Risk analysis of crashes related to e-scoters: proposal of a predictive model

The rapid expansion of electric scooters has revolutionized urban mobility while raising significant safety concerns, particularly regarding crashes. This thesis presents a comprehensive study on the risk analysis of crashes involving electric scooters and proposes a predictive model to estimate the likelihood of crashes on urban road segments and intersections. Conducted in Bari, Italy, this research ad-dresses gaps in the literature by integrating data on electric scooter traffic, road ge-ometric characteristics, and infrastructure within a robust statistical framework. The study begins with an analysis of crash data provided by the Local Police of Bari for the period 2021–2023. Descriptive statistics reveal critical insights into crash dynamics: 70% of the events resulted in injuries, highlighting the vulnerability of scooter users. Crashes predominantly occurred on undivided roads, underscoring potential infrastructure-related risks. The analysis also considers temporal and spa-tial patterns, providing useful insights for identifying high-risk conditions. Data collection for this research followed an innovative approach, addressing the scarcity of standardized data on electric scooter traffic. Through collaboration with Vento Mobility Srl (TIER), access to usage statistics was secured, complement-ed by field surveys to estimate traffic volumes for both shared and private scooters. Additional information on road characteristics, vehicle traffic, and infrastructure was meticulously gathered using tools such as QGIS and Google Maps, ensuring a com-prehensive dataset. A binomial logistic regression model was developed to predict the likelihood of crashes based on key variables such as daily traffic volumes, road geometry, and the presence of specific infrastructure like bicycle crossings or turning lanes. Models were separately developed for road segments and intersections, ensuring greater contextual relevance. Variables were selected based on statistical significance to en-hance the model accuracy while maintaining interpretability. The predictive model results highlight key factors influencing the risk of crashes involving electric scooters. For intersections, high volumes of vehicular and scooter traffic significantly increase crash likelihood. Similarly, for road segments, variables such as segment length and traffic density emerge as critical predictors. These findings underscore the importance of targeted infrastructure improvements and traffic management strategies to reduce risks. This thesis provide a validated predictive model that can support policymak-ers, urban planners, and transportation authorities. The study offers a practical tool to identify high-risk areas and implement safety interventions, promoting sustainable and safe micromobility in urban contexts. By addressing both data collection challenges and the methodological limita-tions of previous studies, this research lays the groundwork for future studies on electric scooter safety. It advocates for the standardization of data collection and the adaptation of infrastructure to support the safe integration of scooters into urban transportation systems, even if changes in regulations should be taken into account in future studies.