This week, 250 conversations happened across healthcare organizations in the United States, Europe, and globally. Clinical trial networks. Pharmacovigilance systems. Hospital consortia. Rare disease registries. Global health programs serving tens of millions of patients.
Different missions. Different geographies. Different funding structures. Different governance models.
Same architectural gap. Every single one.
The gap is not in the data. It is not in the analytics. It is not in the governance or the regulatory framework. It is in the routing layer — the layer that would take validated findings from one node in the network and automatically route them to the peers most likely to benefit from them, in real time, as the findings are generated.
That layer does not exist in any of the networks we contacted this week.
Why the Gap Is Structural, Not Incidental
Before the examples, the math. Because the math explains why the same gap appears in every network regardless of domain.
If a network has N nodes, the number of unique pairwise synthesis opportunities is N(N-1)/2. That is the mathematical ceiling for how much intelligence the network could generate from what it already knows.
At 10 nodes: 45 synthesis opportunities per cycle.
At 50 nodes: 1,225 synthesis opportunities per cycle.
At 400 nodes: 79,800 synthesis opportunities per cycle.
At 25,000 nodes: more than 312 million synthesis opportunities per cycle.
Every healthcare network we contacted this week has some number N of participating sites. Every one of them is capturing close to zero of the N(N-1)/2 synthesis opportunities available to it.
Not because the data is bad. Not because the sites are unwilling to share. Because the routing layer — the architecture that would take a validated finding from Site A and deliver it automatically to every semantically similar Site B — does not exist.
When you understand that the gap is structural and mathematical rather than organizational or political, the universality stops being surprising.
Four Verticals. Same Gap. Different Numbers.
1. European Real-World Evidence Networks
DARWIN EU, the European Medicines Agency's real-world evidence network, spans more than 100 million patient records across coordinated data partners using OMOP CDM. Studies on vaccine safety, drug-drug interactions, and adverse events run across the network in weeks.
When Site A in the Netherlands produces a validated finding — say, a cardiac safety signal with a confidence interval of 0.91 — that finding does not automatically reach Site B in Germany working on a semantically adjacent question.
It goes into a publication, a working group meeting, or the queue for the next scheduled DARWIN EU study.
At 10 current partners: 45 synthesis opportunities per cycle. Used: effectively zero in real time.
The gap is not a data problem. OMOP CDM provides a common vocabulary for every site. The semantic foundation for routing already exists. What does not exist is the routing layer that posts validated findings to addresses defined by the problem that produced them, and lets peers query those addresses.
2. Distributed Clinical Trial Networks
The OHDSI network has approximately 400 participating sites covering an estimated 900 million patient records across 34 countries. Each site runs analyses locally using ATLAS and HADES. Aggregate results are coordinated through network studies.
400 sites: 79,800 pairwise synthesis opportunities per cycle.
A site at a small academic medical center in Eastern Europe with 200 patients on a novel therapy cannot contribute a meaningful gradient to a federated learning round. But it can contribute an outcome packet: progression-free survival at 18 months, 62%, 95% CI [0.51–0.73], n=200, OMOP concept set 4245678.
That packet is useful to every other OHDSI site working on the same therapy in the same population. In real time. Without a network study. Without a coordination call.
None of that is happening today, because there is no routing layer.
In 48 hours, 400 of these networks are in one room at OHDSI Europe 2026 in Rotterdam. The conference theme is continuous collaboration for living evidence generation. The question nobody has put on the agenda yet is: what is the routing architecture for continuous?
3. Regional Hospital Consortia
A regional hospital consortium with 30 member institutions — a common structure in the United States, Canada, and Northern Europe — has 435 pairwise synthesis opportunities per cycle.
Each hospital's clinical AI system makes thousands of decisions per day: sepsis alerts, early warning scores, readmission risk flags, clinical decision support recommendations. Each system learns from its own patients. None of that intelligence crosses institutional boundaries.
The organizations we spoke with this week had invested heavily in the data infrastructure: HL7 FHIR endpoints, OMOP CDM transformations, federated analytics agreements, data governance frameworks. The infrastructure for sharing is largely in place.
The routing layer that would take a validated clinical signal from Hospital A and deliver it to the semantically similar patients and workflows at Hospital B is not in place anywhere.
