Since the growing road network and increasing vehicles and speed leads to increased loss of life and financial losses consequence from road accidents. Improvement plans of the hazardous points are administrative priorities of any government. Any investment in this area in order to improve these unsafe locations is the most feasible plan to be considered. One of the very popular accident severity index models used in all countries is based on linear models to rehabilitate pavements and this paper is aiming at correcting the deficiency of PIARC’s related model i.e. lack of sensitivity to changes in the traffic volume flow, to modify crash severity index (which is based on linear models) making an allowance for the nonlinear effects of traffic on eventful locations on dual carriageways. To do so, traffic volume has been chosen as the hazard criteria and, using multiple regression and statistical models, the coefficients and variables of the new model have been calculated by means of the SPSS software.