AI Clinical Trial Eligibility Agent

How Cassidy automates this using AI

Step 1: Ingest trial protocols and structure criteria

The Workflow triggers when new trials or protocol amendments arrive. Cassidy parses inclusion/exclusion criteria, encoding them into computable logic with Boolean operators, temporal windows, numeric thresholds, and arm-specific genomic rules.

Step 2: Normalize patient data from EHR and ancillary sources

Cassidy pulls clinical data—diagnoses, labs, medications, staging, performance status, and molecular results—from the Knowledge Base and connected systems, normalizing values to standard vocabularies like SNOMED CT, LOINC, and RxNorm.

Step 3: Extract eligibility-relevant facts from unstructured notes

For clinical notes, pathology reports, and radiology findings, Cassidy applies NLP to identify key facts with negation, temporality, and uncertainty attributes, converting them to structured concepts with evidence spans.

Step 4: Execute matching logic per trial arm

The Agent evaluates each patient against trial criteria using Boolean and temporal reasoning—checking recency windows, sequence constraints, dose intervals, and organ function thresholds—and records pass/fail/unknown status for every criterion.

Step 5: Generate explainable rationale with evidence traces

For each match, Cassidy produces criterion-level rationale showing the exact data values, timestamps, note snippets, or variant calls that support the determination, along with clear blockers and unknowns requiring manual review.

Step 6: Deliver results to clinical workflows

Matches surface in-context through EHR integration, presenting ranked trial options to clinicians and candidate lists to study teams—with Human-in-the-Loop review for binning, overrides, and outreach activation.