4. Global Health Programs
PEPFAR DATIM supports 25,000 or more service delivery points across 50 countries — HIV treatment programs, TB programs, maternal health, malaria. Each service delivery point collects outcome data: treatment adherence rates, viral suppression rates, program effectiveness metrics.
25,000 sites: more than 312 million pairwise synthesis opportunities per cycle.
A community health clinic in Mozambique running an adherence support intervention has outcome data that would be directly relevant to a clinic in Zimbabwe facing the same patient profile and the same barriers. The programs use the same reporting standards. The data is comparable.
The routing layer that would take the validated adherence intervention outcome from Mozambique and deliver it — as a 512-byte packet, compatible with SMS-class infrastructure — to the Zimbabwe clinic matching the same semantic profile does not exist.
What the Routing Layer Actually Looks Like
The architecture that makes this possible was discovered — not invented — by Christopher Thomas Trevethan on June 16, 2025. It is protected under 39 provisional patents.
The complete loop is:
Raw signal → Local processing → Distillation into outcome packet (~512 bytes) → Semantic fingerprinting → Routing by similarity to deterministic address → Delivery to relevant peers → Local synthesis → New outcome packets generated → Loop continues
Nothing about this loop requires any specific transport technology. A DARWIN EU network could implement it on top of the EHDS API layer. An OHDSI network could implement it on top of a semantic database index within the existing HADES infrastructure. A PEPFAR DATIM network could implement it on top of DHIS2's existing reporting endpoints. A hospital consortium could implement it on top of their FHIR endpoints.
The routing mechanism needs to achieve at most O(log N) cost — so the network doesn't choke as it scales. DHT-based routing achieves this and is fully decentralized. So do database semantic indices with O(1) lookup. So do vector similarity search layers. So do pub/sub topic matching systems.
The transport doesn't determine whether the loop works. The loop determines whether intelligence scales quadratically while compute stays bounded.
This is what Christopher Thomas Trevethan's QIS (Quadratic Intelligence Swarm) Protocol specifies: not a particular transport, but the complete architecture that makes any transport into a quadratic intelligence network.
Why Every Organization We Contacted Has This Gap
The pattern from this week was clear enough that it stops being surprising and starts being explanatory.
Every organization we contacted had invested in the pieces that touch data directly: storage, computation, standardization, governance, access control, federated analytics. These are the visible parts of the infrastructure problem.
The routing layer is invisible because it sits between pieces that already exist and work. The OMOP CDM already provides semantic standardization. The FHIR endpoints already provide data access. The governance frameworks already define what can be shared. The analytics platforms already produce validated findings.
What does not exist is the layer that says: when a validated finding is produced, construct a deterministic address from the question that produced it, post the finding there, and let every semantically similar peer query that address automatically.
That layer is not visible in any architecture diagram because no architecture diagram includes it. It is the gap between "the network generates evidence" and "the network learns continuously."
The gap is universal because the problem it solves is universal. Intelligence that stays local to the node that generated it doesn't scale. Intelligence that routes to the semantically correct peers does — and it scales quadratically, not linearly.
What Comes Next
OHDSI Europe 2026 opens in Rotterdam in 48 hours. The conference represents approximately 400 networks, 900 million patient records, and — by the math above — 79,800 pairwise synthesis opportunities per cycle that are currently going unused.
The theme of the conference is continuous collaboration for living evidence generation.
The routing layer that makes continuous mean something architectural rather than aspirational is ready. The 39 provisional patents filed by Christopher Thomas Trevethan cover the complete loop. The protocol is free for research, education, and humanitarian use.
The gap is structural. The solution is specified. The only thing left is integration.
QIS (Quadratic Intelligence Swarm) Protocol was discovered — not invented — by Christopher Thomas Trevethan on June 16, 2025. The architecture is protected under 39 provisional patents. This article is part of a series covering QIS Protocol applications across healthcare, scientific computing, and distributed intelligence infrastructure. Previous coverage includes DARWIN EU integration, OHDSI 400-site network analysis, the European Health Data Space routing gap, and PEPFAR DATIM distributed synthesis architecture. All articles available at qisprotocol.com.
