Homomorphic Encryption for Privacy-Preserving Healthcare Informatics: A Secure Weighted Risk Score Framework Using OpenFHE and Synthetic FHIR Data

This paper presents an end-to-end privacy-preserving healthcare analytics framework that combines Synthea-generated Fast Healthcare Interoperability Resources (FHIR) records with the Cheon-Kim-Kim-Song (CKKS) homomorphic encryption scheme implemented in OpenFHE. The framework extracts five normalized features—age, systolic blood pressure, body mass index, cholesterol, and a binary heart-condition indicator—and evaluates an illustrative weighted risk score while the patient data remain encrypted. The experimental dataset contains 15 synthetic patients distributed across three representative institutions and uses a CKKS batch size of 32 slots. For the reported representative evaluation, the plaintext score was 87.60 and the decrypted encrypted score was 87.58, corresponding to a 0.022% relative error and a 71.01 ms end-to-end runtime. The score is a demonstration model for evaluating encrypted computation and is not a clinically validated diagnostic index. The results show that a shallow linear healthcare computation can be evaluated with low numerical error while preserving confidentiality during processing. The study also identifies limitations related to circuit depth, incomplete hardware metadata, and the need for validation using clinically established models and larger datasets.