This study investigates the relevance of certain parameters for virtual testing of the driver's side OOP problem and attempts to answer the following questions: Which level of detail is needed in the airbag models to assess occupants' injury values for OOP scenarios? What is the influence of the airbag cover on the occupant response for OOP situations and how to accurately model the airbag cover? Are current dummy models suitable to assess the localized human injuries and/or is it beneficial to include human models for injury assessment? The combined multibody-FE code MADYMO was used. A virtual set-up of a folded driver airbag including cover, steering wheel and a simple seat forms the base model. Occupants were positioned according to the two OOP driver positions defined in FMVSS 208, i.e. chin on module to maximize neck injury risk and chest on module to maximize chest injury risk. The 5^th percentile Hybrid III. dummy model as well as the 5^th percentile human model were used and different airbag inflation and cover modeling techniques were investigated. Attention was focussed on analyses of the injury values obtained from the simulations of the different situations (with / without cover, use of gasflow, human / dummy model). From this study, it is concluded that it is important to take into account the geometrical effects of the airbag cover on airbag deployment and hence injury generation in OOP simulations. Adding a gasflow description to the airbag deployment simulation resulted in differences in the early deployment phase. If the dummy's injuries are determined in this phase (dependent of scenario), gasflow contributes to improved airbag modeling at the cost of higher CPU times. From the human model simulations, it is concluded that for this specific scenario, using dummy models could lead to underestimation of human injuries in the chest region, in contrast to overestimation of the head acceleration and the neck loads in the rebound phase. Therefore, human models could provide additional value compared to dummy models.