Interoperability Initiative Adds Structure To Clinical Text

Industry-wide alliance Health Story aims to use natural-language
Health Data Security: Tips And Tools
Health Data Security: Tips And Tools
(click image for larger view and for slideshow)
Unstructured documents in electronic health records--including dictated notes, reports, and other kinds of free text--can't be used for clinical decision support or quality reporting, but they are nonetheless important to physicians who are treating a patient, especially if doctors don't know that patient's history. If the documents could be more structured, without requiring physicians to use point-and-click templates to enter the data, they would yield valuable information.

Now an industry-wide initiative known as Health Story has shown how to approach this issue in a new way by using natural-language processing. Health Story, an alliance of healthcare vendors, providers and associations, not only outlines a way to add structure to unstructured documents, but says the new approach can also improve communication between disparate EHRs used in different organizations.

Health Story has developed implementation guides for nine clinical document types that conform to the Clinical Document Architecture (CDA) of the data standards organization Health Level Seven (HL7), and has an associate charter agreement with HL7. The XML-based CDA documents--collectively known as the Consolidated CDA--include the revised Continuity-of-Care Document (CCD) that the Center for Medicare and Medicare Services (CMS) has proposed as the sole format for the clinical summaries that must be exchanged during transitions of care under its Meaningful Use Stage 2 rules.

[ Which healthcare organizations came out ahead in the IW500 competition? See 10 Healthcare IT Innovators: InformationWeek 500. ]

The other Consolidated CDA document types are consultation note, diagnostic imaging report, discharge summary, history and physical, operative note, procedure note, progress notes, and unstructured documents. All of these except the last one contain structured elements.

Although it might seem strange to include "unstructured documents" as a CDA document type, a scanned paper document could be given a CDA header and sent from one EHR to another. That puts the information in the hands of clinicians who might really need to see a discharge summary or a consultant report, said Robert Dolin, MD, leader of Health Story and vice chair of HL7, in an interview with InformationWeek Healthcare.

Dolin views the process of adding structure to the free text in CDA documents as incremental. "There are about 1.2 billion clinical documents out there, and it's relatively easy to turn those into minimally structured documents," he said. "If you do that, the next year you can add a little more structure. With that strategy, you embrace the front-line clinicians right from the get-go, and you minimize the disruption to clinical workflow. You get all of this data flowing."

To create structured, coded data from free text, some of the companies participating in Health Story are using natural language processing (NLP) software from M*Modal, a Medquist subsidiary; Nuance; and A-Life, a division of Optum. This is the same kind of approach that some EHR vendors are using to extract data from dictation transcribed by speech-recognition programs.

Dolin admitted that "there's a known error rate with natural language processing." For example, the software might pick up a mention of a heart condition in the family history section of a note and mistakenly think that should be added to the patient's problem list.

"My belief is that when you're using NLP, the folks who are receiving the CDA documents need to be aware that that's where the structure came from," he said. After that, they can decide how they want to use the information. But clinician oversight is essential, he added. Terms in CDA documents should not populate EHR fields automatically, but a physician might add them to the appropriate fields.

To get provider adoption, Health Story will have to persuade EHR vendors to build the CDA document types into their products. It already has a big leg up because EHRs certified for stage 1 Meaningful Use are already capable of exchanging CCDs, Dolin noted. Health Story recently released what it calls a "green CDA" toolkit to aid developers in creating other CDA documents.

Health Story developed its implementation guides in cooperation with HL7 and Integrating the Healthcare Enterprise (IHE). Other organizational affiliates include the American College of Physicians (ACP), the America Health Information Management Association (AHIMA), the Association of Healthcare Documentation Integrity (AHDI), the Clinical Documentation Industry Association (CDIA), and the Health Information Management and Systems Society (HIMSS).

Vendor members of Health Story include Fujitsu, Inofile, Lantana Consulting Group, M*Modal, Nuance, Optum, and Verizon. Altogether, more than 300 organizations have been involved in the effort.

The Office of the National Coordinator of Health IT (ONC) also helped Health Story by integrating the initiative into its Standards and Interoperability Framework wiki. In a press conference at the recent HIMSS meeting in Las Vegas, Doug Fridsma, director of ONC's office of standards and inoperability, said, "Health Story is one of the first success stories in the interoperability of simple documents. It is a critical piece for maintaining what is the structure of the data we're trying to convey."

Healthcare providers must collect all sorts of performance data to meet emerging standards. The new Pay For Performance issue of InformationWeek Healthcare delves into the huge task ahead. Also in this issue: Why personal health records have flopped. (Free registration required.)

Editor's Choice
Brian T. Horowitz, Contributing Reporter
Samuel Greengard, Contributing Reporter
Nathan Eddy, Freelance Writer
Brandon Taylor, Digital Editorial Program Manager
Jessica Davis, Senior Editor
Cynthia Harvey, Freelance Journalist, InformationWeek
Sara Peters, Editor-in-Chief, InformationWeek / Network Computing