Hailed as heir-apparent to the Web throne now occupied by HTML (HyperText Markup Language), XML (Extensible Markup Language) promises a standardized way for systems to talk to each other--without sacrificing extensibility (a major drawback to HTML). Think of it as a dialect related to HTML, within the mother-tongue of Structured General Markup Language (SGML). And since it’s vendor and platform-independent, XML is a dialect with immediate opportunities.
Businesses will soon deploy all kinds of applications using XML to move structured data around on intranets and the Internet this year, predicts Jon Udell, BYTE magazine’s executive editor, new media. Many vendors are actively harnessing the technology into electronic data interchange solutions under such ventures as Open Financial Exchange, where the World Wide Web Consortium recommended XML standard is the basis for Internet commerce. In healthcare, it may become the "open sesame" on patient health information now scattered throughout the care delivery enterprise.
Medical records killer app?
SGML/XML technology brings together two very important issues under one standard: a Web-speak data-exchange format and industry specificity. If SGML/XML technology delivers on its promises, it will revolutionize the Web and related Web-based applications--and just may enable the development of killer apps for electronic medical records.
Because SGML/XML applications conform to machine processable data structure--as opposed to most current applications that dictate data structure--the technology can facilitate the exchange of intelligible medical record information to caregivers anywhere in the system, says John Spinosa, MD, PhD, director of pathology, Central Laboratory, Pathology Medical Group, La Jolla, Calif.
With XML and SGML underlying technology, users can search for information in ways that take advantage of narrative text, pieces of text or any other data bits on the fly. Tags within the SGML document identify data, effectively making the document a database.
To facilitate the care delivery process while respecting the privacy rights of the patient, "a minimum data set must belong to the patient to accompany that patient through the network of health services," says Spinosa. Patient information should not be locked up by who or what created it, he says. All commercially available electronic medical record applications do a great job--so long as the product controls the entire data entry, storage and retrieval operations. Once the data is exported, the information transfers as an ASCII file with all the intelligence and lookup tables residing outsideof the ASCII data file. He calls that a "gotcha" both for the patient and for legitimate government interests. Information inaccessibility at the local health delivery level results in inconveniences and inefficiencies for the patient and provider; at the national level, such data barriers stymie public health efforts.
If it is possible to exchange information easily and quickly, there is no need to have all the information in one place. Under SGML/XML, the system only has to manage where the information is--the metadata--not the information itself.
If the caregiver can readily access the information, the issue of the data storage site becomes much less important. Spinosa suggests this could lead to models of cooperative business arrangements to finance data storage and transfer operations.
Although security is a hot button in electronic medical record issues, the issue of data ownership and management is fast becoming important. Related (but not restricted to privacy), ownership is also about maintaining control of personal information. Spinosa agrees with Tom Lincoln, MD, Rand Corp., Santa Monica, Calif., who says, "The medical record is owned by the patient, and the care providers have fiduciary responsibility to take care of it." In that line, he does not believe that patients want their health records made accessible in their entirety to all providers at all times. XML, he thinks, will add a layer of anonymity that will allow patients to maintain a sense of ownership.
Rethinking the paradigm
Throwing technology solutions at a cultural problem doesn’t resolve the medical records problem, says Anil Sehti, chief technology officer, Sequoia Software, Inc., Columbia, Md. Given a flexible database and direct data capture by the physician using a keyboard, many medical records storage and access problems could have been accomplished on any hardware platform under any operating system long ago--if technology alone were the answer. For him, XML overcomes many of the difficulties surrounding patient medical records, while supplying a means for legacy systems to achieve compliance with emerging standards. The medical record problem is entirely too complex for conventional technologies, he continues, suggesting that if decision makers disregard the power of SGML and XML, they may end up waiting for products that cannot be built.
The Kona Architecture
As a means to demonstrate the utility of using SGML for the exchange of clinical information, a privately-funded group of vendors, providers, physicians and SGML experts met last summer to develop a proposal for the exchange of electronic health records and documents, and to jump-start the standards process. The group constructed the Kona Architecture, based on SGML, as a guideline for an early draft submission. Subsequently, a Health Level Seven (HL7) Special Interest Group (SIG) formally adopted Kona as a working proposal. Plans are to bring a fully implementable proposal to HL7 for balloting.
