When is a sore throat not a sore throat? To a computer, it's when the sore throat is called pharyngitis, a throat infection, an irritation of the throat or throat soreness. The same thing goes for high blood pressure when it's called hypertension, hypertensive disorder or elevated blood pressure.
Those simple differences in clinical terms to describe much the same thing, easily translated in doctors' heads when they come across them in fact-finding, are an obstacle in building clinical information systems and an electronic health record.
Earlier this month, the federal government took steps to forge agreement on hundreds of thousands of clinical terms, agreeing to pay $32.4 million to license a leading database of terms and offer it free to the healthcare industry beginning in January 2004. The terminology system, called the Systematized Nomenclature of Medicine, or SNOMED, was developed during the past 40 years by the College of American Pathologists at a cost of $35 million, says Diane Aschman, chief operating officer of the organization's division for terminology development.
"The importance of this is it's really the foundation for a standardized medical record that could be shared," Aschman says. "It codes the detail of what happens to the patient and why the patient requires various procedures and drugs."
"This system," says HHS Secretary Tommy Thompson, "will prove invaluable in facilitating the automated exchange of clinical information needed to protect patient safety, detect emerging public health threats, better coordinate patient care and compile research data for patients participating in clinical trials."
In addition to licensing the core SNOMED computer architecture and its 344,000 terms, HHS commissioned the Institute of Medicine to design a standard model for an overall electronic health record, which it also will share free of charge with all sectors of the U.S. healthcare system.
Already the Healthcare Information and Management Systems Society has submitted to the IOM a detailed proposal that was negotiated during the past seven months and approved as an industry-consensus model the day before the HHS announcement, says Stephen Lieber, HIMSS president and chief executive officer. Lieber says the HIMSS document will be "a starting point" for the IOM project and will jump-start a standards-adoption process by an official industry agency called Health Level 7 that calls for completing an electronic health-records framework in January 2004.
The federal moves are the latest in a series of standards-adoption initiatives in which HHS and the U.S. Defense and Veterans Affairs departments have thrown their weight behind common formats for exchanging clinical information electronically (March 31, p. 10).
Providers and their IT vendors say the rapid adoption of the SNOMED standard in new information systems and through updates to existing systems will pave the way for better performance of new clinical decision-support information systems. Those computer programs are capable of analyzing patient data to head off harm and guide doctors to the best treatment, but the software systems are only as good as the data they're fed (Feb. 10, p. 6).
The universal adoption of a standard for clinical terms also promises to gradually remove millions of dollars in upfront expense and annual information technology maintenance costs now devoted to translating terms from one computer system to another, says Nancy Brown, senior vice president of strategic planning for the information solutions group of San Francisco-based McKesson Corp.
The terminology barrier
Much of the work-and cost-in implementing a clinical information infrastructure in a healthcare system involves connecting information systems for laboratory, pharmacy, nursing documentation, patient vital signs and other components that all contribute important data on the same patient, Brown says.
With so many ways to express the clinical terminology for complaints, disorders and observations, IT vendors and their clinical consultants were forced to arbitrarily pick one definition for each of the thousands of medical terms a computer has to recognize when creating, amassing or storing clinical data.
The problem began when hundreds of vendors settled on subtle differences in identifying and narrating medical problems. Resolving those differences when exchanging information from one computer system to another through the use of complex software interfaces can cost healthcare organizations millions to engineer and additional annual expenses of updating the work for new or changed clinical terms.
That was the case at Oakland, Calif.-based Kaiser Permanente, which "has implemented a wide variety of clinical systems over the past decade," says John Mattison, assistant medical director of clinical-systems development. "Unfortunately, every vendor has used their own proprietary set of medical terms and definitions."
The considerable work to fix the confusion is often complex and tedious, Mattison says. "We have spent millions of dollars doing this work over the years, and it continues to represent a substantial amount of overhead," he says.
