Determining pedestrian accident hotspots on road segments is a crucial part of the pedestrian safety assessment as it is used to prioritize problematic parts of a road network for in particularly planning and implementation strategies. Moreover, the spatial pattern of the pedestrian accidents may change over time due to several factors. In order to better understand pedestrian safety conditions, pedestrian accident patterns have to be analysed with regard to both space and time. This paper adapts such a spatio-temporal hotspot detection method for the analysis of pedestrian accidents. In this study, 189 traffic accidents involving pedestrians that resulted in injury or fatality on the Eskisehir Motorway (Turkey) between the years of 2005 and 2010 are mapped with their spatial and temporal information. Network-based Kernel Density Estimation is used to examine the hotspots of pedestrian accidents and their changes over the years. Then, the significances of the results are evaluated by using Network-based the Nearest Neighbor Distance and the K-function methods. The impact of land use change and taken measures are evaluated based on spatio-temporal hotspot analysis.