Data-Driven Safety Screening for Smart-City Urban Bicycle-Route Networks

Smart-city mobility strategies aim to shift urban car trips to cycling, and that shift materializes only where streets are safe enough for ordinary riders. Many U.S. cities nevertheless expand cycling networks administratively, designating existing arterials rather than redesigning them, although whether designation carries safety information has not been tested at a network scale. Integrating a statewide crash file with a linear-referenced roadway inventory, this study analyzes 6930 unique police-reported pedalcycle crashes in Tennessee (2002–2026). Designated routes carry 17.4% of crashes and are faster (41.0 versus 37.4 mph), busier, and wider than other crash locations (all p < 0.001). In county-clustered logit models, designation was associated with higher fatal odds (OR = 1.697, 95% CI 1.15–2.50) but not significantly with fatal-or-serious odds or, in a zero-truncated negative binomial model, with crash frequency, and exposure-normalized rates are lower. The result is therefore a conditional-severity association consistent with the arterial character of designated corridors, not evidence that designated routes are generally more dangerous to ride. Designated routes hold 15.6% of analyzed crash-bearing mileage but 24.2% of fatal crashes and 20.5% of the USD 5.99 billion societal cost, and 92.3% of their mileage is high-stress. The interpretable logit (cross-validated AUC = 0.766) matches gradient boosting and random forests, and an empirical Bayes corridor index validated on 2019–2025 crashes captures 37.6% of later fatal and serious crashes on screened crash-bearing segments (about 22% statewide) within 10% of their mileage, comparable to crash density ranking; the results support pairing designation with speed management or separation, which the index can help prioritize within an urban safety digital twin.