"Once all clinical-system vendors use SNOMED as their common terminology, these problems will all but disappear," Mattison says. As a dividend, Kaiser's annual software maintenance expenses "should decline in excess of $1 million each year," he says.
Up to now, the effort to identify all data that had to be altered for other computer systems was likely to be uneven and often incomplete, Brown says. For example, a search of patient history and recent tests might miss information that wasn't properly identified in an outpatient database. Or the information from some physician practices and other care sites might not be available at all because of the expense and trouble of putting it through the translation process and keeping the interfaces current. That increased the chances a clinician could be presented a picture lacking all the factors needed to make the best decisions, she says.
But to get doctors to use clinical systems with confidence, IT vendors have to improve the sophistication and impact of their decsion-support programs. Foolproof standards for pulling in the right patient data "are absolutely critical to all our advancements in clinical information technology," Brown says.
For one thing, the healthcare industry's ability to adhere to recommended best practices for treating medical conditions won't rise higher than current levels until computerized alerts can be made much more specific, Mattison says. Alerts generated when doctors order tests and medications tend to go off too often, resulting in "alert fatigue" that leads physicians to disregard the warnings or complain about them, he says.
The problem is that without specific medical documentation, a computer can't zero in on a match between a general description of a potential danger and the narrower clinical situation facing a patient given his history and unique medical condition, Mattison says. With SNOMED as a foundation, alerts will be much more likely to communicate a true danger, and doctors will be more inclined to take them seriously, he says.
Because a significant amount of medical record information for patient care and public health reporting is in physician offices, small practices can't be discouraged by the cost of recording and contributing electronic data, industry observers say.
The upfront licensing cost for the government is a small price to pay compared with the potential to lower the implementation and maintenance costs of crucial clinical systems, especially for small providers, an HHS spokesman says.
Putting reporting on autopilot
Aschman says her interest in creating and retrieving clear, concise information goes back 20 years to the advent of DRG-based reimbursement for Medicare services, when she worked at medical products distributor Baxter International.
The shift in payment approach called for a good sense of what it cost to provide a medical service and what was being done for a patient, but there was no easy way to investigate the healthcare process except with "armies of people with pencils" poring through paper charts, she says.
In 1993, new reporting requirements from the National Committee for Quality Assurance involving measurement of health plan performance focused industry energy on collecting data to see if providers were meeting requirements of the Health Plan Employer Data and Information Set, or HEDIS. "People were scrambling to answer what on the surface were basic questions," Aschman says. But the evidence was fragmented, scattered and not in a form that could be gathered, aggregated and shared, she says.
A third wave of clinical reporting requirements is ready to descend, this time from the Centers for Medicare and Medicaid Services. The CMS is completing a pilot to measure the quality of care rendered for certain conditions, such as pneumonia and heart attacks, by tracking whether specific tests and interventions were done (June 30, p. 14).
Those reporting chores, which the CMS eventually would like to expand industrywide, are on top of expanded requests for evidence of clinical performance and patient safety from the NCQA and the Joint Commission on Accreditation of Healthcare Organizations, as well as new demands for supplying up-to-date information to local public health agencies.
A common medical vocabulary will help make those reporting requirements an automatic process using the same data over and over instead of a labor-intensive onslaught on patient charts for each agency, Aschman says.
Among the HIMSS model's essential requirements for an electronic health record is that healthcare vendors' clinical applications must support mandatory reporting and other initiatives such as disease management without additional data entry by clinicians.
The clinical terminology licensed by HHS is far from static, and the pathologist group itself faces annual maintenance expenses just to keep up with new wrinkles in medical care-recent examples include the emergence of severe acute respiratory syndrome as well as exotic diseases such as monkeypox.
Aschman says the $32.4 million from the government will help underwrite the cost of annual updates and maintenance, freeing experts in the SNOMED division to spend more time improving the quality of computer tools that capture and represent medical knowledge.