The fulfillment of wide-ranging requirements in regard to safety and comfort of drivers is a challenging task for many automotive manufactures. In order to ensure the reliable and efficient testing of advanced driver assistance systems (ADAS), a data-driven catalog consisting of relevant traffic scenarios is of major importance. In this context, this paper presents a two-layer method for the mining of critical scenarios from accident data containing information on various categories. Firstly, the Extended Successive Odds Ratio Analysis is conducted, in which all the extracted combinations of risk-inducing attributes can be taken into account as the foundation of scenario description. Afterwards, the Successive Association Rules Analysis completes the scenario description using further attributes whereby an optimization problem is formulated to find the best associated attribute successively. The relevant traffic scenarios extracted from real-world data by applying this novel approach give the developers an opportunity to test the functionality of ADAS.