When they dream of the latest addition to their clinical information technology systems, healthcare IT leaders still have wide-open spaces in which their thoughts may roam.
A recently released survey by the Medical Group Management Association revealed just 28% of 560 group practices responding had an electronic health record, or EHR, system, and only 10% said their systems had clinical decision-support functions. In March, the Journal of the American Medical Informatics Association published the results of a 2002 survey of 626 responding hospitals in which just 16.1% had fully or partially installed computerized physician order entry, or CPOE, systems, the basic launchpad for first-generation computerized decision-support tools.
Though even EHR and CPOE systems remain beyond the grasp of a majority of office practices and hospitals, some tech leaders are pushing forward beyond adverse drug alerts and beta-blocker reminders to a second generation of more sophisticated computer-based decision-support features. They are developing tools that medical informaticist William Bria, M.D., calls "stick shakers."
Bria, president of the Association of Medical Directors of Information Systems, takes the term from a safety device added to jet aircraft to help pilots transition from propeller-driven planes. A biplane's control stick starts to shake as it reaches stall speed, Bria says, but with a jet there was no mechanical linkage to the stick, so it didn't shake until the pilots vibrated, too.
"These guys were falling out of the sky unless they felt something physical," Bria says. "So they built a tiny shaker in the pilot's hand.
"What we need in medical care today is stick shakers," Bria says, but not just "dumb alerts" like basic prompts. The new second-generation systems should "provide more useful and sophisticated information in a place and time that's appropriate."
They would be "passive but intelligent agents that are working alongside you with artificial intelligence-type programming that makes itself known only when a number of tried and true events occur. It's showing up, just slowly, but surely."
AI on watch over me
One physician IT leader who is dreaming of what never was and asking why not is Richard Kremsdorf, M.D., president and chief executive officer of San Diego-based CliniComp International, a developer of decision-support software.
Kremsdorf, a pulmonologist and critical-care specialist, also has served as vice president of clinical information systems at Catholic Healthcare West, San Francisco, and was medical director of clinical information systems at ScrippsHealth, San Diego.
Unlike alerts, CliniComp's new Essentris OnWatch artificial intelligence system is designed to catch errors of omission, which cause more deaths than the 98,000 per year attributed to preventable inpatient medical errors in the 1999 Institute of Medicine report. "Errors of omission are basically unseen," Kremsdorf says. "There are lots of times where there is a bad outcome and you know, if the system worked differently, it didn't have to be that way."
The OnWatch system, being piloted at Sharp Memorial Hospital in San Diego, monitors every patient in the hospital, centrally locating and displaying alerts to supervisors only on those patients whose data has been flagged by a rules engine in the software as being or trending outside preset norms. Distress triggers include infection indicators, fluid status imbalance, respiratory compromise, laboratory results and hemodynamic instability.
Kremsdorf says that in most hospitals, even in the intensive-care unit, data may be acquired electronically, but it gets written on a piece of paper and that paper is somewhere -- maybe in the right place at the right time, but only in one place.
"So that data is only available to the individual who goes to that piece of paper, and all the decision support is done by the person who has found that piece of paper in that person's head. How do they find where there is trouble that will benefit from their expertise? Someone tells them."
With OnWatch, "We're talking about data that can be seen by multiple persons and identify and explain in a dashboard which patients are having troubles."
A diabetes control team at Sharp "has really glommed onto this," Kremsdorf says. Data on blood-sugar levels are entered into the hospital's lab IT system, which flows to OnWatch. "That number is electronically sent, and we run a rule against it and determine whether it?s a value that should be displayed on our electronic dashboard."
Control-team members aren't scanning individual charts or looking at data from hundreds of patients. The computer does that for them.
David Nash, M.D., is chairman of the department of health policy at Jefferson Medical College of Thomas Jefferson University in Philadelphia. He also teaches the Advanced Applications in Quality Management program of the American College of Physician Executives and co-edited the recently released graduate text for health professionals, The Healthcare Quality Book: Vision, Strategy and Tools.
Nash says he foresees hospitals and physician offices won't be the only places where decision-support tools will reside. Today, he says, patients can log onto the Internet and complete a do-it-yourself Framingham heart-risk assessment online. In the future, Nash predicts, direct connections by patients to "a whole host of interactive IT-type tools" will be the norm for the management of chronic diseases.
Cerner Corp. has pledged to spend
$25 million over 10 years to develop one such patient-provider link. The idea is to create electronic health records for up to 1.3 million children with juvenile diabetes and improve communications, including the transmission of test results, between the patients and their doctors.
Neal Patterson, CEO and chairman at Cerner, says physicians will enroll patients in the first phase of the program through 12 host children's hospitals and medical centers. Physicians already using a Cerner system would integrate the diabetes program into their daily workflow, while those not using a Cerner system would access the records through a secure Web site.
Patterson says current-generation glucometers and insulin pumps gather data but have no way of transmitting it to a central database in real time. But he says he believes that in the future, the devices will be able to transmit that information directly into an electronic medical record and set off an alert if the readings fall outside a prescribed range. Cerner is working with the manufacturers to add that capability, he says.
Daniel Nigrin, M.D., chief information officer for participating Children's Hospital Boston and an assistant professor of pediatrics with the Harvard Medical School, says current communication systems are "inconvenient and inefficient."
"There is so much data that needs to be communicated back and forth between patients, nurses and nurse practitioners and doctors, and right now it happens in clunky ways, such as by phone," Nigrin explains. He says the Cerner project will allow faster transmission of the information, the eventual goal being the transmission of real-time information to the doctor.
"This is the first time something like this has been done gratis and on such a large scale," Nigrin says. "The idea is not revolutionary in itself, but the push to extend this to all comers, whether Cerner clients or not, is pretty significant."
Internist Scott Weingarten, M.D., is a clinical professor of medicine at the University of California, Los Angeles and CEO of Zynx Health, a developer of evidence-based order sets and clinical guidelines. Zynx is in negotiations with vendors of clinical information systems, such as CPOE, to add a "content authoring space" to their systems to aid rapid deployment and updating of clinical pathways and other decision-support procedures, Weingarten says.
"There is never going to be shrink-wrapped decision support," he says. "It always has to be customized and debated by physicians. So we provide them with the ability to customize their system, and you need to provide the clinical justification that's just one click away."
Steve Epstein is CEO of MEDai, a developer of artificial intelligence software that calculates from claims and clinical data the expected quality and cost outcomes of a patient hospitalization. The system works on about 40 disease states, and cuts through the fog of variables. It "really gets down to the three or four drivers that are different for your patient," Epstein says.
For now, the system operates retrospectively, but Epstein says the goal is to have it working with the physician in real time. The last barrier is the timely, electronic input of relevant data, he says.
Toward that end, MEDai has contracted with clinical IT vendor MedPlus, a division of national laboratory operator Quest Diagnostics, with the goal of developing an electronic medical record system equipped with artificial intelligence from MEDai. Development began in March, he says.
"We will be the intelligent agent in the background looking for patterns, to put the intelligence in the front end" at the point of care, Epstein says.
Modern Healthcare reporter Mark Taylor contributed to this story.