Counting crashes is not enough: developing an exposure-based high risk network for e-scooter safety

Shared e-scooters are now common in urban transportation, but safety planning remains limited by a lack of exposure-based, spatially detailed crash risk measures. This study applied an exposure-based approach to identify high-risk e-scooter corridors in Nashville, Tennessee, combining 92 police-reported, motor-vehicle-involved e-scooter crashes with trip-kilometer exposure data over a 45-month period. Crashes were assigned to corridors by rider direction of travel, and rates were calculated per 100,000 vehicle kilometers traveled (VKT) citywide and by corridor. Across more than five million e-scooter kilometers of exposure, the citywide crash rate was 1.84 per 100,000 VKT, and 82.6% of crashes occurred at intersections or driveways. The 25 corridors accounted for 75.0% of crashes and 70.9% of VKT, with a combined rate (1.94 per 100,000 VKT) statistically indistinguishable from the rest of Nashville (1.58; p = 0.34). Corridor-level rates were nonetheless heterogeneous (chi-square = 96.17, df = 24, p < 0.001), with unadjusted rates ranging from 0.69 to 24.49 crashes per 100,000 VKT; empirical Bayes shrinkage reduced this spread to roughly 3.7-fold. Three corridors, Shelby Avenue, 21st Avenue, and West End Avenue, exceeded the citywide rate individually, with the first two remaining significant after correction for multiple comparisons. Exposure-adjusted corridor rates gave a different picture of risk than crash counts alone: citywide averages concealed substantial street-level variation, supporting corridor-level exposure measures as a screening tool for micromobility safety, though individual designations remain exploratory rather than confirmatory. These results let agencies combine trip and crash data to target safety investment at high-risk corridors.