Inside BEKplatform’s Ontology: The 24 Million-Term Engine Powering Better Patient Matches

Why Language Is the Hidden Barrier to Clinical Trial Enrollment

Clinical trial recruitment is hard—often harder than it should be. Sponsors miss enrollment targets, sites burn hours on chart review, and eligible patients frequently go unnoticed, buried in narrative notes or mislabeled fields.

BEKhealth set out to solve this using AI. But not the kind that simply scores patients or surfaces keyword hits. At the heart of BEKplatform lies something far more foundational: a proprietary medical ontology made up of more than 24 million terms, synonyms, and lexemes.

It may sound like a giant dictionary—but it’s actually the engine that enables BEKplatform to find more eligible patients, faster and with greater accuracy. Here’s how it works—and why it matters.

Why Patient Matching Fails Without the Right Language

Poor enrollment is one of the leading reasons why clinical trials fail—delaying treatments and driving up costs. According to Tufts CSDD, nearly 80% of trials don’t meet enrollment timelines.

And it’s not due to a lack of data. Health systems are full of it—but that data isn’t always usable.

Here’s why:

• Medical records mix structured fields (like ICD-10 codes) with unstructured text (like consult notes).

• Clinicians document in inconsistent ways: “heart attack,” “MI,” and “myocardial infarction” all refer to the same condition.

• Trial eligibility criteria are nuanced—like “Stage II HER2-negative breast cancer in postmenopausal women with no thromboembolic history.”

• Most systems rely on keyword matching, which misses the subtle but critical connections needed for accurate pre-screening.

That’s where BEKplatform’s ontology makes the difference.

What Is the BEK Ontology?

Think of it as a clinical-grade translation engine for research. BEKhealth built its ontology by aggregating more than 400 medical libraries and training it against 70,000+ real-world trial protocols.

It maps:

• Medical terms, concepts, and diagnosis patterns

• Synonyms, abbreviations, and shorthand

• Relationships between conditions, symptoms, medications, labs, and procedures

This means BEKplatform doesn’t just “search”—it understands context. And it doesn’t just match against patient data—it matches against the specific eligibility logic of real trial protocols.

It aligns with trends in semantic interoperability, which organizations like HL7 and ONC have long championed to improve healthcare data usability.

From Fragmented Data to Real Patient Matches

Imagine a trial requiring patients who:

• Have had type 2 diabetes for more than 5 years

• Are currently on second-line therapy

• Show signs of progressive kidney dysfunction

• Have no recent history of cardiovascular events

A traditional pre-screener may miss this patient unless every box is manually checked. But BEKplatform, powered by its ontology, can:

• Infer second-line therapy from “on metformin and insulin”

• Recognize kidney decline through eGFR trend analysis

• Interpret notes like “no CV events since 2020” as a negative cardiac history

This level of contextual interpretation aligns with what NIH researchers have identified as a critical gap in automated patient matching tools—especially when unstructured data is involved.

Real-World Impact

BEKhealth’s customers are seeing measurable results:

• 10x more protocol-matching patients identified

• 2x more patients enrolled per study

• Hundreds of hours saved in manual chart review

In one example, a lung cancer study site pre-screened 10 eligible patients in just three weeks and enrolled three. With manual chart review, that outcome would have taken significantly longer—or been missed entirely.

This shifts how sponsors view feasibility and how quickly sites can activate and recruit.

Why This Matters for Everyone in the Ecosystem

Sponsors:

• Faster recruitment reduces delays and costs

• Real-time feasibility helps optimize study design and site selection

• Better site performance improves trial predictability

Sites:

• Less time on chart review, more time with patients

• Actionable pre-screening lists accelerate enrollment

• Greater ability to support multiple studies at once

Patients:

• Better access to trials, especially for those historically overlooked

• Faster connection to treatment options

• A more equitable chance at participation—something highlighted by groups like CISCRP in their push for more inclusive recruitment practices

Ontology That Works Like a Human—at Machine Scale

BEKplatform’s ontology isn’t a feature. It’s the foundation. It allows BEK’s AI to interpret complex medical data with nuance—across both structured and unstructured fields—and match patients to trial criteria with unmatched precision.

And unlike opaque systems, BEKhealth’s AI is designed for transparency. Human experts stay in the loop, and the system evolves through real-world use.

You don’t need to memorize the 24 million terms behind the scenes. You just need to know they’re working for you—quietly powering a smarter, faster, and more inclusive way to match the right patient to the right study.