What an unmatched study is
An unmatched study is one whose patient or order information doesn't line up with a valid record in the systems of record: the RIS, the EHR or the master patient index. Sites call them different things: unverified, orphaned, broken, exception or QC-queue studies. The symptoms are the same:
- A patient ID that doesn't exist, or belongs to someone else
- A temporary or placeholder identity, such as a trauma alias
- A missing, duplicated or invalid accession number, so there is no order or report to attach to
- Demographics that disagree with the patient record: different birth date, different name spelling
Some PACS hold these studies in a reconciliation queue until someone fixes them. Others accept them and file them under whatever identity they arrived with. Either way, they accumulate. An archive that has run for fifteen years has unmatched studies, usually more than anyone has counted.
How they happen
Manual entry
The exam was acquired without a modality worklist and the technologist typed the patient details at the scanner.
Downtime
The RIS or EHR was down, so studies were acquired under temporary identities and some were never reconciled.
Emergency aliases
Trauma patients registered under a temporary name and ID, merged later in the EHR but not in the PACS.
Wrong selection
The wrong patient or order was picked from the worklist, then corrected in one system but not the other.
Outside studies
Images imported from CDs or other facilities with the sending site's patient IDs.
Merges that didn't arrive
Patient merge or update messages from the EHR that the PACS never received or couldn't apply.
Why they matter in a migration
Inside the old PACS, unmatched studies are a nuisance. During a migration they become a risk:
- Wrong patient records in the new system. A target that accepts studies as they arrive will create a new patient for every bad ID, or worse, file the study under a real patient who happens to share the ID.
- Rejections. Targets that validate against the MPI or require an order will refuse them, and they show up as unexplained failures in the migration report.
- Lost priors. If they are left behind, nobody can find them when a radiologist needs the comparison study.
- Legal hold and retention. They are still part of the medical record. "We didn't migrate them" is rarely an acceptable answer.
The migration is also the best chance you will get to fix them. Every study is being touched anyway.
Finding them
Start from the source inventory you build for validation, and compare it with the systems of record:
- Patient check. Compare every Patient ID and issuer with the MPI or EHR. Flag IDs that don't exist, and IDs whose name, birth date or sex disagree beyond normal formatting differences.
- Order check. Compare accession numbers with the RIS. Flag studies with no accession, an unknown accession, or an accession shared by studies of different patients.
- Merge history. Pull the patient merge history from the EHR or interface engine (ADT merge messages). Flag studies still filed under a retired ID.
- The PACS's own queue. Export whatever the source holds in its QC or exception list.
The output is a list of studies with a reason code for each, which turns a vague worry into a number you can plan around.
Resolving them
Resolve in order of confidence, and keep a record of every change.
- Deterministic fixes first. A retired ID with a known merge to a surviving ID. An accession that exists in the RIS under a single patient. These can be corrected by rule.
- Strong matches next. Combinations of identifiers, birth date, sex, study date and modality that point to exactly one patient and one order. Correct them, and log the evidence for each.
- Everything else goes to a person. Never auto-merge on name alone. Two patients with the same name and birth date is not a rare event in a large health system.
Corrections should change the patient and order attributes, not the image data, and they should be traceable: what was changed, from what, to what, on what evidence. Your HIM or imaging informatics team will want that trail.
Watch the accession numbers. Some archives group studies by accession. Two studies that wrongly share one can be merged into a single study in the target. Clean them before migrating, not after.
What to do with the rest
After reconciliation, a set of studies remains that can't be safely matched. There are three reasonable choices, and most projects use more than one:
| Option | When it fits | Trade-off |
|---|---|---|
| Migrate into a quarantine | The target has a reconciliation workflow, and staff to work the queue | Keeps them in the clinical system, but the queue has to actually get worked |
| Preserve outside the PACS | Studies must be kept for retention, but don't belong in the production archive | Clean production system; studies stay searchable and restorable on demand |
| Leave in the source | The source system stays available read-only for a defined period | Cheapest short term, but the problem comes back when the source is retired |
Preserving them outside the PACS is what DICOMatics Attic is for: historical, orphaned and unmatched studies on storage the hospital owns, searchable, and restorable to the production PACS when someone needs one.
Questions
What is an unmatched study in PACS?
A study whose patient or order information does not match a valid record in the RIS, EHR or master patient index. It may have a wrong or temporary patient ID, a missing or invalid accession number, or demographics that disagree with the patient record. Many PACS hold these in a QC or exception queue; others store them like any other study.
What causes unmatched studies?
Mostly manual entry and downtime: exams acquired without a modality worklist, trauma patients registered under temporary names, studies acquired during a RIS or EHR outage, the wrong patient picked from the worklist, outside studies imported with foreign IDs, and patient merges that never reached the PACS.
Should unmatched studies be migrated?
They should be resolved or deliberately handled, not copied blindly. Migrating them as-is can create duplicate or wrong patient records in the new system. The usual approach is to reconcile what can be matched, quarantine what needs clinical review, and preserve the rest outside the production PACS.
Can unmatched studies be matched automatically?
Partly. Matching on identifiers, accession numbers, merge history and dates can safely resolve a large share. Matching on name alone is not safe. Anything the rules cannot resolve with confidence should go to a person for review.