Spatial analysis of macro-level bicycle crashes using the class of conditional autoregressive models

Area of Science:

• Transportation Safety
• Spatial Statistics
• Epidemiology

Background:

• Bicycle crashes exhibit spatial clustering at the census block group level.
• Understanding contributing factors is crucial for targeted safety interventions.
• Existing models may not fully account for spatial dependencies in crash data.

Purpose of the Study:

• To investigate the relationship between bicycle crash frequency and contributing factors at the census block group level in Florida.
• To apply and compare Conditional Autoregressive (CAR) models within a hierarchical Bayesian framework.
• To identify variables significantly associated with total and fatal-and-severe injury bicycle crashes.

Main Methods:

• Utilized four years (2011-2014) of bicycle crash data from Florida.
• Employed hierarchical Bayesian Conditional Autoregressive (CAR) models, specifically Besag's and Leroux's models.
• Assessed model performance and identified significant variables using 95% Bayesian credible intervals.

Main Results:

• Besag's CAR model demonstrated superior fit compared to Leroux's model for crash prediction.
• Identified 21 significant variables for total crashes and 18 for fatal-and-severe injury crashes.
• Positive associations found for population, vehicle miles traveled, urban road density, and bicycle activity; negative associations for educational attainment and rural road density.

Conclusions:

• Spatial modeling is essential for accurately analyzing bicycle crash data.
• Demographic, socio-economic, roadway, and activity-related factors significantly influence bicycle crash risk.
• Findings can inform targeted safety strategies to reduce bicycle crashes and injuries.