SNF Interface Acceptance: Reconcile ADT, Pharmacy and Laboratory Events

Begin with the operating path

For each connection, record sender, receiver, participating facilities, event types, direction, source identifiers, expected timing and the place unresolved work appears. Name the parties responsible for transport, mappings and operational review separately. One vendor may deliver a message successfully while another rejects its facility code.

The 2025 SAFER System Management guide explains why interfaces need attention to mapping, maintenance and different suppliers' change cycles. The acceptance pack below is an original exercise for the buyer and participating technical teams; it is not a universal message specification.

Reconcile events at several boundaries

Use a synthetic set of ten expected result events, E01–E10. Inject these conditions deliberately:

The ten-message transport total looks complete but three expected events have not reached the intended queue: E08, E09 and E10. Each needs a different investigation. Clinical review of the seven routed items is yet another state. Do not report “ten results complete” from this data.

A useful local reconciliation has one row per expected event, with source key, receipt reference, identity disposition, queue reference and latest relevant status. Protect clinical content in approved systems; a broader project record can use synthetic or restricted reference keys.

Test ADT order and duplication explicitly

Prepare one synthetic arrival at 10:00, transfer at 12:00 and correction at 13:00. Deliver the correction before the transfer, then resend the arrival. The participating systems must define expected behavior: controlled sequencing, a visible exception or another reviewed mechanism. The acceptance result is the agreed current state plus understandable history.

Save message IDs and event IDs. The same event retransmitted differs from a new event that resembles it. A system that suppresses all similar messages may lose a legitimate correction; a system that treats every receipt as new may create duplicate episodes. Inspect the person, episode and location relationships after recovery, not merely the acknowledgment log.

Use a separate unknown-facility case. Its expected result can be an unresolved item rather than guessed placement. A regional connection needs facility-specific mapping tests even when the transport endpoint is shared.

Translate pharmacy status into exact meaning

Ask what creates each visible label: queued, sent, transport acknowledged, content accepted, rejected, corrected or pending. None should imply medication administration unless the actual administration workflow provides that evidence.

In a synthetic case, transport acknowledgment turns an icon green but content is later rejected because the destination facility is unmapped. Check whether the later rejection reaches an ordinary monitored view. If users still interpret green as finished, the evaluation has found a status-design or workflow problem even though the original acknowledgment was technically correct.

After correction, preserve the relationship to the original attempt. Clinical and pharmacy leaders define ordering and follow-up procedures; the technical test establishes what information those procedures can rely on.

Rehearse outage recovery without inventing missing events

Pause a test endpoint, allow a known batch to queue, then restore it through the approved test process. Reconcile expected unique events, previously accepted events, replayed messages and unresolved exceptions. Confirm that reprocessing an event does not silently create a second task or replace a newer correction.

The cost and support scope should include partner participation, monitoring and later changes. ONC EHR Contracts Untangled, September 2016 treats integration responsibilities as a procurement concern. A setup charge and one successful message do not establish who will investigate a rejected mapping at night or retest after a partner upgrade.

An independent facility may use an outsourced interface coordinator; specify a reachable operational handoff and the source records that coordinator needs. A group may centralize monitoring, but local staff still need visibility into their unresolved items. Test the case where central transport is healthy and only one facility's routing is broken.

Accept the failure path as part of the connection

Keep the initial inputs, expected outcomes, configuration versions and actual results. Include at least one deliberately detectable failure; a clean run provides little assurance if the checker never notices the injected fault. Retest changed mappings against related sites and messages, not just the first repaired example.

The final inventory should identify active versus planned connections and the last tested scope. That distinction prevents an architecture diagram from becoming an unsupported claim that a future integration is already available.