Mayo Clinic Builds Next-Gen Health Information Exchange

Open-source natural language processing software will provide additional context so that clinicians and researchers can better use patient data from a variety of sources.

Ken Terry, Contributor

July 20, 2011

4 Min Read

17 Leading EHR Vendors

17 Leading EHR Vendors

(click image for larger view)
Slideshow: 17 Leading EHR Vendors

Using a Beacon grant provided by the Office of the National Coordinator of Health IT, the Mayo Clinic is building what may be the next generation of health information exchanges (HIE) with a group of healthcare providers in southeast Minnesota. In this real-world demonstration, Mayo will apply the computer tools it is developing through a federal SHARP grant to create new methods of mining electronic health record (EHR) data.

"We're building a patient data repository for southeast Minnesota that will execute on a lot of the promise and principles we're articulating in our SHARP grant," said Christopher Chute, MD, a Mayo Clinic epidemiologist and the principal investigator on Mayo's SHARP grant.

Mayo is using the government's CONNECT software to establish the HIE. The world-famous group practice also is working with the pioneering Indiana Health Information Exchange to build a data repository like the one in Indianapolis. "We'll use that to do population health management, aggregate outcomes analysis, and comparative effectiveness research pooled across healthcare encounters by different providers with the same patients," said Chute.

What's different about Mayo's approach is that it will use natural language processing to identify terms in disparate EHRs so that they can be mapped to a normative terminology such as SNOMED CT. This "semantic interoperability" will make it possible to create databases that transcend the differences between the vocabularies in different EHRs.

In contrast, Chute said, alternative HIE approaches such as that found in Microsoft's Amalga product simply provide a way of viewing data in disparate systems. "The Amalga tool pays very little attention to data normalization," Chute said. "It basically provides common views of the information and facilitates broad-scale review across multiple patients."

The researchers are using Mayo's open-source natural language processing software within a framework established by the IBM Watson Center to provide additional context so that clinicians and researchers can better use patient data from a variety of sources. The Watson supercomputer, which made headlines when it beat human contestants at "Jeopardy," uses the same UIMA (unstructured information management architecture) framework. So far, the Mayo scientists have applied their new tools to records of 30 patients with diabetes at Intermountain Healthcare and about 10,000 patients from the Mayo Clinic. These tests demonstrated their ability to run through the process from end to end, said Chute, but did not show whether they could normalize data from different EHRs. That's one of the goals of the Beacon demonstration project. The three legs of the SHARP stool, Chute said, include automating aspects of terminology mapping to a standard lexicon and "syntactic normalization" of defective HL7 or Continuity of Care Document (CCD) messages. Put another way: The new system can repair ill-formed HL7 or CCD messages by analyzing the messages to determine what's wrong with the syntax of the message and correct it so that its content can be understood. The third leg of the stool is natural language processing, which will be applied to clinical notes containing data that doesn't show up in administrative codes, lab results, or medication lists. Analyzing signs and symptoms, for instance, can make the categorization of diagnoses much more accurate, he said. In the long run, Chute predicted, the tools Mayo is developing might be useful in clinical trials--for example, in identifying candidates for studies--and in triggering clinical decision support. Someday, he added, it might be possible to apply natural language processing to physician dictation, so that terms could go into discrete fields in an EHR. While much research must be done before that happens, he noted that the current mode of documentation in EHRs tends to slow doctors down. "I think we'll end up with a hybrid of some structured data, augmented by a large amount of dictated or textual data. And I hope we can leverage the content of that textual data in a reliable and efficient way." Mayo's SHARP tools and its natural language processing program are open source, noted Chute, and any developer can get a free license. Commercial applications may be based on these tools, but Mayo isn't looking to profit from them, he said. "We're not a software company. We're not trying to make money off this environment, we're trying to help our patients." In the new, all-digital InformationWeek Healthcare: iPads are leading a new wave of devices into the exam room. Are security, tech support, and infection control up to the task? Download it now. (Free registration required.)

About the Author(s)

Ken Terry


Ken Terry is a freelance healthcare writer, specializing in health IT. A former technology editor of Medical Economics Magazine, he is also the author of the book Rx For Healthcare Reform.

Never Miss a Beat: Get a snapshot of the issues affecting the IT industry straight to your inbox.

You May Also Like

More Insights