PACS and DICOM Radiology Integration: A Hospital IT Guide

Modality worklist lets the technologist select a scheduled study on the modality rather than typing patient details at the console. This is not a convenience feature. Manual entry at the modality is the single largest source of patient identity errors in radiology, because a mistyped name or a wrong identifier produces a study that lands in PACS unmatched or, worse, attached to the wrong patient.

The worklist query should return enough context for the technologist to confirm they have the right patient and the right examination: identifiers, name, the requested procedure, the accession number, and the referring clinician. Returning too broad a list encourages selection error, so the query filters — by modality, by station, by date — matter as much as the connection itself.

Emergency and unidentified patients need a defined path. Trauma cases arriving without registration cannot wait for a worklist entry, so the laboratory of record needs a temporary identity convention and a merge procedure for reconciling the study once the patient is registered. Unplanned merges are inevitable; unplanned merges without a procedure are how duplicate records propagate.

When a modality sends images to PACS, the transfer succeeding is not the same as the images being safely retained. Storage commitment is the DICOM mechanism by which the archive explicitly confirms that it has taken responsibility for the objects, which is what allows the modality to clear local storage. Without it, a modality either fills up or deletes on an assumption.

Hospitals discover this the hard way when a modality's local disk is cleared after a transfer that PACS did not fully persist. Configuring storage commitment, and verifying that the modality actually acts on the response rather than ignoring it, is a specific commissioning test worth doing rather than assuming.

The related check is object completeness. A study can arrive with fewer series or images than were acquired, and nothing in a simple transfer confirms the count. Reconciling expected against received image counts, and alerting on incomplete studies, catches transfer truncation before a radiologist reports on a partial study.

The administrative side runs on HL7 messaging: an order message when a study is requested, status updates as it is scheduled and performed, and a result message carrying the radiologist's report back to the clinical record. Each message type needs an agreed structure and, more importantly, agreed behaviour on the edge cases — cancellation, amendment, and a report issued preliminary then finalised.

Preliminary and final reports are the most commonly mishandled case. A preliminary read issued overnight and a final report issued the next morning must both reach the record, with the final clearly superseding the preliminary rather than appearing as a second independent report. Clinicians acting on a superseded preliminary because both are visible without hierarchy is a real patient safety exposure.

Newer integrations increasingly use HL7 FHIR R4 resources alongside or instead of v2 messaging, particularly where the hospital is aligning with ABDM and needs radiology reports to participate in a shareable health record. The transport choice does not change the workflow questions; it changes the representation. Either way, the report must be identifiable, attributable, and versioned.

The accession number is the join key between the order, the study, and the report. If the modality writes an accession that differs from the one in the order — because it was typed, truncated, or formatted differently — the study will not match, and the images will sit in PACS unattached to any request. This is the most common single defect in new integrations.

The prevention is that the accession must flow from the order into the worklist and from the worklist into the DICOM header without human retyping. The detection is a reconciliation report that lists studies without matching orders and orders without matching studies, reviewed daily during stabilisation and weekly thereafter.

Someone has to own that unmatched queue. Left unowned, it accumulates silently, and each entry represents either a patient whose images cannot be found from the clinical record or a request that appears never to have been performed. HealUDoc dashboards can surface unmatched radiology studies alongside pending laboratory orders so that a single operations review covers both.

Clinicians need images where they are already working, which means launching a viewer from the patient's record in context rather than logging into a separate PACS client and searching. Context launch passes the patient and study identity to the viewer, so the clinician lands on the right study without a second search — and without the chance of opening the wrong patient.

Zero-footprint browser viewers have made this substantially easier than the plugin era, and they matter especially for referring clinicians in OPD and for consultants reviewing remotely. Diagnostic reporting still requires calibrated displays and the full reporting workstation; clinical review does not, and conflating the two requirements makes access harder than it needs to be.

Access control needs to be as considered as availability. Image access should follow the same role and treatment-relationship rules as the rest of the record, and the access log needs to cover image viewing as well. Under the DPDP Act 2023, imaging is personal health data like any other, and a viewer that bypasses the record's access controls is a genuine compliance gap rather than a technicality.

Imaging storage grows faster than most hospital IT plans anticipate, driven by thin-slice CT, tomosynthesis, and increasing study volumes rather than by patient numbers. A capacity plan should be built on measured growth per modality rather than on a headline figure, and it should account for the fact that retention obligations for imaging are long.

Tiering keeps this affordable: recent studies on fast storage for immediate retrieval, older studies on slower or cloud storage with an accepted retrieval delay, and a documented policy for what moves when. The clinical requirement to state is retrieval time for a prior study during an active consultation, because that is the constraint that determines the tier boundary.

Assume you will migrate archives at least once. Vendor-neutral archive design, standards-conformant storage, and a documented understanding of how studies would be exported are what make that migration a project rather than a crisis. The time to establish an exit path is during procurement, not when the incumbent contract is expiring.