Blogs, Genomics

HL7 Order to Result Reconciliation: How Diagnostic Labs Find the Orders Their Interfaces Lose

HL7 order to report

Ask a lab interface team how their HL7 feeds are performing and most will point to the interface engine dashboard. The channels are running, the queues are clear, and the acknowledgments are coming back as accepted. Ask how many orders from last month never received a final result, and most teams need several days to find out, because very few labs measure it.

That second question is what HL7 order to result reconciliation answers. Anyone who has supported lab interfaces through a few EHR migrations learns that delivering messages and completing orders are separate problems.

An engine can deliver every message it was given and still leave orders open, because the result that should have closed them was never sent, was sent with the wrong identifier, or was filed by the receiving system against a different order.

This article explains how reconciliation works in HL7 v2 lab interfaces, which identifiers it depends on, where orders most often go missing, what CLIA expects of interfaced results, and how to move reconciliation from an occasional audit to a daily control.

What is HL7 order to result reconciliation?

HL7 order-to-result reconciliation is the routine check that every lab order received through an HL7 interface has a matching result sent back, that every result traces to a real order, and that every message was acknowledged. Matching normally runs on the placer order number in ORC-2 or OBR-2 and the filler order number in ORC-3 or OBR-3.

In practice, it means following each order from the first order message, usually ORM^O01 or OML^O21, through every result message, ORU^R01, until the order reaches a final state. That final state is a verified result, a documented cancellation, or a correction that has reached every system that received the original.

Reconciliation builds on the basic split between HL7 ORM vs ORU lab orders and results. The order message carries the request, the result message carries the answer, and every request should eventually receive one.

Why doesn’t interface monitoring catch missing results?

A lab order interface engine monitors messages, and most of the failures that matter happen around orders. An acknowledgment code of AA, as defined in HL7 table 0008, tells the sender that the receiving application accepted the message. It does not tell anyone that the result filed to the correct chart, that it matched the right order, or that a clinician ever saw it.

The published data point the same way. A CAP Q-Probes study reported in CAP TODAY found that 99 percent of results were transmitted accurately from the LIS to the EHR, yet only 79.5 percent of laboratory directors reviewed results in the EHR when it was first implemented. Transmission was rarely the problem. Verification was.

Order completeness is a separate measure again. A 2016 study in Clinical Laboratory Science reviewed inpatient lab orders at a 571-bed community health system before and after its EHR implementation. Orders with missing results fell from 16.5 percent to 11.3 percent, and duplicate orders fell from 9.1 percent to 5.8 percent. The EHR improved both figures, but neither reached zero, and neither would have appeared on an interface dashboard.

Reconciliation is only as reliable as the identifiers behind it. Before looking for gaps, confirm which of these fields each interface actually populates. Many reconciliation problems turn out to be identifier problems that have existed since go-live.

IdentifierHL7 fieldAssigned byWhat usually breaks it
Placer order numberORC-2, OBR-2Ordering system, usually the EHRPaper requisitions entered in the lab without the EHR number, or reorders that reuse a number
Filler order numberORC-3, OBR-3The lab, often the accession numberRe-accessioning, split specimens, and redraws under a new accession
Message control IDMSH-10, returned in MSA-2The sender of each messageEngines that generate a new ID on every retry
Patient identifierPID-3Registration or the ordering systemMRN merges, and outreach patients without a hospital MRN
Ordered testOBR-4Ordering system, mapped to the lab codeCompendium mismatches that split or merge tests
Parent reference for reflex testsOBR-29, with OBR-26 for the parent resultThe labReflex and add-on tests sent without a parent reference

The HL7 standard treats ORC-2 and OBR-2 as the same placer identifier, and ORC-3 and OBR-3 as the same filler identifier. When both segments carry a value, they must match, as the NIST HL7 v2 OBR definition sets out.

The ORC segment is also optional in ORU messages, so some senders place the filler number only in OBR-3, a point Rhapsody’s OBR segment guide makes clearly. A reconciliation query that reads only ORC fields will report orphan results that are not orphans at all. Read both segments every time.

Where do lab orders and results lose each other?

Across most diagnostic lab interfaces, missing or mismatched results fall into six recognizable patterns. Each one has a different cause, so each needs its own check.

1. Orders that never receive a result

Every new order, identified by ORC-1 = NW under the order control codes in HL7 table 0119, should eventually receive at least one ORU^R01 with the same placer number. Compare the latest status of each order against the expected turnaround time for that test. Without a turnaround reference, a slow sendout and a lost result look identical.

Not every unanswered order is an error. A specimen that never arrived is a legitimate outcome, but the lab should close it with a cancel message, typically OC for order canceled, so the ordering system does not show the order as pending indefinitely.

The most common real causes are orders entered manually at accessioning without the placer number, re-accessioned specimens that lose their link to the original order, and test code mappings that split one ordered panel into separately resulted components.

