1. Healthcare Analytics
  2. Process
  3. Run reconciliation and correction
Process · Run reconciliation and correction

Run reconciliation and correction

Resolve failed or unreconciled runs and correct a published revision without losing the original.

Step by step

  1. Find the run Data Engineer

    Data Engineer opens runs that ended Failed or reconciliation required and reads the failure reason, input hash, source revision and result reference.

    • Failed run identified
    • Updates Operation runs
  2. Confirm the steps Data Engineer

    Data Engineer checks each numbered step. A step left Unknown is confirmed against its external source using the receipt ID, receipt date and evidence before anything is retried.

    • Step outcome confirmed
    • Updates Operation steps
  3. Retry or reconcile Data Engineer

    Data Engineer retries the run with the same operation key, or records why it cannot be completed. A retry must not write a second result for the same definition and window.

    • Run completed once
    • Updates Operation runs, Metric results
  4. Open a correction Analytics Manager

    When a published result is wrong, Analytics Manager starts the correction workflow. The prior revision is kept, a successor revision is written by incrementing Revision, and state history records the full transition.

    • Successor revision written
    • Updates Outcome cohorts, Operation runs, State history
  5. Review the successor Clinical Reviewer

    Clinical Reviewer reviews the successor revision and approves it with Decision: Pending → Approved, or returns it with a reason. The earlier approval does not carry over.

    • Successor approved or returned
    • Updates Review decisions
    ApprovalA person approves before the work moves on.
  6. Republish Analytics Manager

    Analytics Manager publishes the approved successor through the publication workflow. The original result stays in place and the new one is tied to its own run.

    • Corrected rates published
    • Updates Operation runs, Outcome cohorts

Where it usually breaks

And how to keep it from happening.

  • A step marked `Unknown` is retried without checking the external system, and the effect happens twice.
  • A correction overwrites the published revision and the earlier rate can no longer be explained.
  • The successor revision is published on the strength of the original approval.
  • A failed run is deleted instead of reconciled, and its failure reason is lost.
Data Engineer
, Analytics Manager, Clinical Reviewer.

Set up Healthcare Analytics your way.

Start free, then use Proto to change fields, steps and approvals before your team goes live.

How to get started