Amidst rapid advancements in precision and translational medicine, cross-institutional and cross-regional sharing of medical research data is increasingly critical. However, such data contains sensitive patient privacy and institutional intellectual property, subject to stringent security and ethical constraints, while cloud adoption exacerbates data leakage risks – making the balance between data utility and privacy protection a core challenge. As a key solution for medical data sharing, existing Attribute-Based Encryption (ABE) schemes suffer from limitations including unidirectional privacy protection, difficulty in tracing key leakage, and inefficient policy matching. To address these, we propose PSI-TR-ABE, a Private Set Intersection-based Traceable Policy-Hidden ABE scheme, which integrates Private Set Intersection(PSI) protocols with Paillier homomorphic encryption to achieve bidirectional policy-attribute privacy protection, employs an Linear Secret Sharing Scheme(LSSS)-based policy pre-verification mechanism to reduce futile decryption, and embeds hashed user identities into key structures for precise traceability. Theoretical proofs and comparative evaluations demonstrate enhanced efficiency without compromising security, providing a novel solution for secure and efficient medical data sharing.