2. Results that match no order

Reversing the direction is where patient safety risk tends to sit. A result that matches no order, often called an orphan or unsolicited result, lands in an unmatched queue in the receiving system, and someone has to file it by hand. Every manual match carries a chance of filing the result to the wrong patient or the wrong order.

The volume can be significant. One children’s hospital described in CAP TODAY reported thousands of interface errors a year from its reference lab interfaces, each a result that could not post to the EHR because it did not match an order, and each corrected manually.

Typical causes include outreach orders placed on paper or by phone, reflex tests created in the LIS without a parent reference, and reference lab results returned without the original placer number.

3. Messages without a clean acknowledgment

Match every outbound message to its acknowledgment through MSH-10 and MSA-2. Look for messages with no acknowledgment at all, messages rejected with AE or AR, and messages resent after a timeout.

Retries deserve close attention. When a receiver processes a message but its acknowledgment is lost on the way back, the sender resends it. Unless the receiver detects the duplicate, the chart now holds the same result twice, and if the engine assigned a new MSH-10 to the retry, the duplicate is difficult to trace later.

4. Result status sequences that should not occur

OBR-25 carries the result status using the values in HL7 table 0123. The most common are P for preliminary, F for final, C for a correction to a final result, X when no results are available because the order was canceled, and A when some but not all results are available. Listing every status sent for each order in time sequence exposes problems that single messages never show.

Status sequenceWhat it usually indicates
P, then FNormal progression
F, then CNormal correction, provided it reached every destination
F, then F with different valuesA correction sent as a new final, which some receivers will overwrite without flagging the change
C with no earlier F at that destinationThe original never arrived there, or arrived under a different identifier
F, then XA cancellation sent after a final result was released
A for longer than expectedPartial results held up by one pending component

Check OBX-11 as well, since it carries the status of each individual observation. A corrected report can show C in OBR-25 while the changed observations still read F if the mapping was never configured for corrections.

5. Corrections that reach some destinations but not others

Corrected results carry the most regulatory weight of any reconciliation check. For each correction, list every destination that received the original: the ordering provider’s EHR, every provider named in OBR-28 Result Copies To, the patient portal, and any health information exchange. Then confirm that the correction reached the same list.

A correction that reaches the ordering provider but not a copied specialist leaves a clinician working from a value the lab has already withdrawn. Labs that want this handled by design rather than by audit usually build clinical report amendment tracking into the reporting layer, so every report version and every recipient is recorded together.

6. Cancels and duplicate orders

The final pattern covers orders that should not have produced results and orders that exist twice. Look for orders canceled by the ordering system (ORC-1 = CA) but resulted anyway, cancel requests that received no response, placer numbers reused across different orders, and the same test ordered twice for the same patient within a short window.

Duplicates are rarely an interface defect. They usually start at order entry, when two providers order the same test or a resent order arrives as a new one. They still consume specimen, reagent, and billing effort, and they make the other five patterns harder to read.

Stopping them before they reach the LIMS is far less expensive than cleaning them up afterward, which is the purpose of duplicate lab order detection at intake. Labs that want the rule applied the same way every time build automated order validation for duplicate lab orders into the order workflow itself.

When most of the gaps trace back to incomplete or duplicate orders, the fix belongs at the front door.

What does CLIA require for interfaced lab results?

CLIA does not mention reconciliation by name, but two parts of the test report standard apply directly. Under 42 CFR 493.1291(a), a laboratory must have manual or electronic systems that ensure results are accurately and reliably sent from the point of data entry to the final report destination in a timely manner, and the regulation specifically includes results reported electronically to interfaced systems.

Under 493.1291(k), when a laboratory detects errors in reported results, it must promptly notify the responsible individuals, issue corrected reports, and keep duplicates of both the original and the corrected report. For an interfaced lab, that obligation extends to every system that received the original. A documented, repeatable reconciliation process is one of the clearest ways to show an inspector that both requirements are being met.

How do you run an HL7 order to result reconciliation?

A first reconciliation can usually be completed by one analyst with access to interface message logs. The steps below follow the order most experienced interface teams use.

  1. Choose the right window. Reconcile orders received between 45 and 15 days ago rather than the most recent 30 days. That gives most tests time to result, so open orders represent real gaps. Extend the window for long assays such as sendouts and genomic panels.
  2. Extract four message sets. Pull inbound orders, outbound ORU^R01 results of every status, acknowledgments, and cancel messages for the window. Many US interfaces follow the HL7 2.5.1 Laboratory Orders Interface (LOI) and Laboratory Results Interface (LRI) implementation guides, which makes field locations predictable.
  3. Match in both directions. Join orders to results on placer and filler numbers, reading both ORC and OBR. Then join results back to orders, and every message to its acknowledgment.
  4. Sort exceptions into the six patterns. Classify each gap as an unanswered order, an orphan result, an acknowledgment problem, a status sequence problem, an incomplete correction, or a cancel or duplicate issue.
  5. Separate explained from unexplained. A rejected specimen with a documented cancel is explained. An order with no result and no cancel is not. Only unexplained items need a root cause, but any patient mismatch or missed correction needs one immediately.
  6. Fix causes, then repeat. Correct the underlying mapping, workflow, or configuration, and run the reconciliation again after 30 days and after every interface change.

