Assessment of the influence of road geometric design parameters on crash frequency and severity at Bwaise interchange

Abstract

Road geometric design parameters are fundamental determinants of road safety, governing vehicle operating speeds, driver workload, braking performance, and the margin available for collision avoidance. Deficiencies in horizontal curvature, longitudinal gradient, lane width, and intersection configuration systematically elevate both crash frequency and severity, yet localized quantitative analyses linking such deficiencies to crash occurrence at urban hotspots in Uganda remain scarce. This study assessed the influence of road geometric design parameters on crash frequency and severity at Bwaise interchange, Kawempe Division, Kampala, over the period 2020-2024. Twenty segments, ten circulatory arcs, and ten approach/exit arm segments produced a panel of 100 segment-year observations from 74 spatially confirmed crash events sourced from Kawempe Police Station and the KCCA Bloomberg Road Safety Database. Geometric parameters were extracted from Ministry of Works and Transport as-built drawings and Google Earth Pro profiles, with spatial allocation in ArcGIS Pro 3.4.2. Poisson, Negative Binomial, and Zero-Inflated Negative Binomial regression were employed for crash frequency, ordered logistic regression for severity, and Crash Modification Factors derived from model Incidence Rate Ratios using the AASHTO Highway Safety Manual framework. The study recorded a mean crash rate of 21.01 crashes per kilometer per year across 20 segments, with four primary hotspots identified. Damage-only crashes dominated at 65.92%, while 18.1% involved injury or fatality; all recorded fatalities occurred on approach arm segments with steep gradients, three of which exceeded the recommended maximum of 4%, the worst recording 7.01%. The Negative Binomial model was confirmed as the most appropriate crash frequency estimator. Each one percentage point increase in longitudinal gradient was associated with a 17.8 -19.5% increase in expected annual crash count. Horizontal curve radius was the strongest geometric predictor, with each one-meter increase reducing expected crashes by 2.9-5.0% (r = -0.675). Crash Modification Factor analysis indicated that gradient reduction from 7.01% to the recommended 4.0% was associated with a 39-41% crash reduction, the single highest-impact intervention at the junction. Ordered logistic regression confirmed gradient and curve radius as significant predictors of crash severity. These findings confirm that geometric design non-compliance is systematically associated with elevated crash frequency and severity at Bwaise interchange.