For EHDS implementers, NFDI4Health architects, GDI infrastructure teams, and anyone building distributed health data infrastructure that must route intelligence across jurisdictional boundaries without centralizing governance.
The Broker Assumption
Every major European health data infrastructure project shares an unexamined architectural assumption: intelligence synthesis requires a central broker.
- European Health Data Space (EHDS): Secure Processing Environments (SPEs) at national level, with a central access point for cross-border secondary use queries
- NFDI4Health: A central metadata repository that indexes distributed health data across German research institutions
- Genomic Data Infrastructure (GDI): A federated analysis platform with central coordination for cross-border genomic queries under the 1+ Million Genomes initiative
- OHDSI Europe: A coordinating center model where studies are designed centrally and distributed to DataPartners
In each case, the architecture works the same way: queries flow from the center to the edges, results flow from the edges to the center. The center synthesizes. The edges wait for the next query.
This is hub-and-spoke. It works. It has known limitations:
Single point of governance: Every cross-border query must be approved by the central body. For EHDS, this means the national SPE approval chain. For NFDI4Health, this means the consortium governance structure. The bottleneck is governance, not technology.
Linear intelligence scaling: The hub receives N contributions and produces one synthesis. The N(N-1)/2 pairwise synthesis opportunities between edge nodes are invisible to the hub.
Latency: Each query cycle — design, approve, distribute, execute, return, synthesize, publish — takes weeks to months. Between cycles, the network does not learn.
Sovereignty tension: Every cross-border synthesis requires data (even aggregate data) to flow to a jurisdiction-external processing point. This creates ongoing tension with national data sovereignty requirements.
What if the broker were not required?
Routing Without the Broker
Christopher Thomas Trevethan discovered the Quadratic Intelligence Swarm (QIS) protocol on June 16, 2025. The protocol routes validated outcomes between distributed nodes using semantic addressing — with no central broker, no central aggregator, and no central governance approval per routing event.
Here is the architectural difference:
Hub-and-Spoke (Current European Model)
Node DE ──→ Central Broker ←── Node FR
Node NL ──→ Central Broker ←── Node IE
Node IT ──→ Central Broker ←── Node ES
↓
Synthesis
↓
Published result
- The broker receives all results
- The broker performs synthesis
- Intelligence grows as O(N) — one synthesis per query cycle
- Every routing event requires broker governance approval
QIS Peer-to-Peer Outcome Routing
Node DE ←→ Node FR (semantic address match)
Node NL ←→ Node IE (semantic address match)
Node DE ←→ Node NL (semantic address match)
Node FR ←→ Node ES (semantic address match)
...
N(N-1)/2 active synthesis paths
- No broker receives aggregate data
- Each node synthesizes locally, on its own infrastructure, under its own jurisdiction
- Intelligence grows as Θ(N²) — every pair of nodes with matching clinical problems learns from each other
- Routing is determined by semantic address, not by governance approval per event
The governance shift: instead of approving each cross-border query, the governance framework approves the protocol — the rules governing what constitutes a valid outcome packet, what can be included, and what the semantic address space looks like. Once the protocol is approved, routing is automatic. This is how internet routing works: the protocol (BGP, DNS) is governed; individual packet routing is not.
How It Works for EHDS Secondary Use
EHDS Regulation 2025/327, Articles 34-50, establishes the framework for secondary use of electronic health data across EU member states. The implementation challenge is architectural: how do you route health intelligence across national boundaries while keeping personal data inside national Secure Processing Environments?
The Current EHDS Model
- A researcher applies for data access through a Health Data Access Body (HDAB)
- The HDAB approves the request and grants access to the national SPE
- The researcher queries data inside the SPE — results stay within the SPE
- For cross-border queries, the researcher applies to multiple HDABs, runs the query in each national SPE, and synthesizes results externally
This works for episodic research studies. It does not enable continuous intelligence routing between national health systems.
The QIS-Enhanced EHDS Model
- Each national SPE runs validated analyses on its health data (no change from current model)
- After a validated analysis completes, the SPE distills the outcome into a ~512-byte packet containing only derived statistics — no personal data, no pseudonymized identifiers, no record-level information
- The packet is fingerprinted using standardized terminology (SNOMED CT, ICD-10, ATC) as the semantic address
- The packet is deposited at its semantic address — accessible to any national SPE managing patients with the same clinical profile
- Other national SPEs query the semantic address space and pull back outcome packets from peer jurisdictions
- Each SPE synthesizes incoming packets locally, inside its own national infrastructure
What crosses the border: ~512-byte outcome packets containing population-level derived statistics. Not personal data. Not pseudonymized data. Not data that falls within EHDS personal data governance scope.
What stays inside each jurisdiction: all patient-level data, all computation, all synthesis. The SPE remains sovereign. The intelligence still compounds across borders.
How It Works for NFDI4Health
NFDI4Health is building a national research data infrastructure for personal health data in Germany. The core challenge: German health data sits in distributed institutional repositories — university hospitals, public health institutes, epidemiological study centers — each with its own governance, ethics board, and technical infrastructure.
