Recent transportation bills have required each state to have a Highway Safety Improvement Program which emphasizes a data-driven approach to improving highway safety. This data-driven paradigm, coupled with the fact that approximately 40 percent of traffic crashes in the U.S. occur at intersections, has led to substantial research focusing on intersections. This study focuses on three areas of intersection safety: vehicular crashes, pedestrian and cyclist crashes, and crashes near ramp-terminal intersections. The impact of geometric characteristics on vehicular crashes at intersections using five years of crash data at an aggregate and disaggregate level. The within sample predictive ability of negative binomial models estimated using aggregate crash data (with empirical Bayes methodology) was compared to that of a disaggregate model estimated using a site-specific random effects negative binomial framework. Pedestrian and cyclist crashes are often difficult to model on a large scale as exposure information is typically not collected or maintained by road agencies. To this end, the characteristics affecting pedestrian and cyclist crashes at intersections have been examined using census tract-level commuter information from the American Community Survey in lieu of observed pedestrian and cyclist volume. Finally, ramp terminal intersections provide important points of connection between restricted access roadways (such as interstates) and adjacent land. The safety performance along the corridors adjacent to the ramp terminal intersections is directly related to the proximity between the ramp terminal and access points such as driveways and intersections. This study explores the effect of ramp terminal and access point proximity on corridor safety and provides a framework for road agencies to evaluate corridor-level safety implications based on the proximity between ramp terminals and access points, the volume of the crossroad, and the volume of the access point.