Temporal Reasoning for Eligibility

The automated extraction and normalization of time-bound logic from unstructured eligibility criteria. The engine identifies and structures complex temporal expressions such as 'within 28 days of enrollment', 'no myocardial infarction in the past 3 months', or 'progressive disease after at least 2 prior lines of therapy'.

• Converts free-text durations into machine-readable intervals (e.g., P28D, P3M)
• Resolves anchor events like 'enrollment date' or 'start of treatment' to specific timestamps
• Handles relative temporal operators: before, after, during, within
• Distinguishes between point-in-time events and durational states

The reconstruction of a patient's clinical timeline by ordering discrete medical events based on their timestamps. The engine assembles a longitudinal patient record from disparate data sources—encounters, lab results, medication orders, and diagnoses—to validate complex sequential logic.

• Orders events like diagnosis → first-line therapy → progression → second-line therapy
• Detects gaps and inconsistencies in timestamped data
• Validates that required sequences occurred in the correct temporal order
• Supports both absolute dates and relative ordering when precise timestamps are unavailable

The automated verification that a sufficient drug-free interval has elapsed between a patient's last dose of a prohibited medication and the trial's screening date. This is critical for ensuring patient safety and protocol compliance.

• Calculates the interval between last administration date and screening date
• Cross-references prohibited medication lists from trial protocols
• Accounts for drug half-life and known pharmacokinetic washout durations
• Flags borderline cases where the washout window is within a configurable margin of error

The evaluation of a patient's disease trajectory over time to determine if it meets protocol-specified progression requirements. The engine analyzes temporal patterns in tumor markers, imaging reports, and clinical assessments to validate criteria like 'documented progression within 6 months of last therapy.'

• Tracks RECIST criteria assessments across multiple timepoints
• Calculates progression-free survival intervals from treatment start dates
• Identifies the date of first documented progression for eligibility anchoring
• Distinguishes between radiographic progression and clinical progression events

The identification of conflicts where multiple time-dependent criteria impose incompatible windows on a patient's record. The engine detects when two required events must occur within overlapping but mutually exclusive timeframes.

• Evaluates all active temporal constraints simultaneously
• Detects impossible combinations (e.g., 'within 14 days' AND 'after 28 days')
• Prioritizes the most restrictive window when conflicts are resolvable
• Generates explicit conflict explanations for clinical reviewer audit

The resolution of temporal expressions that reference a dynamic anchor point rather than a fixed calendar date. The engine binds relative references like 'date of informed consent,' 'Cycle 1 Day 1,' or 'time of randomization' to the appropriate patient-specific or protocol-defined timestamps.

• Maintains a registry of all protocol-defined anchor events
• Dynamically resolves anchors based on patient-specific milestone dates
• Handles nested relative references (e.g., '30 days after the last dose of prior therapy')
• Supports both forward-looking and retrospective temporal reasoning