The current NFDI4Health approach centralizes metadata (what data exists where) while keeping data distributed (actual records stay at institutions). Researchers discover datasets through the central metadata catalogue and negotiate access institution by institution.
QIS adds the missing layer: continuous outcome intelligence routing between NFDI4Health institutions.
- Charité Berlin runs a treatment outcome analysis on its local cohort. Distills the validated result into an outcome packet. Deposits at the semantic address (ICD-10 + ATC + outcome type).
- University Hospital Heidelberg manages patients with the same condition on the same treatment class. Queries the semantic address. Pulls back Charité's outcome packet.
- Heidelberg synthesizes locally. Its next analysis starts with Charité's intelligence, not from zero.
- Heidelberg's own validated outcome deposits at the same address. Charité queries and synthesizes.
No central NFDI4Health system processed the synthesis. No cross-institutional data sharing agreement was needed per routing event (the protocol approval covers the framework). Each institution's data stayed in its own repository. The intelligence compounded across institutions automatically.
How It Works for GDI Genomic Data
The Genomic Data Infrastructure consortium is building federated genomic data access across 23 European countries under the 1+ Million Genomes initiative. The challenge: polygenic risk scores (PRS), pharmacogenomic variant associations, and rare variant population frequencies require synthesis across national genomic datasets — but genotype sequences cannot cross borders.
QIS applies to genomic outcome routing:
- A GDI node computes a PRS for a specific variant set in its national cohort
- The validated PRS delta is distilled into an outcome packet: variant reference IDs (rsIDs, not sequences) + ancestry population tag + disease association + effect size + confidence interval
- Fingerprinted on variant set + ancestry + disease → deterministic semantic address
- Deposited and routable to every GDI node computing PRS for the same variant-disease association
What routes: statistical associations between variant sets and disease outcomes. Effect sizes. Population frequencies. All derived — no genotype sequences.
What stays local: every individual genotype, every sequence file, every phenotype linkage. The genomic data never leaves the national node.
The synthesis across 23 countries builds a continental PRS model without any country sharing any individual's genome. And because QIS imposes no minimum cohort size, even countries with small genomic cohorts (rare ancestry populations, rare variant carriers) contribute equally. The most informationally valuable observations — the rare ones — are included, not excluded.
The Compute Economics
The reason European health data projects default to hub-and-spoke is not that brokerless routing is theoretically impossible. It is that full-mesh routing (every node talks to every node) requires O(N²) compute, which is impractical at scale.
QIS breaks this assumption:
| Architecture | Intelligence Output | Compute Cost | Intelligence/Compute Ratio |
|---|---|---|---|
| Hub-and-spoke (broker model) | O(N) | O(N) | 1.0 |
| Full mesh (naive peer-to-peer) | O(N²) | O(N²) | 1.0 |
| QIS outcome routing | Θ(N²) | O(N log N) | O(N / log N) — grows with network size |
For a 27-member-state EHDS network:
- Hub-and-spoke: 27 intelligence units at 27 compute units = ratio 1.0
- QIS: 351 intelligence units (27×26/2 pairwise paths) at ~89 compute units (27 × log₂27 ≈ 27 × 4.75) = ratio 3.9
The ratio improves as the network grows. At 100 nodes: 4,950 intelligence units at ~664 compute units = ratio 7.5. The broker model stays at 1.0 regardless of scale.
What This Means for European Health Data Strategy
The European health data infrastructure stack is being built right now. EHDS is in implementation phase. NFDI4Health is building its metadata layer. GDI is deploying federated access nodes. OHDSI Europe is expanding its DataPartner network.
The routing layer — the protocol that determines how validated intelligence moves between nodes — has not been chosen. The default is hub-and-spoke because that is what existing infrastructure provides.
QIS offers the alternative: peer-to-peer outcome routing with no central broker, quadratic intelligence scaling, logarithmic compute cost, and EHDS SPE-native privacy by architecture. The protocol is transport-agnostic — it runs on whatever infrastructure each national node already has. The semantic address space uses the clinical terminologies (SNOMED CT, ICD-10, ATC, LOINC) that European health data standards already mandate.
The infrastructure exists. The standardization is done. The protocol that activates the N(N-1)/2 synthesis paths between existing nodes is discovered.
The Discovery
Christopher Thomas Trevethan discovered the Quadratic Intelligence Swarm protocol on June 16, 2025. The breakthrough is the complete architecture — the loop that enables real-time quadratic intelligence scaling without compute explosion, not any single component. 39 provisional patents filed. Humanitarian licensing ensures the protocol is free forever for nonprofits, research institutions, and educational use.
For European health data architects: the QIS protocol specification, OHDSI synthesis layer reference, and the 20 most common technical questions are published.
This is part of an ongoing series on QIS — the Quadratic Intelligence Swarm protocol — documenting every domain where distributed outcome routing closes a synthesis gap that existing infrastructure cannot close.
Top comments (0)