Inferensys

Glossary

Eligibility Rule Engine

A deterministic software system that evaluates a set of structured patient facts against a predefined library of clinical trial eligibility rules to produce a pass/fail decision.
Developer reviewing semantic search engine results on laptop, relevance scores visible, technical search demo.
DETERMINISTIC CLINICAL LOGIC

What is an Eligibility Rule Engine?

A deterministic software system that evaluates a set of patient facts against a predefined library of clinical trial eligibility rules to produce a pass/fail decision.

An Eligibility Rule Engine is a deterministic software system that evaluates a set of structured patient facts against a predefined library of clinical trial criteria to produce a binary pass/fail decision. Unlike probabilistic AI models, it applies explicit, auditable logic—such as IF hemoglobin < 9 g/dL THEN exclude—ensuring that every eligibility determination is transparent, repeatable, and fully traceable to the originating protocol requirement.

The engine operates by ingesting a computable phenotype definition and executing it against a patient's normalized clinical data, resolving complex logical expressions, temporal constraints, and value comparisons. It serves as the final, high-precision gate in a hybrid matching architecture, validating candidates identified by semantic search before site contact, thereby eliminating false positives caused by ambiguous criteria interpretation.

CORE CAPABILITIES

Key Features of an Eligibility Rule Engine

An eligibility rule engine is the deterministic backbone of automated clinical trial screening, evaluating structured patient facts against a library of computable criteria to produce definitive pass/fail decisions. The following features define a production-grade system.

01

Deterministic Criteria Evaluation

The engine applies Boolean logic (AND, OR, NOT) and comparison operators (=, >, <, ≥, ≤) to patient facts with absolute reproducibility. Unlike probabilistic AI models, a deterministic engine guarantees the same input always yields the same output—critical for regulatory audit trails.

  • Evaluates atomic criteria like Hemoglobin >= 9.0 g/dL or Diagnosis == 'Non-Small Cell Lung Cancer'
  • Supports nested logic groups for complex compound criteria
  • Produces a definitive PASS, FAIL, or INCONCLUSIVE (insufficient data) result per criterion
02

Temporal Constraint Resolution

The engine validates time-window constraints against a patient's longitudinal record to enforce criteria like washout periods, disease progression timelines, and recency requirements.

  • Evaluates constraints such as Prior chemotherapy completed > 28 days before enrollment
  • Reconstructs clinical event sequences from timestamped data points
  • Handles relative time offsets (e.g., within 6 months of diagnosis) and absolute date thresholds
  • Flags violations of minimum or maximum interval requirements between events
03

Ontology-Backed Terminology Normalization

The engine maps clinical concepts to standardized terminologies (SNOMED CT, LOINC, RxNorm, ICD-10-CM) before evaluation, ensuring that synonymous expressions are consistently interpreted.

  • Normalizes high blood pressure and essential hypertension to the same SNOMED concept
  • Converts lab values to consistent units (e.g., mg/dL vs. mmol/L) before comparison
  • Resolves medication synonyms to RxNorm ingredient-level concepts for concomitant medication checking
  • Eliminates false negatives caused by terminology mismatch
04

Criteria Decomposition Engine

Complex, multi-part eligibility criteria are automatically decomposed into atomic, independently evaluable units before execution. This enables granular pass/fail tracking and clear screen failure reporting.

  • Parses a criterion like Histologically confirmed NSCLC, Stage IIIB-IV, with ECOG 0-1 into three separate evaluable facts
  • Maintains parent-child logical relationships between decomposed atoms
  • Enables partial matching and weighted scoring when full criteria cannot be met
  • Supports criteria weighting to prioritize critical inclusion factors over secondary ones
05

Audit Trail and Explainability

Every eligibility decision is accompanied by a complete, immutable decision trace that records which patient facts were evaluated, which criteria passed or failed, and why.

  • Logs the exact patient value compared against each criterion threshold
  • Provides a human-readable justification for every FAIL result (e.g., Hemoglobin 8.2 g/dL < required 9.0 g/dL)
  • Supports regulatory compliance with 21 CFR Part 11 audit trail requirements
  • Enables rapid screen failure analysis and protocol refinement
06

Protocol Amendment Handling

The engine supports versioned rule libraries that automatically detect and integrate changes from formal protocol amendments without requiring full system redeployment.

  • Maintains distinct rule sets for each protocol version with effective dates
  • Automatically applies the correct version based on the patient's screening date
  • Provides diff reports showing exactly which criteria were added, removed, or modified
  • Ensures continuity of screening operations during protocol transitions
ELIGIBILITY RULE ENGINE

Frequently Asked Questions

Clear, technical answers to common questions about the deterministic software systems that evaluate patient facts against clinical trial criteria to produce pass/fail eligibility decisions.

An eligibility rule engine is a deterministic software system that evaluates a set of structured patient facts against a predefined library of clinical trial inclusion and exclusion criteria to produce a binary pass/fail decision. Unlike probabilistic AI models, it operates on explicit, auditable logic. The engine ingests a patient's normalized clinical data—such as diagnoses coded in ICD-10-CM, lab results in LOINC, and medications in RxNorm—and sequentially executes a series of IF-THEN rules. For example, a rule might state: IF hemoglobin < 9 g/dL THEN EXCLUDE. The engine resolves all logical operators (AND, OR, NOT), validates temporal constraints (e.g., 'surgery within the last 6 months'), and outputs a final eligibility determination along with a trace of every rule evaluated, ensuring full algorithmic explainability for regulatory audits.

Prasad Kumkar

About the author

Prasad Kumkar

CEO & MD, Inference Systems

Prasad Kumkar is the CEO & MD of Inference Systems and writes about AI systems architecture, LLM infrastructure, model serving, evaluation, and production deployment. Over 5+ years, he has worked across computer vision models, L5 autonomous vehicle systems, and LLM research, with a focus on taking complex AI ideas into real-world engineering systems.

His work and writing cover AI systems, large language models, AI agents, multimodal systems, autonomous systems, inference optimization, RAG, evaluation, and production AI engineering.