Secure Cohort Discovery

The protocol distributes a single feasibility query across multiple hospital sites simultaneously. Each site executes the query locally on its own patient database, producing an encrypted or secret-shared partial result. No raw patient data leaves the hospital's firewall. The computation uses secure multi-party computation (MPC) to aggregate results across sites, ensuring that even the aggregator learns only the final cohort count, not individual site contributions. This architecture supports complex inclusion/exclusion criteria including ICD-10 codes, lab values, medication histories, and genomic markers.

Before counting cohorts, sites must identify patients present in multiple databases to avoid double-counting. The protocol employs Private Set Intersection (PSI) to match patient identifiers across institutions without revealing the identifiers themselves. Each site inputs its set of hashed patient IDs, and the protocol outputs only the intersection size or deduplicated union. Bloom filters and oblivious pseudorandom functions accelerate this matching while maintaining cryptographic privacy. This step is critical for multi-site trials where patients may receive care at multiple facilities.

To prevent inferential attacks where a query result of 0 or 1 could reveal information about a specific individual, the protocol enforces a minimum cohort size threshold. If the aggregate count falls below a configurable threshold (typically 5-10 patients), the system returns a masked response such as '< 10' rather than the exact number. This suppression mechanism is combined with differential privacy noise addition to provide formal privacy guarantees. Researchers can iteratively refine their criteria until they achieve a statistically meaningful cohort size.

Every cohort discovery query is recorded in an immutable, cryptographically signed audit log. This log captures the query parameters, the requesting researcher's credentials, the timestamp, and the aggregate result without storing any patient-level data. Zero-knowledge proofs can be attached to each query execution to verify that the protocol was followed correctly without revealing the underlying data. This audit trail satisfies HIPAA accounting of disclosures requirements and provides institutional review boards with transparency into feasibility assessment activities.

Researchers interact with the system through a federated dashboard that provides real-time feedback on cohort sizes as they adjust inclusion and exclusion criteria. The dashboard displays aggregate statistics such as demographic breakdowns, comorbidity distributions, and medication usage patterns across the eligible cohort. All visualizations are generated from privacy-preserving aggregates, never from individual records. The interface supports drag-and-drop criteria building with auto-suggestions for standard terminologies like SNOMED CT and LOINC, accelerating protocol design.

Secure cohort discovery protocols are designed to operate across jurisdictions with conflicting data protection regulations. The cryptographic architecture ensures that data never crosses borders—only encrypted intermediate values are exchanged. Attribute-based access control enforces site-specific policies, allowing each institution to define which data elements are queryable and by whom. The system supports GDPR, HIPAA, and PIPEDA compliance simultaneously, with configurable data residency constraints that respect each country's sovereignty requirements while enabling global clinical research.

A cryptographic protocol that allows two or more parties to compute the intersection of their private datasets without revealing any elements outside the intersection. In cohort discovery, PSI enables a researcher to find patients matching trial criteria across multiple hospitals without either party exposing their full patient roster.

• Key property: Only the intersection is revealed; non-matching records remain cryptographically hidden
• Common techniques: Diffie-Hellman-based PSI, oblivious transfer-based PSI, and fully homomorphic PSI
• Real-world use: Finding overlapping patients between a pharmaceutical company's trial criteria and a hospital's EHR system

A cryptographic method for distributing a secret value among multiple parties, where each party receives a mathematically meaningless share. Only when a qualified subset of shares is combined can the original secret be reconstructed. This is the foundational primitive that enables secure multi-party computation.

• Additive secret sharing: A value x is split into random shares x₁, x₂ such that x = x₁ + x₂ mod p
• Threshold schemes: Require t-out-of-n shares for reconstruction, providing fault tolerance
• Role in cohort discovery: Patient counts and query predicates are secret-shared across computing nodes so no single node sees the full query or result

A class of protocols that allows a central server to compute the sum of values from multiple clients without inspecting any individual contribution. In the context of cohort discovery, secure aggregation enables a researcher to learn the total patient count across sites without seeing per-site counts.

• Masking technique: Clients add pairwise random masks that cancel out in the aggregate sum
• Dropout robustness: Protocols handle clients that disconnect mid-computation without compromising privacy
• Threshold variant: The aggregate is only revealed if a minimum number of clients participate, preventing inference from small subsets

A fundamental cryptographic primitive where a sender transmits one of potentially many pieces of information to a receiver, but remains oblivious to which piece was selected. OT is the atomic building block for garbled circuits and general-purpose secure computation.

• 1-out-of-2 OT: The sender has two messages m₀, m₁; the receiver chooses bit b and learns only m_b
• OT Extension: Efficiently amplifies a small number of base OTs into millions using only symmetric-key operations
• Relevance: Underpins the secure comparison operations needed to evaluate patient inclusion/exclusion criteria across private databases

A cryptographic protocol introduced by Andrew Yao that enables two parties to jointly evaluate a boolean circuit over their private inputs without revealing those inputs to each other. One party garbles (encrypts) the circuit; the other evaluates it obliviously.

• Free-XOR optimization: XOR gates require no cryptographic operations, dramatically reducing cost
• Half-gates technique: Reduces communication to 2 ciphertexts per AND gate
• Application: Encoding clinical trial eligibility criteria as a boolean circuit that evaluates patient records without exposing the records or the exact criteria logic

The application of secure multi-party computation to statistically infer unobserved genetic variants in a study cohort by referencing a private haplotype reference panel. This enables researchers to enrich genomic datasets without exposing either the study genotypes or the reference panel.

• Hidden Markov Models: The imputation algorithm is implemented as an MPC circuit operating on secret-shared data
• Privacy guarantee: The reference panel owner learns nothing about the study cohort; the researcher learns only imputed variants
• Clinical relevance: Essential for identifying genetic markers in rare disease cohorts where pooling data across institutions is legally restricted