A clear read for the decision in front of your team.
Every drug page on the public reference web has been continuously edited, challenged, reverted, and negotiated for years — by a distributed community of editors with diverse expertise, competing interests, and established editorial norms. This is the largest adversarial claim-testing process in the world. We measure which claims survive it.
Whether you sit on the committee or prepare for it: we show what changed, what holds, where the argument is vulnerable, and what would tip the conclusion — before the decision locks.
You Tell Us
The decision in front of your team
Every brief starts with a real decision: a P&T committee date, a safety review deadline, a formulary submission window. You tell us the disputed claim, the decision context, and who needs the read.
- A coverage review for an expanding indication
- A safety signal assessment before committee circulation
- A due diligence window on a pipeline asset
We Map the Evidence
The claim is scored across four measurement lanes.
Refract, our open-source observation engine, tracks every revision event in the public record — citations added and removed, guideline updates, label changes, policy shifts. The measurement pipeline scores each lane using weighted signals, contradiction penalties, and temporal boundaries:
- Clinical truth — whether the evidence direction supports the claim
- Ratification — whether authoritative bodies have endorsed or challenged the position
- Economic stake — the material impact if the claim holds or falls
- Operational feasibility — whether the institutional path to action is clear
The Engine Produces a Posture
A defensible read, not a recommendation.
The scoring engine combines the lane scores into a decision posture — one of: monitor, prepare, active review now, or narrow action justified. A committee-facing brief is generated from the assessment, with source references, lane-level scores, identified contradictions, and explicit re-review triggers.
- Scored posture — where the evidence lands across all four lanes
- Strongest contradiction — the single most material challenge to the assessment
- Re-review trigger — "Re-run if clinical truth changes materially"
- Deterministic hash — the assessment is reproducible and auditable
You Decide
The posture is delivered. The trigger is set.
The first page shows the decision, the scored posture, and the timing. The appendix carries the lane-level scores, source audit, and re-review trigger. If the evidence moves before your committee meets, the pipeline detects it and the re-review condition fires.
A reviewer can challenge the framing, adjust the recommendation, or wait for stronger evidence. The brief makes the decision discussable — no more arguing about what the evidence says.
Why It Works
Engineered reproducibility, not analyst judgment.
The assessment produces a deterministic hash. Same inputs, same posture, every time. No analyst drift, no recency bias.
Once the evidence map exists, re-running the assessment is a computational diff against the previous state. The re-review is included because it costs nearly nothing to re-run.
Evidence outside the decision window is rejected by the pipeline — not excluded by an analyst who may not notice the date. Every source carries a verified timestamp.
What You Receive
A brief, not a black-box summary.
The output is a structured document: first-page posture readout, scored evidence lanes, source audit, and re-review trigger. Designed to circulate before committee — not after.
