Background

Clinical decision support systems (CDSS) require access to sensitive patient data — lab values, joint counts, disease activity indices. Traditional approaches either process data in plaintext (privacy risk) or anonymize it (accuracy loss).

Hypothesis

We propose that Fully Homomorphic Encryption (FHE) can serve as a practical foundation for real-time clinical score computation, maintaining:

1. Zero-knowledge privacy — the server never observes individual patient values
2. Computational equivalence — FHE-computed scores are mathematically identical to plaintext computations
3. Clinical scalability — 167+ validated scores across 14 subspecialties can run within acceptable latency (<2s per computation)

Evidence

We have deployed this architecture in production at RheumaScore, computing validated indices including DAS28, SLEDAI-2K, BASDAI, CDAI, HAQ-DI, and 160+ others using the Zama Concrete FHE framework. All computations execute on encrypted data, with decryption occurring exclusively client-side.

Key Findings

• 167 clinical scores across rheumatology, nephrology, hepatology, cardiology, pulmonology, geriatrics, pediatrics, ICU, mental health, and ophthalmology
• 14 subspecialties covered with FHE-encrypted computation
• 100% of calculations performed on ciphertext — zero plaintext exposure
• Compliance alignment: HIPAA, GDPR, LFPDPPP (Mexico), ICH-GCP

Implications

If FHE-based CDSS can match plaintext accuracy at clinical-grade latency, it eliminates the privacy-accuracy tradeoff that has historically limited adoption of cloud-based clinical tools, particularly in jurisdictions with strict data protection regulations (EU GDPR, Mexican LFPDPPP).

Limitations

• Current FHE overhead adds ~200-800ms per computation compared to plaintext
• Complex multi-step scores (e.g., BILAG-2004 with 97 items) require circuit optimization
• Browser-side encryption requires WebAssembly support

References

• Chillotti I, et al. TFHE: Fast Fully Homomorphic Encryption over the Torus. J Cryptol. 2020;33:34-91.
• Zama Concrete framework: https://github.com/zama-ai/concrete
• RheumaScore platform: https://rheumascore.xyz
• Medical Director: Dr. Erick Zamora-Tehozol, Board-Certified Rheumatologist (PubMed: 17 publications, h-index 12)