Author(s): Batomen, Cloutier, Carabali, Hagel, Bondy, Di Ruggiero
Several Canadian cities are implementing traffic calming measures (TCMs) which consist of engineering physical modifications of the built environment, to prevent traffic collisions. TCMs are generally implemented following an isolated or area-wide approach. Although many studies have reported reductions in collisions and injuries tied to the presence of isolated TCM, they were limited by their pre-post designs, which cannot account for the underlying trend in the number of collisions, regression to the mean, collision migration, and dispersion. Compared to the isolated approach, the area-wide approach of TCMs has an advantage given that it should not be limited by collision migration and dispersion mentioned above. However, previous studies were still limited by their pre-post designs.
This study aims to assess the effectiveness of an area-wide approach of TCM implementation to reduce collisions, injuries, and fatalities using a longitudinal design.
Seven TCMs (crosswalk, curb extension, diverter, flashing crosswalk, mid-block narrowing, raised crosswalk, and speed bump) were evaluated in the city of Montreal from 2012 to 2019. The number of severe and deadly collisions was the primary outcome. Secondary outcomes included severe and deadly car occupants and pedestrians’ injuries. Two units of analysis were considered (intersections and census tracts). Fixed effects regressions were performed (Conditional logistic for intersections and Conditional Poisson for census tracts), using each unit as its own control while accounting for the underlying trend in outcomes. Lagged values of the cumulative number of TCMs at a given year was used as the independent variable in our models.
There was a steady increase in the number of TCM overtime from a cumulative total of 209 in 2012 to 1,769 installations in 2018. Speed bumps (48%) and curb extensions (28%) were the most implemented TCM and most of the installations occurred on local streets (72%). During the study period, 1,751 severe or deadly collisions were recorded, and two-thirds occurred on arterial roads. Overall, there were no associations between the presence of TCM and the primary outcome. However, subgroup analyses of intersections on local roads suggest a 60% reduction in collisions, OR:0.27 (95%CI: 0.07 – 1.00). Regarding secondary outcomes, they were some suggestions of an 11% reduction in car occupants’ deaths and injury rates, RR: 0.89 (95%CI: 0.74 – 1.06) when using census tracts as the unit of analysis.
Although most collisions occur on arterial roads, TCMs are generally implemented on local roads. A 60% reduction in collisions on intersections made of local roads was observed. However, when considering all types of intersections or census tracts as units of analysis, we did not find any associations with study outcomes. Despite the strength of our design, unmeasured time-varying confounding cannot be excluded. Further studies applying mediation analyses to disentangle the potential mediation and confounding effects of traffic volume on the relation between TCM and collisions are needed. Sensibility analyses and Bayesian spatial models should also be explored to account for measurement errors and the spatial correlation in the data.
Our results suggest that TCM reduces severe and deadly collisions on intersections made of local roads.