Deidentified outcome facts

A published rate is only as good as the observations under it. Deidentified outcome facts are those observations: one row per fact, tied to its parent cohort, with a sequence, a description, the date it was observed, the cohort date and the counts it contributes. Each fact keeps its source reference and the consent basis under which it was used. Identifiable patient data stays in the source system; the app holds the deidentified fact and the reference back to it.

Facts arrive with Outcome = 'Pending'. The Outcomes Analyst checks each one against the source snapshot and the outcome definition, then accepts or rejects it. Rejected facts stay on the cohort with their outcome recorded, so the reviewer can see what was excluded. Supporting documents, such as an extract log or a data-quality note, are attached to the cohort as evidence and wait for verification.

• Accept or reject explicitly. A fact moves Pending → Accepted or Pending → Rejected; nothing leaves the pending queue silently.
• Consent travels with the fact. consent basis is recorded per fact as well as per cohort, so a mixed cohort can be explained.
• Evidence waits for a reviewer. Evidence documents start with Verified = 'No' and only a Clinical Reviewer marks them verified.

Agree the consent bases your organization uses and the exact wording to record for each. Decide which source reference identifies a fact uniquely, so a reloaded extract does not duplicate rows. Agree the rejection rules for facts outside the cohort window, facts without a consent basis and facts whose counts do not reconcile with the snapshot. Confirm where deidentification happens before load, and who in your organization signs off that process.

Load facts for a cohort from a snapshot whose record count you know. Include one fact with no consent basis and one observed outside the cohort window. The Outcomes Analyst should reject both with the outcome recorded, and accept the rest. The accepted patient count on the facts should reconcile with the patient count held on the cohort.

Reload the same extract. The source references should match the existing facts and no duplicate rows should appear. Then ask the Clinical Reviewer to open the cohort’s evidence and mark the extract log verified. The Outcomes Analyst should not be able to verify evidence for their own cohort.

• Deidentified facts awaiting acceptance — pending facts with their parent cohort, observation date, counts, consent basis and source reference.
• Evidence awaiting verification — evidence documents not yet verified, with document date, external reference and attachment.

• Cohort construction to validation: Build a cohort revision, load its facts and accept or reject each one.
• Independent review to publication: Approve the exact revision and publish its outcome rates.

Review the implementation plan and the data model before preparing your import files.