11 minute read

Action in Digital Health and Clinical Radiology

The College is finalising a white paper on digital health and clinical radiology. This sets out key steps in the application of digital health to clinical radiology to improve the workflow for radiology practices, and the outcomes for patients. Already, there has been action on some of the key proposed steps, building on earlier work by the College and others.

At the request of the Federal government, the College and the Australian Diagnostic Imaging Association (ADIA) have submitted a report on the terminologies used for the identification of radiological procedures, with recommendations on how to adopt a standardised “Radiology Referral Set” with minimum disruption to existing systems.

The Federal Budget handed down in May 2021 included provision of $7.2 million for work on electronic referrals in radiology. This will build on earlier work by the College on the information content required in referrals, and on the College’s position paper “Building eReferral: Safety and Patient Choice Position Statement.” The original version of the position paper was prepared in 2018, and this is currently being reviewed in the light of recent experience (especially since the outbreak of the COVID-19 pandemic, and the widespread adoption of ‘telehealth’ measures). There will continue to be a strong emphasis on the importance of patient choice of provider. It will also be important to ensure that this great opportunity to improve the quality of referral information is not lost by merely replicating existing paper-based processes.

The Australian Digital Health Agency (ADHA) has been working with the providers of secure messaging in health care to define standards and profiles that will allow secure messaging to be widely available in health care, with seamless access to the clients of other providers, rather than access limited to each provider’s silo. Secure messaging will play a key role in safeguarding patient privacy and confidentiality as digital data exchange becomes ubiquitous, avoiding the potential hazards of email and social media. The Agency has been trialling prototype secure messaging solutions with several of the leading messaging suppliers this year.

Much consideration has been given to strategies for improving access to images from previous studies. The best solutions here will build on work from other key steps in the white paper (the Radiology Referral Set and eReferral), and will also be influenced by the infrastructure put in place by State jurisdictions. That said, the white paper recognises that the steps need to be worked on concurrently as much as possible, both to optimise the interplay between them, and to minimise delays.

Radiology Referral Set: a Key Tool for our Digital Future

New digital technologies and applications are in the process of revolutionising many fields of activity, including health care and, specifically, radiology. In planning and preparing for this, the College has identified some key tools that will be needed to help improve our processes and outcomes.

Interoperability

A plethora of new applications in health are being launched by many different groups, and it is critical that the data from one application can be read and treated appropriately by the next application in the chain. This requires the systems and applications to be not just interoperable (where the output from one application can be read and used by another), but semantically interoperable (where each system attributes the same “meaning” to a given output).

In radiology, one of the key data elements is the name of the imaging procedure or test that is to be done. We are all familiar with the list of exam names that we use in our own practices—“of course” a “CT brain” means a CT scan of the brain. But when we look to develop applications that will handle procedures done at many different practices, we find that everyone has their own list, and their own “obvious” names for tests.

It is critical that test names be interoperable—a referral system must know that the practice RIS will implement the same procedure that the referrer entered into the referral system; a PACS system must recognise that images of test “x”, requested by an image retrieval system, correspond to images obtained as part of tests labelled as “w” in its own system.

Why do we need (another) RRS ?

When the American College of Radiology was establishing its Dose Index Registry for CT, they found that data being submitted for CT scans of the chest could have any of more than 20 site-specific descriptors; a similar range was found for CT scans of the brain (CT head, CT cranium, head CT, cerebral CT, Computed Tomographic scan of the head, etc., etc.).

Traditionally, the reconciliation of names from different systems has been done by a human—often one in the booking office, with or without assistance from senior radiographers and radiologists.

If we are to capture the efficiencies of digital techniques (which are needed to cope with the ever-rising volume of work), the various software systems need to be able to achieve this semantic interoperability with as little human intervention as possible.

This can be done by developing a standard set of terms for imaging procedures, each with a clearly defined meaning—an X-ray foot is one procedure, an X-ray ankle another, and an X-ray foot and ankle could be a simple addition of the first two, or a third procedure, in which, say, only a single lateral projection is obtained.

This standard set of terms is sometimes called an “orderables catalogue” or an “examination catalogue”; in order to emphasis the consultative nature of our profession we have preferred the term “Radiology Referral Set” (RRS)—the set of procedures that can be requested from radiology practices.

Any system referencing an element of the RRS must be able to rely on any other system attributing the same meaning to that element.

Therefore, the vision is that systems sending information about tests will always reference terms from the RRS. This does not mean that every existing term has to be converted to a new term from the RRS, with all the legacy terms discarded, extensive changes to existing systems, and the associated disruption and cost. What it does mean is that there will need to be mappings from existing catalogues to the RRS, so that the application can accurately convey what is intended. This could be (but does not have to be) entirely ‘’under the hood”: a user could enter the same legacy term they always have; the application would then map it to an appropriate term in the RRS, and send that to the next application (with or without the original “legacy” term).

Over time, it might be expected that existing systems would adopt the standardised RRS for simpler operations, and that new systems might use the RRS from the outset.

Priority Interoperability areas for action based on workstreams agreed with CIOs

Choosing an RRS

Once it is accepted that a standardised Radiology Referral Set would be a valuable tool for interoperability in radiology, the obvious next question is how to choose or produce such a set.