What does a first reconciliation usually reveal?

The first run almost always surfaces problems older than anyone expected, often dating to the original go-live. The distribution of exceptions usually says more than the total, and in NonStop’s interface work the patterns tend to point in consistent directions.

  • Unanswered orders and orphan results together almost always indicate an identifier problem at intake. Orders are entering the lab without a placer number, or with one that does not survive accessioning. EHR order validation software that checks each order before it reaches the LIMS addresses the cause rather than the symptom.
  • Acknowledgment problems on their own usually point to engine configuration, such as timeout and retry settings or a receiver rejecting messages for format reasons that nobody is monitoring.
  • Status sequence and correction problems together indicate that the reporting and amendment workflow is the weak point, rather than the connection itself.
  • Cancel and duplicate problems call for order entry rules applied before an accession number is assigned.

When the exceptions are spread across all six patterns, the interface can no longer be treated as the source of truth for order status. That is usually a design problem, and tuning individual channels will not resolve it.

If your first reconciliation turned up more than you expected, or you have not been able to run one yet,

How do you make reconciliation a daily control?

A one-time reconciliation finds existing gaps. A daily process stops new ones from accumulating. Labs that do this well tend to build it in four layers.

  • Capture identifiers in flight. Have the interface engine write placer number, filler number, patient identifier, result status, and acknowledgment code to a reconciliation table for every order and result message. Teams using Mirth Connect order integration can do this within each channel.
  • Report exceptions every morning. Query that table for the six patterns above and send the exceptions to a named owner, with an agreed time to resolve each one.
  • Give clients visibility. Many of the calls asking where a result is stop once ordering practices can check status through a provider order status portal.
  • Measure completeness alongside speed. Track open orders, orphan results, and late corrections next to turnaround time, so lab turnaround time analytics reflect lost orders as well as slow ones. Teams investing in clinical lab turnaround time automation should track completeness too, because a faster workflow that loses orders is not an improvement.

NonStop builds this control at two points. At intake, Intergenix checks every order against completeness, validity, and business rules, reports failures at the field level, deduplicates before validation, and routes failed orders to a structured resolution queue, while keeping the LIMS, EHR, and billing system synchronized. NonStop reports that Intergenix reduced integration effort by 80 percent and saved 8 to 12 hours a week of manual re-entry.

Across the interface itself, NonStop builds Mirth Connect channels for each system pair and runs the old and new interfaces in parallel until the output matches order for order and result for result. A real-time interface health dashboard then monitors the connection after go-live. This is the core of NonStop.io’s HL7 interface development for labs and order-to-report workflow automation.

Questions lab informatics teams ask about reconciliation

What is the difference between interface monitoring and order to result reconciliation?

Interface monitoring checks whether messages are moving and being acknowledged. Reconciliation checks whether each order is complete, meaning it received a final result, every result traces back to an order, and every correction reached the systems that received the original. An interface can pass monitoring and still fail reconciliation.

Which HL7 fields link a lab result back to its order?

The placer order number in ORC-2 or OBR-2 and the filler order number in ORC-3 or OBR-3. The ordering system assigns the placer number, and the lab assigns the filler number, which is often the accession number. When both segments carry a value, the pair must match.

What is an orphan result in a lab interface?

An orphan result is an ORU^R01 message that cannot be matched to any order the receiving system knows about, usually because the placer order number is missing or unknown. Most EHRs hold these in an unmatched results queue until someone files them to the correct patient and order by hand.

Does CLIA require labs to reconcile interface results?

CLIA does not use the word reconciliation. However, 42 CFR 493.1291(a) requires systems to ensure results are accurately and reliably sent to their final report destination, including results reported electronically to interfaced systems. Regular order to result reconciliation is a practical way to show that those systems work.

How often should a lab reconcile HL7 orders and results?

Monthly is a reasonable starting point for a stable interface. Reconcile again after any interface change, EHR upgrade, compendium update, or new client go-live, since those are the points where identifiers and mappings shift. High-volume labs generally move to a daily automated exception report.

Can Mirth Connect reconcile orders and results automatically?

Mirth Connect does not include a reconciliation report, but it keeps a message history for each channel and can write fields such as placer number, filler number, result status, and ACK code to a database as messages pass through. A scheduled query against that table can then produce a daily list of incomplete orders.

Knowing how many orders never came back

Go back to the question at the start of this article: how many orders from last month never received a final result? A lab that reconciles regularly can answer it in minutes, name the cause of each gap, and show an inspector the evidence. A lab that relies on the interface dashboard alone usually finds out from a clinician.

If your team has never measured order completeness, or has measured it and cannot explain what it found, NonStop engineers who build laboratory informatics services for diagnostic labs can review it with you. Bring a sample of message logs and your interface map, and the session will identify whether each gap is a mapping issue, a workflow issue, or a design issue.