Spatial epidemiologic analysis of relative collision risk factors among urban bicyclists and pedestrians

Abstract

Pedestrians and bicyclists are the victims of countless car crashes in U.S. cities as well as around the world. Yet, many dimensions of their involvement in crashes remain rather poorly known. In this article, we follow a spatial epidemiologic approach to study the relative risk factors of bicycle and pedestrian crashes at the neighborhood level in the City of Buffalo, NY over a two-year period. The analysis examines physical road characteristics such as roadway and intersection functional classes, urban density and type of development-business or residential, as well as socio-economic and demographic variables to identify discriminating risk factors between the two non-motorized transportation modes. The analysis underscores significant differences tied to neighborhood ethnicity, educational attainment and land use, while physical characteristics of the road infrastructure register as marginally discriminating factors. Income related socio-economic status is not found to play a prominent role.

Epidemiology and spatial examination of bicycle-motor vehicle crashes in Iowa, 2001–2011

“…Our findings somewhat differed with two spatial studies of pedestrians and bicyclists that examined similar person (age, gender) and population characteristics (population density, education, income) by census tracts in Buffalo, New York (Delmelle and Thill, 2008;Delmelle et al, 2012). For example, they found population density and education (no high school degree) to be important risk factors for bicyclists, while we did not.…”

“…The inconsistencies between our study and these two conducted in Buffalo may be due to differences in outcomes. Their outcomes were relative risk of bicycle vs. pedestrian crashes (Delmelle et al, 2012) and traffic hazard intensity (Demelle and Thill, 2008) and they did not stratify by intersection and non-intersection crashes. Our study also included crash characteristics that they did not examine, such as contributing circumstances, motor vehicle action, and major cause.…”

On accommodating spatial dependence in bicycle and pedestrian injury counts by severity level

“…Before CPB, a common approach was to map the count variable into an approximate continuous variable (typically also applying a log-transformation to ensure positive predictions, and sometimes also normalizing by an exposure measure to obtain crash rates or taking ratios of different types of crashes), and then apply well-established estimation methods developed for continuous models. Examples of such efforts in the safety literature include LaScala et al (2000), Quddus (2008), Ha and, and Delmelle et al (2011). While useful, these efforts may be viewed as approximations, since they generate "continuous" variables from underlying count data.…”

“…We do so because the more disaggregate spatial units (roadway street segment, intersection, Census block, and Census block group) can routinely experience zero pedestrian and bicyclist-related crashes for multiple years at a stretch, which reduces the variability of the count variables across such disaggregate spatial units and decreases our ability to tease out the risk factors associated with pedestrian and bicyclist crash involvement. The use of the more aggregate Census tract level avoids these problems, while also representing a reasonably homogeneous spatial unit of an urban area (see Delmelle et al, 2011).…”

Built Environmental Designs in Promoting Pedestrian Safety

“…Recently, both researchers and practitioners from urban and transportation planning have considered the importance of built environments in facilitating or constraining the likelihood of pedestrian injuries [7]. Specifically, the influences of development patterns on pedestrian crash frequency have been emphasized and examined [2,[8][9][10][11][12][13]. However, few studies have investigated development patterns' impacts on pedestrian injury severity.…”