It is important to determine the genetic types of high risk of birth defects and to promote early heredopathia screening and birth guidance for first class prevention of birth defect like congenital hearing loss. This research applies the decision tree algorithm to analyze the clinical data of nearly 1 000 cases of GJB2 genetic mutation record by using data mining tools. The model of assistant screening for pathogenic genes of hearing loss birth defects is established. By studying the structure of the model tree and samples’ classification results, it is concluded that pure positive samples of hearing loss are obtained in five groups of branches in the model tree. Besides, the set of genetic loci states formed by each of these branches is consistent with the mutation state of pathogenic genes confirmed by clinical studies. The screening model established by the decision tree method can assist doctors to quickly screen out the types of pathogenic genes from the clinical big data.