AMIA: Regenstrief pumps up its clinical decision support

'We're really just beginning to explore the possibilities of NLP'
By Neil Versel
09:27 AM

Regenstrief Institute was a pioneer in clinical decision support, but its system was aging. Most noticeably, the interface in its famed "Gopher" computerized physician order entry system looked the same in 2010 as it did in 1984.

After two years of redesign, the venerable Indianapolis institution unveiled an update last year with more than just a modern look, Regenstrief researchers pointed out Monday at the annual American Medical Informatics Association (AMIA) symposium.

The new Gopher features a graphical user interface that looks like a single, integrated system, but the complex display is generated from multiple layers and plug-ins, research scientist Jon Duke, MD, said. More importantly, it supports a new, multifaceted approach to clinical decision support, featuring order entry, research and results viewing capabilities.

"We recognized that improving quality of care was a priority," Duke explained. "We wanted to promote provider efficiency."

Duke and two Regenstrief colleagues touted five advances in clinical decision support the new Gopher provides over the original:

  • Dynamic alerting, or the mechanics to change alerts based on clinical relevance to a specific patient's needs. Programmers built in 10 theoretical alerting levels, though Duke said the informatics team decided to activate four alerting zones on the screen for now. "Every alert in our system … had a baseline relevance level which determines its display location," Duke said.
  • A "relevance adjustment module" that makes particular alerts "non-interruptive" for certain specialities to help avoid alert fatigue, according to Duke. It also aggregates alerts, providing multiple alerts simultaneously, including multimedia alerts. As Duke explained, Gopher is a "learning" system that tracks each user's behavioral history and can, for example, show if a physician has ignored a specific alert on a specific patient before and ask if the alert should appear in a sidebar on the screen in the future.
  • Chart search, which senior Regenstrief informaticist Douglas K. Martin, MD, called a Google-like search function within the chart. The search is capable of returning information regardless of whether it was in an unstructured note or structured lab result. A simple search takes less than a second, but Martin said the informatics team faced a speed challenge from having 12 million unique patients in the EHR database. Regenstrief decided on just-in-time indexing per patient rather than trying to index the entire, massive repository at once. To speed up workflow, a batch processor can cross-check the appointment file each day to "pre-build" some indexes for patients expected in the office, Martin said.
  • InfoPanel, a communications system for instant messaging between clinicians and displaying whether orders have been fulfilled. "We have mechanisms to show when other people are in the same chart," Duke said. This helps prevent clinicians from signing out of charts without reading new orders others have entered.
  • Recipe authoring, which Duke called an "update on the classic idea of rules authoring," based on If This Then That (IFFTT), a service that links triggers and actions. Each alert has triggers; for example, when a doctor selects a drug, this brings up a series of actions that follow. “You can create what we call recipes and have them go directly into our system," as an element of customization, Duke said.

The recipe authoring is only about 30 percent built so far, Duke said, but the entire project is a work in progress.

Gopher is beginning to incorporate natural language processing (NLP) into the search function, including the ability to analyze physician notes in real time. "We're really just beginning to explore the possibilities of NLP," Duke said.

The next step is adding more logic to CDS-NLP connection, and Duke promised some studies on the efficacy of NLP in the next six months or so.

[See also: AMIA airs clinical decision support concerns to FDA]