405.
- 30.
Boscoe FP, Pickle LW. Choosing geographic units for choropleth rate maps, with an emphasis on public health applications. Cartogr Geogr Inform Sci 2003;30:237–248.
- 31.↑
Rushton G, Armstrong MP, Gittler J, et al. Geocoding in cancer research: a review. Am J Prev Med 2006;30(suppl 2):S16–S24.
- 32.
Hurvitz P. What is the difference between ZIP code “boundaries” and ZCTA areas? Available at: gis.washington.edu/∼phurvitz/zip_or_zcta/index.html. Accessed Aug 13, 2009.
- 33.↑
The International Classification of Diseases, 9th revision. Available at: icd9cm.chrisendres.com/icd9cm/index.php?action=child&recordid=8065. Accessed Aug 13, 2009.
- 34.↑
Becker TM, Wiggins CL, Key CR, et al. Symptoms, signs, and ill-defined conditions: a leading cause of death among minorities. Am J Epidemiol 1990;131:664–668.
- 35.↑
Barker SB, Wolen AR. The benefits of human-companion animal interaction: a review. J Vet Med Educ 2008;35:487–495.
