Skip to main content

Verify any clinical value against its source document

Every event on the Patient Timeline links to the source document and the text span that produced it. Open the value, open the document behind it, compare, and correct or escalate without leaving the workflow.

Patient Timeline with Full Provenance and Source Links

Who this is for​

Anyone who has to defend an extracted value: data quality analysts and curation reviewers; clinical data scientists running acceptance testing during a pilot; registrars validating a field before submission; AI governance and compliance teams documenting how output was checked; clinicians reviewing a sample before the data is used; and analysts about to publish a cohort count someone will question.

Why it matters​

During evaluation, a clinical data scientist samples 50 extracted stage values and checks each against its pathology report. During production, a registrar disagrees with an extracted recurrence date and needs to see what the model read. During an audit, someone asks where a submitted figure came from.

All three are the same question: show me the source. A structured value with no path back to the document it came from cannot be defended to an IRB, a registry auditor, or a skeptical oncologist, and a number that cannot be defended does not get used.

What you gain​

Every timeline event links to the source document and the span that produced it, so verification takes seconds rather than a chart pull. Corrections flow back to the workflow that raised them, which means the review loop closes instead of producing a list of complaints. For evaluations, this is the workflow that turns "the accuracy looks good" into a documented sample your governance committee can read.

Before you start​

  • An active workflow containing patients: a cohort, a registry case, a care gap result, or a curation result
  • Source documents ingested and linked. Timeline events built from structured imports alone will not have documents behind them
  • A defined escalation path for issues a reviewer cannot resolve

Step 1: Open a patient from the active workflow​

  1. Open the relevant workflow, such as a cohort, registry case, care-gap result, or curation result.
  2. Select the patient that requires review.
  3. Open the patient profile, patient details, or patient timeline from the selected workflow.
  4. Review the patient identity and context to confirm that the correct patient is open.
  5. Identify the event, value, gap, registry field, or extracted result that needs source-level review.

Step 2: Review the Patient Timeline​

  1. Review the patient timeline for relevant diagnoses, visits, medications, measurements, procedures, notes, and clinical events.
  2. Use the timeline to understand the sequence of events around the item being reviewed.
  3. Check whether the timeline contains evidence that supports, contradicts, or explains the selected result.
  4. Use filters or timeline navigation if the patient has a large number of events.
  5. Select the relevant event or document reference for deeper review.

Step 3: Open supporting source documents​

  1. Open the source document linked to the selected timeline event, extracted field, care gap, registry case, or curation result.
  2. Review the document title, note type, date, author or provider context, and patient identifiers.
  3. Locate the sentence, section, table, or evidence span that supports the reviewed value.
  4. Compare the source text with the timeline event or extracted result.
  5. Open additional supporting documents if the first document does not provide enough context.

Step 4: Confirm whether the timeline and extracted values match the source evidence​

  1. Compare the displayed value, timeline event, or extracted field against the source document.
  2. Confirm whether the value is correct, incomplete, unsupported, outdated, or contradicted by source evidence.
  3. Check date, clinical concept, negation, laterality, severity, status, and document context where relevant.
  4. Determine whether the issue is a review correction, missing evidence, source-data issue, or workflow limitation.
  5. Record the review decision according to the available workflow.

Step 5: Correct or escalate the issue​

  1. Correct the reviewed value where the workflow supports editing.
  2. Add a review note if the workflow supports reviewer comments.
  3. If the value cannot be corrected in the current workflow, escalate the issue for data review or product follow-up.
  4. Include enough context in the escalation, such as patient, document, date, field, expected value, and observed issue.
  5. Save the correction, review decision, or escalation note.

Step 6: Return to the original workflow​

  1. Return to the cohort, registry case, care-gap result, or curation result where the review started.
  2. Confirm that the updated decision or review note is reflected where expected.
  3. Continue reviewing additional patients, fields, gaps, or cases as needed.
  4. Use the patient timeline and source documents as supporting context for final workflow decisions.

Recipe reference​

Each stage of this scenario is also a reusable building block.

Review a Patient Timeline before taking action​

When to use on its own: When you need patient context before accepting an extraction, confirming a gap, or reviewing a registry case.
Features involved: Patient Timeline, patient details, source document review, patient-level assistant where available.
Edge cases / limitations: Timeline completeness depends on ingested source data. Missing events may reflect missing source data rather than absence of care.
Value: Helps you trust and validate outputs before acting.