This research focuses on the development of a crash risk index that quantifies the crash risk coming from impending rear-end or sideswipe crashes. It proposes a generalized framework that can be adopted for traffic safety improvement on freeways to accomplish this. The proposed generalized framework begins with the collection of potential contributing factors that affect crash risk. It involves traffic parameters, such as basic traffic parameter that can be obtained from detectors, lane-based traffic parameters, number of trucks, ramp flow, and surrogate measures of lane-changing, and environmental characteristics, such as rainfall, snow, visibility, pavement, and lighting condition. The most significant contributing factors are identified by random forest method and three statistical tests for five different segment types. Based on the identified factors Bayesian random intercept logistic regression is used to build crash risk prediction models for five segment types. The outcome of them is used to develop a crash risk index that quantifies crash risk for the segment. The developed crash risk index is applied to a 13.14-mile stretch of I-110 northbound in California to monitor the change in crash risk in real time. The results of monitoring demonstrate that the crash risk indices result in high crash risk before the crash occurs. Lastly, new variable speed limits (VSL), which is a proactive intervention, aiming to reduce the high crash risk indicated by the crash risk indices are proposed. The results indicate that the proposed VSL control reduces the high crash risk below thresholds and achieves a reduction in travel time as well. It is expected that if this whole framework would be utilized on a freeway, it will help traffic operators to identify hazardous traffic conditions better and to implement proactive interventions to alleviate crash risk in the right location at the right time.