Our Take
The real problem isn't matching speed—it's that traditional systems miss eligible patients buried in clinical notes and can't explain their logic to the teams using them.
Why it matters
Clinical trial enrollment is a bottleneck. Patient-trial mismatch wastes weeks. CRCs and medical teams need both accuracy and visibility into why a patient surfaces, not a black-box ranked list.
Do this week
CRCs: audit your current eligibility screening workflow this week to document which patients were flagged as ineligible but shouldn't have been—that's the gap an LLM approach targets.
Paradigm Health brings LLMs into trial eligibility screening
Paradigm Health released a clinical trial matching solution that combines structured data (demographics, lab results, diagnoses coded in EHRs) with unstructured clinical text (notes, imaging reports, visit summaries) to improve patient-trial alignment. The tool uses LLMs to extract nuanced eligibility signals from free-text notes that traditional discrete-field matching misses.
The offering also surfaces natural language summaries explaining why each patient is included or excluded, allowing clinical research coordinators (CRCs) and medical teams to review the reasoning without jumping between systems.
Transparency is the problem, not just accuracy
Traditional patient-trial matching relies on structured data alone: age, BMI, lab values in coded fields. This approach fails in two ways. First, it misses patients whose eligibility depends on clinical context documented only in notes (e.g., a patient whose comorbidity severity is described narratively, not coded). Second, when a patient surfaces or is rejected, the CRC has no clear explanation, creating friction and slowing enrollment decisions.
The source notes that lack of transparency around inclusion and exclusion leaves teams uncertain about why patients were surfaced, creating inefficiencies downstream. An LLM-based system that can read the entire patient record and explain its logic directly addresses this visibility gap. CRCs gain confidence to act on results faster because they can verify the reasoning on screen, without navigating to separate systems or calling analysts.
How to start thinking about this
If you run clinical research operations, the relevant question is not whether LLMs can read notes faster than humans (they can), but whether your current screening process is missing eligible patients and whether you can audit why. Start by sampling rejected patients: pull 20–30 cases flagged ineligible and ask your CRCs whether the reasoning was defensible. If you find false negatives, an LLM-assisted screening system that reads the full clinical record and explains its logic is worth piloting. If your current system is accurate but slow, this tool helps mainly by speeding the review cycle, not by finding new patients.