Efficient transport infrastructures are crucial to improve traffic safety and reduce negative effects of human activities, where effective transport policies play a vital role. It is also important to evaluate the effects of the policies to document and update them with time. The purpose of this study is to examine road traffic crashes with fatalities and severe injuries at the county level in Sweden, considering both planned innovative policies and the unexpected influence of the COVID-19 pandemic. Data on road traffic crashes with fatalities and severe injuries from 2010 to 2021 in twenty-one counties in Sweden were collected, as well as road network data, vehicle registration data, land use data and socio-demographic data. Negative binomial autoregressive models with region fixed effect were proposed to estimate the effects of explanatory variables on crash frequency, along with Generalized Method of Moments (GMM) models. The findings demonstrate that the speed limit changes and the introduction of safety cameras in Sweden is effective in reducing fatalities and severe injuries. Importantly, different regions exhibit distinct responses to these interventions, with varying effects on different types of crashes. Additionally, findings show that other factors, such as land use characteristics, presence of trucks, and population density also influence the crash frequencies. Moreover, the effect of the COVID-19 pandemic was investigated, which shows a reduction in traffic crashes during the pandemic period. Important policy implications based on the findings are discussed to enhance transport safety and resilience. By considering regional variations, this study provides insights into the spatial dynamics of road safety patterns, and supports the Vision Zero policy in Sweden.