Tech that translates doctors' orders

    Natural language processing helps convert physicians' verbal instructions into electronic records.

    By Anna Kattan

    Everyone seems to think electronic health records are a great idea: going digital cuts down on paper storage, makes it easier to transfer patient information from one provider to another, and ultimately will enable the medical industry to create immense clinical databases. Electronic health records are a key component of the Obama Administration's stimulus plan, and dozens of corporations claim they are lined up to create more jobs when the government releases funds for digitizing medical files and other improvements to healthcare information technology.

    Who could quibble with that?

    One group that's not so keen on going paperless is physicians, who often find the current generation of electronic records rigid and counterintuitive. Most doctors prefer to type detailed clinical notes, rather than codes that identify a diagnosis or procedure.

    So how can technology bridge this gap between physicians' preferences and the need to move healthcare into the Internet era? One solution is natural language processing.

    Natural language processing is technology that helps computers understand ordinary human languages like English, German or Arabic. Essentially it takes unstructured, narrative text and transforms it into structured, coded data. That data can then be analyzed by computers and categorized in ways meaningful to researchers.

    "Natural language processing breaks down the structure of the sentence so it can literally read it like a human, understand and comprehend it, and then provide feedback," says David Byrd, a sales and marketing executive at Language and Computing, a privately-held natural language processing company.

    Indeed, scientists have been working for decades to improve machines' ability to comprehend language. Natural language processing was first developed out of a synergy between computer science and linguistics. Since then, the technology has been employed in search engines, spellcheckers, databases and speech recognition systems. In healthcare, it is used to some extent to analyze narrative text for reimbursement and research purposes.

    But natural language processing has not come close to reaching its potential. Experts say developing this technology could play an enormous role in overhauling healthcare because it is the best solution to address unstructured data.

    "We are just now in the beginning stages of the adoption curve for any kind of natural language processing technology. People are just now realizing that there is some value," says Andrew Hickl, the CEO of Language Computer Corporation, a closely held company that develops such technology.

    Cultivating natural language processing would help computers understand the various ways humans express themselves when they write narrative text. This in turn would allow hospitals across the country to share important medical data and expertise, even if they do not have compatible electronic record systems.

    But the Holy Grail of natural language processing, say industry experts, would be to assist physicians in making near real-time decisions when meeting with patients. Getting near-instant feedback from colleagues across the country would improve physicians' ability to diagnose and treat patients. It could also warn them of impending epidemics.

    In fact a number of research facilities and companies are trying to take the technology to the next level. Earlier this month, IBM (IBM, Fortune 500) and the Mayo Clinic took a step forward when they announced the launch of an open-source consortium to try to improve the technology. By putting their knowledge on the open source, they hope to turn clinical notes into computer code.

    Many experts say at a minimum the open-source endeavor will help natural language processing gain traction.

    "I think it (consortium) will accelerate the awareness about the potential of natural language processing by getting more people to use these products," says Brian Hazlehurst, a senior investigator who develops natural language processing at the Center for Health Research of Kaiser Permanente.

    Still, doctors -- or their office managers -- still are going to have to get used to punching in codes and data, especially when it comes to providing payment and billing information. "If we go forward with this sort of technology, it doesn't replace structured electronic medical records. It simply acts as a valuable adjunct," says Charles Jaffe, the CEO of Health Level Seven, a not-for-profit organization that develops standards in healthcare.

    Put another way, technology may adjust to doctors' needs when it comes to patients' health. But doctors will need to learn to adjust to technology if they want to get paid.