HIPAA-Compliant Patient Record Anonymization Pipeline

Design a robust Python script using pandas and cryptography that automatically anonymizes patient medical records while preserving statistical integrity. The solution must handle PHI (Personal Health Information) by replacing identifiable fields with secure hash tokens, maintain referential consistency across multiple datasets, and generate an audit log of all transformations. Implement encryption methods that allow reversible anonymization for authorized medical research purposes, with strict access controls and compliance with HIPAA 45 CFR § 164.514 standards.