Pedestrian Distraction in the District of Columbia Intersections: Behavior and Prevalence Analysis

Abstract: The increasing incidence of pedestrian injuries and fatalities at intersections, often linked to distracted walking, calls for innovative safety measures. This study examines pedestrian behavior at signalized intersections in the District of Columbia, with a focus on distraction prevalence and its impact. Our research employs a mixed-method approach, combining survey data analysis with ongoing video observations at 15 intersections during peak hours (8-9 AM, 12-1 PM, 5-6 PM). Initial survey data from 390 pedestrians showed an uneven distribution among distraction levels: single distraction (40.8%), not distracted (31.0%), and multiple distractions (28.2%). To improve model reliability, we balanced the dataset to equal representation (33.3%) across all categories. Using multinomial logistic regression analysis on the balanced data, we identified significant predictors for each distraction level. For multiple distractions, electronic device usage (OR=2.79, p< 0.001), noise cancellation technology use (OR=2.50, p< 0.001), and respondent condition (OR=2.63, p< 0.001) emerged as strong predictors, with age showing a protective effect (OR=0.87, p< 0.001). Among non-distracted pedestrians, respondent condition (OR=0.18, p< 0.001) showed the strongest negative association, while age (OR=1.22, p< 0.001) emerged as a positive predictor. Single distraction was primarily predicted by respondent condition (OR=2.74, p< 0.001) and time of day (OR=1.45, p< 0.05). The balanced model achieved improved predictive accuracy (60%), with notable performance in identifying non-distracted individuals (F1-score=0.75). While video analysis of pedestrian behavior at intersections continues, these findings highlight the complex interplay between personal, technological, and environmental factors in pedestrian distraction patterns. Our ongoing research aims to develop comprehensive strategies to enhance pedestrian safety in urban environments.