With a permitted left-turn phase, left-turning vehicles irregularly change speed and trajectory to compete for the right of way with straight vehicles approaching from the opposite direction, which brings serious safety hazards to intersection operation. The data collected from three intersections in the high-tech development zone of Xi'an city, and the turning behaviors of left-turning vehicles were classified into three patterns: "no-interference, yield, and rush" based on the interactions between the left-turning vehicles and the straight vehicles approaching from the opposite direction. This study took post encroachment time as the quantitative indicators of the conflict severity, which was classified into 4 levels and the disordered logistic regression model was established to identify the significant factors affecting the conflict severity. The results showed that: the conflict severity of the left-turning vehicles in rush pattern is higher than that in yield pattern. The odds ratio of that the conflict severities were medium, serious conflict respectively increased to 11.598 and 3.348 when the vehicles in rush pattern. The arrival speed of the left-turning vehicles and signal phase were significant factors increasing the severities of conflicts, and the average speed difference and average headway are protective factors. The results of this study can provide a reference for traffic management departments to optimize traffic control schemes.