Rollover accidents are one of the leading causes of death in highway accidents due to their very high fatality rate. A key challenge in preventing rollover via chassis control is that the prediction of the onset of rollover can be quite difficult, especially in the presence of terrain features typical of roadway environments. These road features include superelevation of the road (e.g road bank), the median slope, and the shoulder down-slope. This work develops a vehicle rollover prediction algorithm that is based on a kinematic analysis of vehicle motion, a method that allows explicit inclusion of terrain effects. The solution approach utilizes the concept of zero-moment point (ZMP) that is typically applied to walking robot dynamics. This concept is introduced in terms of a lower-order model of vehicle roll dynamics to measure the vehicle rollover propensity, and the resulting ZMP prediction allows a direct measure of a vehicle rollover threat index. Simulation results using a complex multi-body vehicle simulation show the effectiveness of the proposed algorithm during different road geometry scenarios and driver excitations.