Environmental correlates of urban dog bites: A spatial analysis

ABSTRACT

To be able to design effective urban public health programs to reduce the risk of dog bites and transmission of disease, the very complex factors that lead to bites need to be considered holistically. This research focuses specifically on the role of spatial and environmental factors as urban public health risks. In doing this, it addresses the following research questions:

(1) What are the relative powers of traditional demographic versus environmental variables in explaining dog bites?

(2) Do different areas of the city evidence different correlates of bites?

The answer to the first is that despite a long tradition in the literature, demographic variables do a relatively poor job of explaining variation in the rates of emergency room visits due to dog bites. Rather, environmental and spatial variables, particularly crime, vacancy, and blight, are better predictors of dog bites than traditional demographic variables such as age and gender. However, even the best-fitting regression model leaves dog bites in many areas of a city unexplained; bite covariates differ by neighborhood. Thus, to effectively address the risk of dog bites in urban areas, different policies are required for different neighborhood conditions.

Funding

This research has been supported by a grant from the S3 Collaborative Grant Program at Michigan State University.

Notes

1. Rates of dog bites appear to have dropped recently in cities across the nation, and this also appears to be the case in Detroit based on 2016 data (Feretti, 2016; National Canine Research Council, n.d.).

2. Various other sources also collect dog bite data in Detroit, including the police department, the U.S. Postal Service, and the Detroit Department of Health and Wellness. Coding and comparison of these data sources with the ER data are included in future research plans.

3. Data Driven Detroit (D3; datadrivendetroit.org) is a statewide organization with a focus on the city of Detroit. D3 houses a comprehensive data system that includes current and historic demographic, socioeconomic, educational, environmental, and other indicators. This data system allows analysts to illustrate complex relationships by combining different data sets at the regional, city, and block levels. D3 is an affiliate of the Michigan Nonprofit Association (MNA).

4. Land bank properties in Detroit are owned by the city, county, or state and include vacant, abandoned, and foreclosed properties, including side lots as well as a demolition program to remove unusable structures.

5. The ER data cover all ZIP codes that make up the city of Detroit. There are several codes that include Detroit as well as parts of another city, specifically Hamtramck, Highland Park, Redford, and Harper Woods.

6. Correlations among the explanatory variables are less than 0.56 in all cases except for median income and gender with a correlation of 0.97. This may be inflating the R² for this equation.

7. Several of the environmental variables indicated in Appendix A were removed from the regression due to multicollinearity. Parks, greenways, bus stops, recreation centers, and nonmotorized routes were all correlated with vacancy, blight, and brownfields at over 0.50. State and local land bank parcels were combined because they were correlated at 0.72. Crime involving weapons was selected for inclusion because it had the strongest individual correlation with bites but also because it is most likely to be indicating drug/dog fighting–related crime, which presumably increases dog bites. Types of dog crime, specifically dogs at large and animal cruelty, were also significantly correlated with crime and so were not included in the regression analysis. Population density and brownfields are correlated at −0.65, but both variables were left in the equation to indicate both lack of density and pollution; vacancy is not correlated with either variable. Full correlation results are available from the authors upon request. Different types of crime were explored separately but again were highly intercorrelated.

8. Dark gray ZIP codes are those that have more bites than predicted by the model; white ones are those with fewer bites than predicted by the model.

9. Based on police data from the same period indicating reports of animals at large.

10. In the interests of space, four illustrative maps are presented for each ZIP code. All maps are available from the authors upon request.

11. Based on police data from the same period.

12. This may particularly be the case for the temporally fixed variables such as the location of schools, day cares, liquor stores, and parks.

Additional information

Funding

This research has been supported by a grant from the S3 Collaborative Grant Program at Michigan State University.

Notes on contributors

Joshua Vertalka

Joshua Vertalka recently received his PhD from the Department of Geography at Michigan State University. He also holds a master of urban and regional planning from Michigan State University. He has been studying the 1918–1919 influenza spread in British India and identifying the spatial correlates that predict dog bite emergency room admissions in Detroit. Currently, Josh is using Twitter to predict influenza emergency room demand in New York City and London. He is also identifying approaches that help augment noisy social media data. Other research interests include urban planning, health geography, volunteered geographic information, and smart cities.

Laura A. Reese

Laura A. Reese is Professor of Urban and Regional Planning and Political Science and the founding Director of the Global Urban Studies Program (GUSP) at Michigan State University. Dr. Reese’s main research and teaching areas are in urban politics and public policy, economic development, local governance, and management in Canada and the United States and animal welfare policy. She has written 11 books and over 100 articles and book chapters in these areas as well as public personnel administration focusing on the implementation of sexual harassment policy.

Melinda J. Wilkins

Melinda J. Wilkins, DVM, MPH, PhD is an Assistant Professor and Director of the Online Master of Science in Food Safety program at Michigan State University. Prior to coming to MSU, she was with the Michigan Department of Community Health, Bureau of Epidemiology, Division of Communicable Disease, and she is an alumnus of the CDC’s Epidemic Intelligence Service. Her areas of research include epidemiologic examination of the human–animal interface, including zoonotic disease surveillance, transmission, and prevention.

Jesenia M. Pizarro

Jesenia M. Pizarro is an Associate Professor in the School of Criminology and Criminal Justice at Arizona State University. Professor Pizarro has worked with various police departments throughout the country in joint efforts to curb violence. Her work focuses on the social ecology and social reaction of homicide. Her published work has appeared in the American Journal of Public Health, Justice Quarterly, Criminal Justice and Behavior, and Homicide Studies.