Phenotype Execution Engine

The engine parses and evaluates Boolean logic trees defined in phenotype authoring languages. It resolves nested AND/OR/NOT operators against patient fact tables, applying operator precedence and short-circuit evaluation for performance.

• Supports arbitrarily nested parenthetical groupings
• Handles negation with temporal scope (e.g., "no diagnosis of X in the last 6 months")
• Optimizes query execution by reordering predicates based on selectivity estimates

The engine validates time-windowed criteria by reconstructing patient event timelines from timestamped clinical data points. It determines whether events occurred in a specified sequence, within a defined interval, or relative to an index date.

• Resolves Allen's interval algebra relationships (before, after, overlaps, during)
• Computes gap analysis for washout periods and lookback windows
• Handles relative temporal anchors such as "within 30 days of first diagnosis"

The engine dynamically expands value set references by resolving terminology bindings against hosted or federated ontology services. A single phenotype criterion referencing "Type 2 Diabetes" is expanded into all descendant codes across SNOMED CT, ICD-10-CM, and RxNorm.

• Supports intensional value sets defined by semantic relationships
• Caches terminology expansions to minimize external service calls
• Maintains versioned bindings to ensure cohort reproducibility over time

The engine decouples phenotype logic from the underlying data storage schema. It translates phenotype queries expressed against a canonical information model into the native query language of the target repository, whether that is FHIR, OMOP CDM, i2b2, or a proprietary EHR data warehouse.

• Maps phenotype data elements to physical table structures
• Generates optimized SQL, CQL, or FHIR API calls at runtime
• Enables a single phenotype definition to execute across heterogeneous data sources

Once evaluation completes, the engine materializes the resulting patient set with full line-level provenance. Each patient's inclusion or exclusion is traceable to the specific criteria and data elements that determined the outcome.

• Produces patient-level audit trails for regulatory inspection
• Supports incremental cohort refresh as new clinical data arrives
• Exports cohorts in standard formats for downstream analytics or trial recruitment systems

The engine partitions phenotype evaluation across distributed compute resources to handle populations exceeding millions of patients. It decomposes complex phenotypes into independent sub-queries that can execute concurrently.

• Implements map-reduce patterns for population-scale screening
• Applies predicate pushdown to filter patients before expensive temporal joins
• Monitors query execution with telemetry hooks for long-running phenotype jobs

A machine-processable definition of a clinical condition, expressed as a set of logical expressions and data queries. The Phenotype Execution Engine consumes these definitions to identify patient cohorts. Computable phenotypes translate clinical concepts—such as 'Type 2 Diabetes with HbA1c > 8.0'—into executable rules that combine:

• Data queries against structured fields (labs, vitals, diagnoses)
• Logical operators (AND, OR, NOT)
• Temporal constraints (within 6 months, after diagnosis)
• Value thresholds (HbA1c > 8.0%, eGFR < 60)

A standardized, machine-readable syntax for expressing complex inclusion and exclusion rules. The Phenotype Execution Engine parses CDL to construct its internal execution plan. Key characteristics include:

• Declarative syntax that separates logic from execution
• Support for temporal operators (BEFORE, AFTER, WITHIN, DURING)
• Modular rule composition allowing reuse of sub-definitions
• Ontology binding to standard terminologies like SNOMED CT and LOINC CDL enables portability of phenotype definitions across different EHR systems and data models.

The engine's capability to interpret and validate time-dependent clinical constraints against a patient's longitudinal record. This is critical for trial eligibility where sequence and timing matter. The engine resolves:

• Washout periods: 'No chemotherapy within 28 days prior to enrollment'
• Disease progression timelines: 'Documented progression after first-line therapy'
• Relative time windows: 'HbA1c measured within 3 months of screening'
• Event sequencing: 'Diagnosis of hypertension BEFORE initiation of ACE inhibitor' Temporal reasoning transforms raw timestamps into clinically meaningful sequences.

The process of converting parsed, structured eligibility criteria into executable database queries. The Phenotype Execution Engine performs this translation by:

• Mapping clinical concepts to specific data elements (e.g., 'HbA1c' → LOINC 4548-4)
• Generating SQL or FHIR API calls against the clinical data repository
• Optimizing query execution order to minimize computational cost
• Handling data model heterogeneity across OMOP, i2b2, or native EHR schemas This translation layer abstracts the clinical logic from the underlying data infrastructure.

A deterministic software system that evaluates patient facts against predefined clinical trial rules to produce a pass/fail decision. The Phenotype Execution Engine functions as a specialized rule engine with:

• Forward-chaining inference to derive new facts from existing data
• Conflict resolution when multiple rules produce contradictory results
• Explanation traces documenting which criteria passed or failed
• Incremental evaluation for efficient re-screening when data changes Unlike probabilistic matching, the rule engine provides auditable, reproducible cohort determinations.

The automated assembly of a chronological patient history from disparate, timestamped clinical data points. Before the Phenotype Execution Engine can evaluate temporal constraints, it must:

• Ingest events from multiple sources (EHR encounters, lab systems, pharmacy records)
• Normalize timestamps across different time zones and granularities
• Resolve conflicting records (duplicate labs, corrected reports)
• Construct a unified event sequence ordered by clinical occurrence date This reconstructed timeline serves as the temporal backbone for all eligibility evaluations.