IT: More than a tool for quality improvement
Where will it go next?
For most organizations, health information technology (HIT) is a tool to be used in quality improvement projects, not the end in and of itself. But the future promises to be different: a time when HIT can be the end of the QI process, the improvement personified.
HIT is already used for decision support, and in medication safety issues, says David Bates, MD, MSc, chief of the division of internal medicine at Brigham and Women's Hospital in Boston. There are devices under development which can alert medical personnel when a patient is just beginning to have a problem. "These are expensive, but so are adverse events, so is the ICU," says Bates. He mentions one product being tested that sits between the bed and the patient, tracking his or her pulse, respiration and movement. A nurse is notified if something seems amiss.
Other uses of IT that affect quality include RFID technology that can help keep track of medications, devices, and patients alike. But the future? Bates finds that exciting to contemplate.
"I think we can eliminate more than 90% of the medication harm that happens using IT," he says. "Bar codes at point of care, pumps, checklists. And it can help with communication things that used to be done with a lab calling the unit clerk, who might or might not understand its importance. IT can know who is responsible and contact that provider." He gives an example of incessant paging going out to residents after patients have bypass surgery because their blood count dropped. "That always happens after bypass, so you use IT to figure out when someone has to be called, what it's important to page for," says Bates.
There is the potential for "wow factor" developments, too: a way to verify the concentration of a drug before delivery, for instance. But Bates is most excited about using HIT to improve reliability, communication, and tracking.
For another exciting development, consider the potential of predictive modeling, says David Classen, MD, an infectious disease doctor and chief medical officer at the technology firm CSC, based in Virginia. What if there was an automated program that looked at how out of sync more than a single data point was and the potential for that to result in a problem? "We are still working on this, and currently, we don't collect enough information in a way or in a time frame that it can be of use."
Currently, most vendors are laser-focused on meaningful use, so those who had been working on some of the more ingenious applications that can make healthcare safer, better, or more cost effective are unlikely to return to the workbench until after the meaningful use money has dried up. If healthcare wants some other kind of new invention, they're going to have to find some funding for it, Classen says.
The possibilities for predictive modeling
But Classen thinks that biometric predictive modeling does have potential in the near future. It's something about which physicians like Karen Joynt, MD, MPH, dream. A cardiologist and instructor at Brigham and Women's Hospital and Harvard School of Public Health in Boston, she says she'd love to see something that would look not just at single data points when they go out of whack, but multiple data points and their trend lines. For instance, a patient whose blood pressure is declining, whose heartbeat is intermittently abnormal, whose kidney function is declining that is a patient who is heading for something bad, she says. But right now, things have to become critical before an alarm sounds. "There is a lot of work in rapid response, but from my perspective, the untapped use of data is to prevent something that hasn't happened yet. You could have a lot of power if you could look at vitals and lab work and pick up early warning signs."
The information is there and available, Joynt says, but it isn't spit out when a cluster of things is trending badly. That's what she wants to see, and Classen says it will happen in the near term.
Far further down the road is populationwide predictive modeling to determine things such as whether a particular group of patients is likely to be compliant with their care. "I'd say we are 10 to 20 years from using predictive modeling to estimate risk based on behavior," Classen says.
Among the companies working hard in this area is Xerox. "It's about more than an early warning system," says Justin Lanning, vice president of business development and marketing for ACS, an IT firm recently acquired by Xerox. Dozens of hospitals are moving ahead with this already, Lanning says. Some areas are riper for it than others; for instance, predicting sepsis and other uses in infection control.
The problem is that most people have anticipated that electronic health records would provide all the data one needed and more. Lanning notes that a typical patient has some 2,200 data points collected per stay. But all that data, even in an electronic form, still can't give Joynt information on her patients that warns her something critical may happen in the near future if she doesn't attend to them.
"The electronic record can't handle the intelligence because they are built to help with billing and admissions and discharges," Lanning says. What needs to happen is more cooperation between companies creating the electronic medical records and those who want to make the data more useful, like ACS and other HIT companies. "Then we have the challenge of figuring out what is the meaningful data, what is the combination that means something."
In a few years, Lanning says HIT will be able to make the predictions Joynt is anxious to see. It will work to help figure out which patients are most likely to bounce back to the hospital so that interventions can be taken to ensure that doesn't happen.
The excitement is justified, says Larry Van Horn, associate professor of health care management and executive director of health affairs for Owen Graduate School of Management at Vanderbilt University in Nashville. But he remains skeptical about how far healthcare as an industry will go with predictive modeling and how fast. "We have, as an industry, been woefully inadequate turning data into intelligence, and predictive modeling is one component of that."
Van Horn says that issues of false positives and false negatives, aligning of incentives, and figuring out what the key events are all have to be worked on before the "idealized promises" can be realized. "We have to change people's view. If you have a group of people who have a higher probability of a bad outcome, you might not find something wrong with all of them, but you can avoid complexities on a couple of them, and that will generate savings in the population as a whole," he says. "We have a lot still to learn, though."
For more information on this topic, contact:
- David Bates, MD, MSc, Chief of Division of Internal Medicine, Brigham and Women's Hospital, Boston, MA. Telephone: (617) 732-5650.
- David Classen, MD, Chief Medical Officer, CSC, Falls Church, VA. Telephone: (703) 876-1000.
- Karen Joynt, MD, MPH, instructor in cardiovascular medicine, Brigham and Women's Hospital, Boston, MA. (617) 432-4893.
- Justin Lanning, Vice President of Business Development and Marketing for the ACS, a Xerox Company, Nashville, TN. Telephone: (615) 712-2212.
- Larry Van Horn, Associate Professor of Health Care Management, Executive Director of Health Affairs, Owen Graduate School of Management, Vanderbilt University, Nashville, TN. Telephone: (615) 322-6046.