Clinical Trial Matching Algorithm

Combines deterministic rule-based filtering with probabilistic semantic matching to maximize both precision and recall. Neither approach alone is sufficient for the complexity of clinical trial eligibility.

• Deterministic layer: Executes hard constraints—lab values, exact diagnosis codes, binary inclusion/exclusion flags
• Semantic layer: Uses patient vector embeddings to match clinical narratives where exact terminology differs (e.g., 'shortness of breath' vs. 'dyspnea on exertion')
• Confidence scoring: Each match receives a composite score; high-confidence deterministic matches bypass manual review, while borderline semantic matches route to clinical coordinators

Validates time-dependent clinical constraints against a patient's longitudinal record. Many eligibility criteria involve sequencing, durations, and washout periods that require chronological reasoning.

• Capabilities: Interprets constraints like 'within 28 days of enrollment,' 'no prior therapy in the last 6 months,' or 'disease progression after at least 2 lines of therapy'
• Patient Timeline Reconstruction: Assembles disparate timestamped events—diagnoses, procedures, medication orders, lab results—into a unified chronological sequence
• Temporal operators: Supports Allen's interval algebra (BEFORE, AFTER, DURING, OVERLAPS) for complex clinical event sequencing

Cross-references a patient's active medication list against a trial's prohibited concomitant medications to identify exclusionary drug interactions automatically.

• RxNorm normalization: Maps brand names, generics, and compound formulations to standardized ingredient-level identifiers
• Drug class inference: Identifies exclusion by class (e.g., 'any CYP3A4 inhibitor') even when specific drug names aren't listed in the patient record
• Temporal window checking: Evaluates whether a prohibited medication was taken within the protocol-specified washout window
• Example: A trial excluding 'strong CYP3A4 inducers within 14 days' would flag a patient taking rifampin 10 days prior

Compares a patient's structured genomic variant data against a trial's molecular biomarker requirements with precision. This is critical for precision oncology trials where eligibility depends on specific mutations.

• Variant interpretation: Matches HGVS nomenclature, genomic coordinates, and protein changes to trial criteria specifying alterations like 'EGFR exon 19 deletion' or 'BRAF V600E'
• Biomarker expression thresholds: Evaluates quantitative results—'PD-L1 TPS ≥ 50%' or 'HER2 IHC 3+'
• Complex logic: Handles compound biomarker requirements such as 'KRAS G12C mutation AND no concurrent STK11 mutation'
• Integrates with FHIR Genomics and VCF file parsing for structured variant ingestion

Systematically analyzes why pre-screened patients failed eligibility to optimize recruitment strategies and refine protocol criteria. This creates a feedback loop between screening operations and trial design.

• Failure categorization: Classifies each screen failure by the specific criterion that caused exclusion (e.g., 'failed criterion 4b: inadequate renal function')
• Protocol feasibility signals: Identifies criteria that are disproportionately causing exclusions, potentially indicating overly restrictive or ambiguous protocol language
• Recruitment funnel analytics: Tracks conversion rates from pre-screening → full screening → randomization, highlighting bottlenecks
• Amendment impact simulation: Predicts how relaxing specific criteria would expand the eligible population before formal protocol amendments are filed

The automated extraction and structuring of complex free-text inclusion and exclusion requirements from clinical trial protocols into a machine-readable format. This is the critical first step that converts unstructured protocol documents into structured, logical rules that a matching algorithm can execute. Techniques include: Named Entity Recognition for identifying conditions and medications, relation extraction for linking entities to values, and negation detection to distinguish required from prohibited criteria.

A technique that transforms a patient's entire clinical profile into a dense numerical vector, enabling semantic similarity comparisons with clinical trial requirements. By encoding diagnoses, medications, procedures, and lab results into a unified mathematical representation, the algorithm can perform approximate nearest neighbor searches to find trials that are semantically similar to a patient's profile, even when exact terminology doesn't match.

The AI capability to interpret and validate time-dependent clinical constraints against a patient's longitudinal record. This involves resolving complex temporal logic such as: washout periods (e.g., 'no chemotherapy within 4 weeks'), disease progression timelines (e.g., 'progression after at least 2 lines of therapy'), and sequence validation (e.g., 'surgery followed by adjuvant therapy'). Requires precise clinical event sequencing from timestamped EHR data.

A clinical trial screening system design that combines deterministic rule-based filtering with probabilistic semantic matching to maximize both precision and recall. The deterministic layer applies strict logic to hard criteria (e.g., age ranges, lab value thresholds), while the semantic layer uses embeddings and NLP to match nuanced criteria (e.g., 'history of autoimmune disease') against unstructured patient narratives. This dual approach ensures no eligible patient is missed due to rigid keyword matching.

A quantitative method that assigns a numerical match score to a patient-trial pair based on the weighted fulfillment of all criteria, enabling ranked candidate lists. Key components include: criteria weighting (assigning importance to each criterion), partial match handling (scoring patients who meet most but not all criteria), and confidence scoring (reflecting the certainty of extracted data). This transforms binary eligibility into a nuanced ranking for recruiters.

The systematic review of reasons why pre-screened patients failed to meet trial eligibility, used to optimize recruitment strategies and refine protocol inclusion criteria. By aggregating and categorizing failure reasons—such as concomitant medication exclusions, biomarker mismatches, or comorbidity conflicts—sponsors can identify overly restrictive criteria and adjust protocols to improve enrollment rates without compromising scientific validity.