Database improves health care

This is the final installment in a four-part series looking at the role of information technologies in health care.

When the Washington Post called up Dr. Rob Califf to ask him whether Russian President Boris Yeltsin was healthy enough to run for reelection, Califf turned to his computer for help.

Califf, director of the Duke Cardiovascular Databank, compared Yeltsin's medical history, which includes a recent heart attack, with that of the 70,000 previous patients in the databank to project how Yeltsin's health might fare on the campaign trail.

His investigation highlights a new approach to health care, one with which doctors are more concerned than are reporters confronted with a deadline. The approach, outcomes-based health care, involves gathering medical data on large samples of patients, looking at how they fare, assessing the relative merits of different therapies and then using this knowledge to produce better outcomes for future patients with similar ailments.

While outcomes-based health care has been around for some time in theory, its use in practice is growing exponentially thanks to quantum leaps in technology. This rapid application of technology is changing the nature of health care.

"The old system was one in which physicians studied the chemistry and philosophy of how a disease worked and implemented theories based on this," Califf said. "Care was based on the physician's own observations, and training was essentially an apprenticeship.

"But you can't decide if a therapy is working or not based on each individual case. Now we have the computer to remember and aggregate numbers. We can measure what happened to patients, and if it wasn't successful, we can change our approach," he said.

"We can give multiple examples of where what we thought was right was in fact dead wrong," Califf continued. "Part of this is a humility lesson for doctors. All of us want to believe that our doctor knows what it is best, and it is not very reassuring to know that the doctor is not always right."

Among other things, the Institute often works with drug and biotechnology companies to compare the success of their new products with the performance of previous treatments in the database, so that the companies may present evidence of their products' merits to the Food and Drug Administration for approval, Mark said.

In addition, researchers deal with the economic impact of particular therapies as well as the quality of life of patients who have received them, said Dr. Don Fortin, director of the Medical Center's Health Information Center.

The Institute, which has an annual operating budget of over $30 million according to Califf, also makes statistical models. These models can then be used by health-care providers to understand how much inventory they need or how many procedures are required. This knowledge, in turn, helps them be more informed when bidding for managed care contracts with insurers.

While at present the models are being used mainly for administrative purposes, models can also be used at the individual level to help a physician treat his patient, said Dr. Eric Peterson, assistant professor of cardiology and a member of the Institute's clinical outcomes group. "With a tough decision, you would normally have wanted an old and wise physician. Now a young physician can make a good decision based on examples of thousands of people," Peterson said.

The physician can use these models as a tool at each step in the care of a patient. After initial consultation with a patient complaining of chest pain, the physician can look at a model which incorporates this information and his patient's vital statistics-such as age, weight and other health factors-to help him determine the next course of action.

If the model shows that the patient might have a high likelihood of having a serious problem, the physician may then decide to skip what is typically the next step, a treadmill test, and move on to a higher level, Peterson said. At each successive stage, the physician can consult a new model to help direct his decision-making.

Peterson said he uses the models a great deal in his work. "Most patients in the 1990s are very involved with their care and want to participate in the decisions," he said. "They appreciate knowing the relative successes of different treatments. Knowing the specific benefits and risks of a treatment, we can make a plan that we both become very comfortable with."

While the Cardiovascular Databank is highly advanced, it is certainly not the only one of its kind at the Medical Center. Several projects are underway at the Pickens Health Center to expand the use of databases in treating other diseases.

In one such enterprise, Dr. David Lobach, assistant research professor of biometry and medical informatics, is compiling data on patients who visit Pickens with diabetes mellitus. Lobach is using data from The Medical Record, an internationally-known program developed in the 1970s at the University.

The biggest challenge Lobach faces is organizing the information in ways that make it useful. "We need information coming in on many variables, and we need to be able to record it in a manner that is useful to us," he said.

As this new approach to health care continues to grow, so do some of the controversies surrounding it. One critique is that an outcomes-based approach, taken to its logical extreme, could make the practice of medicine automated, thus de-individualizing the patient.

Fortin disputes this claim, saying that the additional information serves to aid the physician, not dictate his decision. "If we had information on all of the various diseases, we, as a group, would be able to perform better health care by putting aggregated information together with individual information," he said. "I actually think this would make care more individualized."

Another critique of outcomes-based health care is that the information doctors use to determine the relative merits of various therapies is oversimplified, failing to take into account key factors that have significant impact on the outcomes of treatments.

Dr. Ralph Snyderman, chancellor of health affairs and dean of the Medical School, acknowledged that this type of fear has some merit, but said that, for diseases currently being treated using outcomes research, all of the variables are understood well enough to be included in the databanks. In other words, the databases contain all of the factors that could potentially affect outcomes.

"There are certain areas in which the data is very useful, such as in asking questions like, 'Is this particular form of treatment the most effective for a sudden heart attack?'" Snyderman said. "It is not as useful, at least as of yet, in cases that are multifactorial such as the long-term treatment of obesity."

Until then, such patients will have to settle for their doctors' own tried-and-true experience.