Laboratory EngineeringGuide
Results, receipts & uncertain-send incident playbook
Reconcile uncertain transport and callbacks without duplicate clinical orders or inferred release.
On this page
State vocabulary
- Transport receipt/SENT: bytes or HTTP response, not LIS application acceptance.
- HL7 CA: commit acknowledgement; AA: application acceptance. Negative acknowledgements are refusals, not acceptance.
- Accession: laboratory identifier; specimen acceptance/rejection: individual specimen state. Neither proves result finality or whole-kit outcome.
- Result received: a reported revision/status. Clinical evaluation and laboratory-authorized release remain separate. A critical flag never releases a result.
- Only released result revisions reach Partner results. Approved synthetic non-Nexus FINAL then CORRECTED evidence retains earlier revisions and returns the latest released revision. Unknown vendor finality remains pending/unreleased.
Incident triage
- Stop new sends after a possible delivery, 409 conflict, timeout or unknown ACK. Preserve the original order/transmission/idempotency identity and exact environment.
- Read authorized canonical order/transmission and receipt evidence; compare lab kit/panel/specimen/accession identity. Do not put clinical free text or keys into support exports.
- Have lab IT confirm application acceptance or rejection against the original identity. Transport reachability alone is insufficient.
- Use only the existing authorized support/reconciliation or installation-bound result-poll recovery path after fresh evidence read-back. Missing installed authority refuses recovery.
- Keep unresolved state quarantined/unknown; never create a replacement order, blind resend or manual success receipt.
Callback and correction controls
- Unverified callback identity/signature, oversized/malformed body, wrong correlation or unsupported status refuses. Duplicate/replay checks need durable canonical evidence before any commit acknowledgement.
- Synthetic signed callback policies do not prove vendor authentication or hosted ingress. Pure translated event objects are not canonical persisted results.
- For an authorized correction, preserve previous released versions and require current clinical/release policy for the new revision. No cross-lab result interpretation or release from status strings.
Known retry safety gap
Important
Disabled native uncertain-send source paths suppress automatic resend after possible delivery. Legacy executable HTTPS/JSON retains classified retries, including a retryable refused sent-state write, and does not establish the same guarantee. Do not document these as identical policies; reconcile clinically uncertain sends with the authorized laboratory owner.
Escalate missing ACK, kit/sample ambiguity, stale revisions or unsupported status through existing workspace support destinations. Do not fabricate a ticket address or bypass the gate to finish an incident.