Interpreting interpretations: How a portal can bridge knowledge gaps with patients

Twitter icon
Facebook icon
LinkedIn icon
e-mail icon
Google icon
 - woman at computer

Patients are increasingly being offered access to their imaging studies through the use of patient portals, but when the level of access outpaces a patient’s ability to comprehend the information, problems can arise. To avoid this issue, researchers at the University of California—Los Angeles (UCLA) have designed a portal that tailors information in a report for a layperson.

Corey W. Arnold, PhD, of UCLA, and colleagues brought their design to the Radiological Society of North America 2011 annual meeting and shared the feedback they received in an article published June 5 in the Journal of the American Medical Informatics Association.

“Even though radiology test results are one of the most difficult portions of the clinical record for lay people to understand, they are one of the most frequently accessed pieces of information via patient portals when available,” wrote the authors. “This suggests the need for new methods of sharing radiology information with patients.”

The solution proposed by Arnold and colleagues was to add an “interpretive layer” to the patient portal that took clinical information and transformed it into consumer-centric explanations. This was accomplished through the use of a natural language processor (NLP) which can read interpretations and extract concepts.

In development of the system, which was focused on neuroradiology studies, nearly 3,000 brain MRI reports from 277 patients were processed, resulting in 448 unique concepts. After a manual reconciliation process, the researchers culled the list down to 52 terms that make up the possible list of salient findings.

When subsequent studies are added to the portal, the NLP scans the conclusion section of the neuroradiology report and, using the superset of possible salient findings, displays pertinent information to the patient. This information was written by the research team as a reflection of how clarifications of findings would be provided to patients in real life. These explanatory layers of information are added over the clinical data, with the actual findings being highlighted in the displayed images.

While Arnold and colleagues received support for the effort to educate patients through modified informatics tools, they noted lingering concerns that patients may misunderstand results. The researchers also designed the system to omit “negated” findings—such as previous masses that have since been removed—but some errors persisted where these findings were still selected despite the fact they cannot appear in the images.

Because of the potential for misunderstanding, some suggested the portal application could be used during an office visit to allow for patients to view the content in the presence of a practitioner who can answer questions. Arnold and colleagues said they are taking the critiques under consideration and are refining both the performance of the NLP model and the level of readability of clinician-generated definitions.

“There are, and will always be, risks in allowing patients direct access with their records, but evidence indicates that these potential risks are outweighed by the benefits provided by such a system to an engaged patient,” wrote the authors.