With advances in cloud computing, hospitals and healthcare organizations can outsource the storage and management of their encrypted electronic medical records (EMRs) to the cloud services for great flexibility and economic savings. Although encryption helps protecting user data confidentiality, designing secure and practically efficient search functions over encrypted data remains challenging problem. This paper first constructs a multi-field conjunctive keyword search (MCKS) scheme, called MCKS_I, which supports equality query. For more flexible and complex MCKS, such as subset and range query, it then proposes an improved scheme, MCKS_II, based on a novel vector representation of hierarchical attributes. The new schemes are proven to be able to resist known plaintext attack. Extensive analysis and experiments show that the proposed schemes are extremely practical.