By Don Dennison

The trend of health provider organization mergers, acquisitions and affiliations to create multi-facility, multi-care setting networks is well documented. As is the adoption of a shared EMR across these enterprises. Such consolidated enterprises are faced with many opportunities to standardize their data, systems, practices and personnel structure to gain clinical and financial benefits. While the standardization of clinical records and systems is driven by the adoption of the shared EMR, imaging systems often lag behind. And past practices commonly used by imaging staff to consolidate data and systems need to be reassessed once a shared EMR is in place. Imaging informatics professionals need to understand the best practices for managing imaging records as part of a complete clinical record, owned by the EMR.

The consolidated enterprise
When several health care facilities and organizations that have operated independently come together to form a single entity, some normalization of data and systems needs to happen in order to gain the benefits of scale and integration. Also, providing authorized enterprise access to data through a common user interface increases productivity and reduces support complexity and costs. For imaging, this can mean the use of a common enterprise viewer embedded in the EMR user interface, and the use of a shared imaging archive that may be provided as part of a shared enterprise PACS or a vendor neutral archive (VNA).



One of the common benefits of a shared EMR is the availability of an enterprise master patient index (EMPI) which establishes a link across locally assigned patient IDs. This is an essential component for providing a longitudinal patient imaging record throughout the enterprise. It does require the use of imaging systems that can receive the EMPI value and have logic to manage a patient’s imaging records that have been acquired in the different facilities under different patient identity domains.

Consolidating imaging records — traditional techniques
Prior to a shared EMR being in place, when data consolidation from disparate facility systems was performed, rudimentary methods were often used to ensure values for different patient information fields were unique. For example, patient ID and accession number. The uniqueness of these and other fields are critical to prevent the records for different patients, acquired in different organizations that may have used a similar numbering scheme, from being incorrectly linked when stored in the same system.

One method used was to prefix (or suffix) values with a string to uniquely represent the source facility. For example, the Patient ID value “12345” became “GH12345” when migrated from General Hospital’s PACS. This prefixing of patient ID values can be largely avoided if the destination system uses the DICOM attribute Issuer of Patient ID value (populated with the assigning authority value from the original ADT source) to keep each patient record managed within its own domain. The use of this attribute is still not supported in many PACS and even some VNAs, unfortunately.

The link between EMR and imaging records
In an enterprise where the EMR is expected to index and be aware of all of a patient’s records, including those stored in imaging systems, coercing the attribute values within imaging records without also updating or linking the values in the EMR can create issues when trying to access the records from the EMR, or its embedded RIS module.

Also, in many cases, the procedure information, which is often persisted in the DICOM attribute Study Description, was not updated as part of the migration, resulting in some PACS features that depend on the body part and other procedure-specific information to experience issues. Two common examples of PACS features that depend on this information are the rules that determine which of a patient’s prior exams are relevant for comparison with the current exam during diagnostic review, and display protocols (aka hanging protocols) that define the automatic layout and
placement of imaging data on the screen.

If the imaging archive (PACS or VNA) sends study content notifications for the migrated data to the EMR so that it can provide users with a link to the enterprise viewer (or PACS client), a variety of challenges may need to be solved.

First, some EMR configurations will not accept unsolicited results, and the study content notification (typically based on a modified HL7 ORU message) will appear as a form of result and cause an exception. In these cases there are two common approaches when migrating the data: an order using the accession number value from each study to be migrated is added to the EMR (often done as a bulk operation); or the study data must be updated with accession number values from the EMR. For EMR configurations that do accept unsolicited results, there are still other potential issues.

As the study content notification will include information on the performed procedure, which is often extracted from the DICOM attribute Study Description, the value presented within the EMR may differ from those used for exams that are ordered from the EMR. This inconsistency of descriptions can cause some user confusion or limit their ability to know which exam to open and view. As the values presented are extracted from DICOM data that was historically managed by a system other than the EMR, and may have been migrated more than once, or imported to the source system without any reconciliation to consistent values, what is presented can be quite different or limited.

Dealing with procedures and body parts
Most PACS and VNAs rely on some form of mapping tables to assign a normalized body part value to an imaging study. Most values coming from the various modalities are unreliable, so the input to the mapping is normally the order. Many PACS do not persist the body part value within a DICOM attribute when the data is migrated (though one called Body Part Examined exists for this purpose). And even if the value was persisted, the defined list of body parts will vary from organization to organization and system to system.

Often an extract of all the procedures from the source organization’s RIS is performed and an experienced resource performs a mapping from the procedure name value (which is typically the value stored in the Study Description) to one of the managed body part values in the destination system.

Shared EMR with separate facility RIS
Often when a shared EMR is implemented, the included RIS module is adopted at each facility as part of the rollout plan. In the case where the facilities continue to operate their own RIS, unless procedure information is normalized to a common set of values, imaging records will continue to contain facility-specific values, making it much more complex to define system rules and configurations based on procedure-level information.

Advice for imaging informatics professionals
When managing patient imaging records in a consolidated enterprise, there are several approaches that are recommended: Before migrating any data or consolidating any facility systems into a shared enterprise, first do a detailed inventory of the data. If the tools are available, this analysis may be performed by qualified staff, but the source system vendor may need to be engaged.

Assess how the source system manages patient IDs and any related domain information. Determine how ordering, acquisition, verification and quality control workflows are managed, as well as any rules to normalize DICOM attribute values. Assess whether the source system has any controls in place to determine imaging record attribute value compliance or uniqueness. Does the system ensure unique DICOM Study, Series and Instance UID values? Does it make use of the Issuer of Patient ID along with Patient ID values to identify the patient within a domain? Does it ensure accession number value uniqueness?

Assume that the current workflows have not always been in place and that some imaging records will have been acquired elsewhere and entered the system without a complete reconciliation of attribute values. Remember, many source systems have experienced many upgrades, have been managed by several different staff and have received information from many external information sources. What is true today may not have always been true.

Perform queries to identify any attribute value outliers. For example, if the patient ID should always be seven numbers, look for any records with a patient ID value containing a non-numeric value, or of a different length. Define and publish an imaging record quality policy. Basing this policy on how your EMR manages patient information, and guidance from IHE on how to map patient and procedure values to DICOM attributes, is highly recommended. Use the policy to measure the consistency and completeness of all new and historic imaging records.

When identifying the shared enterprise imaging system (PACS or VNA), assess its ability to manage multi-facility and multi-patient ID domain data. Look for tools to allow users at each facility to perform data quality corrections on their own data (without affecting data from other facilities). Also, look for tools to automatically detect, and potentially correct, records that do not comply with your organization’s imaging record quality policy. Make any patient-level information changes at the EMR level and use proven methods (for example, an ADT patient merge message from the EMR to the imaging IT system) to propagate and apply the change to the imaging record.

About the author: Don Dennison is an imaging IT industry consultant who serves on SIIM’s board of directors as well as its Program, Hackathon and eCommunity committees, and chairs the ACR Connect and Informatics Industry Activities committees. He is a frequent speaker and panelist on topics such as medical imaging record interoperability and integration of imaging data within the EMR. He has published numerous articles on imaging informatics.