The aim of clinical decision support implementing electronic health records is to satisfy the physicians' information needs. We are motivated to propose an attention based network on query expansion. Considering the difficulty and cost of medical text annotation and inspired by the idea of migration learning, we chose the non-medical dataset for model training, and migrated to medical datasets. The model utilizes LSTM to obtain sentence representation and adopt attention mechanism to obtain entities representation. The proposed approach can dynamically select related entities as expansion of the query. At the same time, we not only consider the score of a single term as an expansion term, but also consider the score of term combination. We conduct the experiments on the three standard datasets of TREC Clinical Decision Support Track, where the approach has a promising overall performance over the strong baseline.