Cohort Definition Language

CDL uses a declarative syntax where rules describe what constitutes cohort membership, not how to retrieve it. This separates clinical intent from execution logic.

• Inclusion criteria: Conditions that must be true for a patient to qualify
• Exclusion criteria: Conditions that disqualify an otherwise eligible patient
• Rule composition: Atomic criteria combine via logical operators (AND, OR, NOT)

Example: INCLUDE (Diagnosis.type = "T2DM") AND NOT (Medication.code IN {"insulin"})

CDL natively supports time-window logic to express sequence-dependent eligibility rules that are common in trial protocols.

• BEFORE / AFTER: Ordering of clinical events relative to each other
• WITHIN [n] DAYS: Proximity constraints between events
• DURATION >= [n]: Minimum length of a condition or treatment
• FIRST_OCCURRENCE: References the earliest instance of an event

Example: Diagnosis.date AFTER Procedure.date WITHIN 30 DAYS

CDL references standardized terminologies rather than free-text, ensuring unambiguous interpretation across heterogeneous data sources.

• SNOMED CT: Clinical findings and procedures
• RxNorm: Medication concepts
• LOINC: Laboratory tests and results
• ICD-10-CM: Diagnosis and procedure codes

Value sets group related codes: Condition.code IN ValueSet("Cardiovascular_Disease") expands to hundreds of mapped codes automatically.

Complex protocols require hierarchical rule nesting to express precedence and scoping correctly.

• Parentheses define evaluation order for compound expressions
• Sub-groups can represent distinct eligibility domains (e.g., lab criteria, medication criteria)
• Each group evaluates independently before combining at the parent level

Example structure:

codeCopy
INCLUDE (
  (Age >= 18 AND Age <= 65) AND
  (Diagnosis.code = "T2DM") AND
  (Lab.HbA1c.value > 7.5)
)
EXCLUDE (
  (Medication.code IN {"insulin"}) OR
  (Condition.code IN ValueSet("Renal_Impairment"))
)

CDL supports numerical comparisons against clinical measurements with unit awareness for precise eligibility filtering.

• Comparison operators: >, <, >=, <=, =, !=
• Unit specification: Lab.HbA1c.value > 7.5 unit="%"
• Range expressions: Lab.eGFR.value BETWEEN 30 AND 60 unit="mL/min/1.73m2"
• Trend operators: Lab.HbA1c.value DECREASING OVER 6 MONTHS

This enables criteria like "eGFR must be between 30 and 60 mL/min" to be evaluated correctly regardless of source data units.

CDL supports named rule definitions that can be referenced across multiple cohort definitions, promoting consistency and reducing duplication.

• Define-once, use-many: Common criteria like "Adult Type 2 Diabetic" become reusable modules
• Parameterized rules: Modules accept arguments for flexible instantiation
• Version control: Rule modules carry version identifiers for audit trails

Example: DEFINE AdultT2DM AS (Age >= 18 AND Diagnosis.code = "T2DM") then INCLUDE AdultT2DM in any cohort.

The software component that runs a computable phenotype definition against a clinical data repository. It interprets CDL expressions, resolves temporal constraints, and queries both structured fields and unstructured notes to return a final patient cohort. Key capabilities include:

• Logical resolution: Evaluating nested AND/OR/NOT operators
• Temporal reasoning: Validating event sequences and time windows
• Value set expansion: Resolving ontological codes to their descendants
• Data source federation: Querying across EHR, claims, and lab systems simultaneously

The process of converting parsed, structured eligibility criteria into executable database queries. This translation layer bridges CDL's declarative syntax with the imperative query languages of target data repositories. A single CDL criterion like 'HbA1c > 7.0% within last 6 months' must be decomposed into:

• A value set lookup for HbA1c LOINC codes
• A temporal filter for encounters within 180 days
• A numeric threshold evaluation (> 7.0)
• A unit normalization step to handle varying lab result formats

The systematic review of reasons why pre-screened patients failed to meet trial eligibility, used to optimize recruitment strategies and refine protocol inclusion criteria. When a CDL-defined cohort is executed, the failure signature—which specific criteria excluded each patient—is logged. Analysis of aggregate screen failures reveals:

• Overly restrictive criteria causing unnecessary exclusion
• Ambiguous criteria requiring clarification in the CDL expression
• Data quality gaps where missing lab results cause false exclusions
• Protocol amendment opportunities to improve enrollment rates

The assignment of relative importance scores to individual inclusion and exclusion criteria to prioritize patient matches. While CDL defines binary eligibility (pass/fail), criteria weighting enables ranked matching for scenarios where multiple trials compete for the same patient. Critical safety criteria receive maximum weight, while procedural preferences receive lower weight. The weighted score supports:

• Trial recommendation engines that suggest optimal matches
• Site feasibility assessments estimating enrollment potential
• Patient prioritization when screening capacity is constrained