To this end, the College and the Australian Diagnostic Imaging Association (ADIA) have been reviewing the existing terminologies for radiology examinations, with a view to making recommendations to the Commonwealth Department of Health. The Australian Digital Health Agency (ADHA) has also been involved. The project has been run on a tight timeline, in order to ensure that its results will be available for the next project in the proposed pipeline, which will look at electronic referral processes in radiology. At this stage, the scope of the project has been limited to that of Category 5 (the diagnostic imaging section) of the Medicare Benefits Schedule, but it is expected that the chosen RRS will be expanded over time to cover all diagnostic imaging procedures.

The RRS project has involved a review of existing terminologies for radiological examinations, a practice survey of terminologies in current use, and the construction of some selection criteria, based on feedback from the survey, and on the logical requirements of the applications that may make use of the RRS (electronic referral, image retrieval, clinical decision support, etc.).

A key requirement of an RRS is that its terms be logically related to each other, so that applications using the RRS can make valid inferences about them. It should be possible for an application to recognise that a “CT Pulmonary Angiogram” is one of many examinations that will image the chest; and that it is also a kind of CT examination (which might be important for, say, applications concerned with radiation doses, or one managing timeslots on a practice’s imaging devices). This is done by structuring the terms in the RRS in a hierarchical fashion, specifying which term is a special case of which broader term. Terminologies that have a logical structure of this sort are known as “ontologies”. For an RRS to best contribute to the efficiency gains offered by automation of radiology workflow, it needs to have the structure of an ontology.

The Options for an RRS

Most “orderable catalogues”/request sets currently in use (including the Medicare Benefits Schedule) are simply lists of terms, with no defined relationships to each other (even though it is usually possible to at least identify the relevant imaging modality from such terms). The Radiological Society of North America (RSNA) recognised the need for a more structured and standardised vocabulary in radiology in the 1990s.

After a review of the then-existing general medical terminologies, the RSNA decided that none of them adequately supported the needs of radiology, and commenced on the development of the “RadLex” set of terms for radiology. This is not limited to names for examinations—it includes many other terms relevant to radiological procedures, anatomy, imaging findings, and radiological diagnosis. Much of this work was originally done by expert consensus, however the need for a logical structure was later recognised, with RadLex restructured and then formally recognised as an ontology. A subset of terms specifically used for naming procedures was developed, and referred to as the “RadLex Playbook”—a structured list of radiological procedures (or “plays”, in the American sporting sense).

Recently, the US National Institutes of Health (NIH) sponsored a mapping project between the RadLex Playbook and one of the more general structured medical terminologies, LOINC (Logical Observations, Identifiers, Names and Codes—widely used in some pathology applications).

Many other medical specialties have recognised the need for controlled terminologies, in which the terms have strictly defined meanings and logical inter-relationships. Pathology has been a pioneer in this field, and was the original source of SNOMED (Systematized NOmenclature of MEDicine), which has since been broadened into “SNOMEDCT” (SNOMED-Clinical Terms), by the incorporation of clinical terms from a British terminology, the Read codes. SNOMED-CT is now the dominant general medical terminology worldwide, and was adopted as a national standard by Australia (on advice from the then National E-Health Transition Authority) in 2005. The Australian “localisation” of SNOMED-CT is managed by the National Clinical Terminology Service, an arm of the ADHA, under license from the International Health Terminology Standards Development Organisation (IHTSDO).

Next steps

The RRS project has completed its review of the existing terminologies that could be used as the basis for an RRS, and submitted its report to the Federal Government at the end of June. It was clear that adapting an existing set of terms would involve much less effort than attempting to build a new one from scratch. The requirement that the standardised RRS have the structure of an ontology sharply limits the number of possible candidates.

It was noted that existing terminologies for procedures, such as that in the Medicare Benefits Schedule, and the various ad hoc lists developed inhouse, are deeply embedded in current workflows, and that these will continue to have key roles in practice. In other medical fields, the introduction of standardised terminologies has been gradual, and often behind the scenes, with little change to the terms employed by clinical users. In some cases, an application may present the result of its mapping from the legacy term submitted to a proposed standardised term, and seek user confirmation that the mapping has been appropriate. Synonyms for, or definitions of, the standardised terms can be provided, to make this easier for the user.

There will be a substantial amount of work required to map the existing terms to whichever new set of terms is adopted. An appropriate funding model for this will be required, along with access to suitable software to assist the mapping process (purpose-designed software tools are available for all of the major candidates).

It is also well recognised that no terminology is ever perfect, that new terms will need to be introduced, and later refined, and old terms retired (or even deprecated). Such maintenance of a terminology is a substantial task, for which ongoing resources and expert support will be required.

The adoption of a standardised RRS will allow the unambiguous identification of any radiological procedure. This will help to ensure that the test performed is the one that best matches the patient’s clinical need. It will also help to enable the accurate retrieval of the specific test report and/or images whenever needed. The RRS will thus be a key foundation for e-Referral, image exchange, and clinical decision support.

Dr Nick Ferris

Digital Health Working Group

Chair, eHealth Reference Group

This article is from: