Vulnerable road users (VRUs)-related crashes are recognized as an important public safety problem. However, few macro-level studies of VRUs-involved crashes have considered the long-term spatial, temporal, or spatiotemporal effects in the crash risk. This study analyzes the bicycle and pedestrian crash risk in different injury severities by using three multivariate Bayesian space–time models. These models address different spatiotemporal effects to account for possible correlations across injury severities over space and time. Various explanatory variables are used to examine the contributory risk factors, including socio-demographic features, roadway structures, and weather characteristics. Spatio-temporal conditional autoregression with an ANOVA style (ST-CARanova) models outperform other two space–time models in most circumstances. The long-term spatiotemporal effects, such as relatively high temporal autocorrelations, significant spatial heterogeneity, and weak spatiotemporal interactions, are found in this study. The increase of female ratios, young people ratios, unemployment rates, and annual average high temperatures could increase the county-level crash risk of cyclists and pedestrians. The findings provide useful insights for policy makers to improve the safety of cyclists and pedestrians.

与弱势道路使用者 (VRU) 相关的碰撞被认为是一个重要的公共安全问题。然而，很少有涉及 VRU 的碰撞的宏观研究考虑了碰撞风险中的长期空间、时间或时空影响。本研究使用三个多元贝叶斯时空模型分析了不同伤害严重程度下的自行车和行人碰撞风险。这些模型解决了不同的时空效应，以解释不同空间和时间的损伤严重程度之间可能存在的相关性。各种解释变量用于检查促成风险的因素，包括社会人口特征、道路结构和天气特征。在大多数情况下，具有方差分析风格 (ST-CARanova) 模型的时空条件自回归模型优于其他两个时空模型。本研究发现了长期的时空效应，例如相对较高的时间自相关、显着的空间异质性和弱时空相互作用。女性比率、年轻人比率、失业率和年平均高温的增加可能会增加县级骑自行车者和行人的撞车风险。研究结果为政策制定者提高骑自行车者和行人的安全提供了有用的见解。