Although SIG members agree that much work remains to be done, committee member Liora Alshuler, writer, consultant and author of the book ABCD… SGML, hopes that the proposal will be ready for balloting this year. It is extremely ambitious, she says, but it is possible: "Today, proposals that do not move forward with Web-speed fall by the wayside. If standards organizations don’t pull together and fast-track their development efforts, the need for exchange mechanisms is so strong and the opportunities so great that private interests will step into the void and bring a standard, that, for better or worse, will become a de facto standard."
SGML/XML has also been considered as a dynamic mechanism for claims attachments--nonstandard reports created in response to third party insurer queries seeking additional information.
The HL7 SIG for claims attachments is developing mechanisms, as charged, within the context of X.12 and HL7--where, says Clement J. McDonald, MD, existing balloted and ANSII-approved mechanisms appear to be adequate. "The deadline is fierce," says McDonald, distinguished professor of medicine at Indiana University School of Medicine and director of the Regenstrief Institute, both in Indianapolis. "Choosing a new approach that has not been defined in detail or gone through the standards consensus process will only complicate and delay the process."
This does not necessarily rule out use of the technology for claims attachments, he says. Any of the messaging syntaxes could be interconverted as long as the underlying model is the same. It is likely that SGML/XML will be an accepted syntax within HL7 within the next year or so. At that time the work in HL7 can be converted to SGML with the appropriate tag definitions and HL7 message models.
Cutting healthcare keys
One of the strengths of XML is that an industry can define its own information requirements with tag sets--technically, document type definition. It is also a holdup. "In the long run, SGML/XML will provide solutions to many problems, but it is not a silver bullet," notes McDonald.
At present, healthcare tag sets have not been defined or standardized--nor has the architecture--but Spinosa notes that upfront investments are the cost of commonality. Once defined, he points out, modifications and extensions are much easier and more powerful than the current message-based approaches. Most industry leaders look to the HL7 SGML/SIG to set the standard and create the first cut of a tag set to define the patient encounter.
Under a multimillion-dollar grant awarded by the U.S. Commerce Department National Institute of Standards and Technology, Sequoia Software, Columbia, Md., has undertaken a project to develop a national master patient index (MPI) across a national computer backbone using SGML/XML.
The MPI will correlate and cross-reference computerized patient health records from multiple sites and will generate a summary of patient encounters that resembles a search engine "hit list." Each "hit" actively links to a specific clinical encounter independent of the storage site.
Acknowledging that medical specialties have very specific and usually well-defined information requirements, Sequoia and its partners are currently working in two specialty areas: radiology and anesthesiology. In fact, in December at the Radiology Society of North America conference in Chicago, software developers Sequoia Software and Kurzweil AI, Waltham, Mass., along with primary author for the new DICOM (Digital Imaging and Communications in Medicine) imaging standard, Dean Bidgood of Duke University, Durham, N.C., demonstrated a new use of SGML/XML technology with DICOM to generate a structured report alongside a radiological image.
The transition from legacy systems to new technology environments continues to be an issue for most healthcare officers. Sehti thinks one advantage of SGML/XML is that the technology is centered on the Internet. "It may be possible," he suggests, "to completely bypass the costly client/server technology stage and jump directly to Internet technology."
Votes still out
No technology can have it all, but who’s talking about the downside of XML? Not many. Those not shouting and waving tend to be standing on the sidelines… watching. BYTE’s Udell sees few obstacles to using XML in server-based applications, but warns that its use in client-based applications will be severely hampered by incompatible Microsoft and Netscape implementations of Dynamic HyperText Markup Language (DHTML) and browser document object models.
With XML, it is possible to have all the benefits of Java in a browser, says Sehti, noting that "Microsoft sees XML as a Java killer." For medical records applications, the power of Java may not be necessary for data transfers when easy-to-use SGML/XML can fit the bill, he says.
Although XML is still a new standard, a number of applications are likely to bring it into the mainstream. Metadata applications, considered the most logical methods to share and describe data, have the Resource Description Framework leading the list. In addition, a number of industry-specific applications are under way, particularly from chemists, who have the Chemical Markup Language and mathematicians, with the Mathematical Markup Language. Another very promising area for development is e-commerce and Electronic Data Interchange.
Though SGML/XML may not be the final answer, its flexibility and portability may be a valuable new tool in solving data-sharing dilemmas for all those patient records "locked up" in legacy systems throughout the enterprise.
Charlene Marietti is senior technology writer at Healthcare Informatics.
