Changes in pedestrian and bicyclist crash severity before and after COVID 19 using machine learning models

Pedestrians and bicyclists are among the most vulnerable road users and account for a substantial share of traffic injuries and fatalities. The COVID 19 pandemic altered mobility patterns and driving behavior, potentially affecting crash severity. This study examines pedestrian and bicyclist crash severity in Florida by comparing 2017–2019 with 2021–2023, excluding 2020 as a transition year. Classification models included LDA, QDA, Logistic Regression, Random Forest, and XGBoost, with severe injury specified as the positive class. Model performance was comparable and modest relative to the majority class baseline, with severe injury recall ranging from 18.6 to 31.0%. A likelihood ratio test rejected parameter stability between periods (χ² = 44.07, df = 21, p = 0.002). However, differences between pedestrian and bicyclist models were substantially larger (χ² = 978.54, df = 21, p < 0.001), indicating that road user mode was a stronger source of variation than study period. Permutation importance identified posted speed limit, time of day, and non-motorist age as important severity predictors, while sex and weather contributed relatively little. Severe injury share increased above pre-pandemic levels in 2021 before declining to 21.5% by 2023. Mode specific analysis showed similar effect directions but different magnitudes. Posted speeds of 40 MPH and above were associated with severity increases of approximately 28% points for pedestrians and 13% points for bicyclists. The findings support context specific countermeasures, with speed and nighttime treatments emphasized for pedestrians and roadway treatments differentiated by road user mode where supported by evidence.