Home > Applications > BI and Analytics

Computerworld - At Partners HealthCare System Inc., which includes some of Boston's most prestigious teaching hospitals, a knowledge management system for physicians has reduced serious medication errors by 55%. John Glaser, CIO at Partners, has been at the helm of this effort for more than a decade, bringing just-in-time knowledge to bear on life-and-death decisions. In this month's Harvard Business Review, he reports on progress and lessons learned with co-author Thomas H. Davenport, director of Accenture Ltd.'s Institute for Strategic Change in Cambridge, Mass. Computerworld's Kathleen Melymuka, who first reported on the project a decade ago, spoke with Glaser recently to get an update about how the system is working.

Going way back, what was the impetus behind the knowledge management initiative? Two things were at play: We know physicians are human beings and there's too much for them to know. If we could build in help for that decision at the time of decision—where decisions are played out as action—that would have a lot of power.

That thought ran in parallel with a Harvard study that was looking at the rate of medical errors at Brigham [and Women's] and Mass. General [hospitals]. The error rate was just too high, and very simple things were causing a good chunk of the problem—things like not knowing a patient was allergic to a drug or forgetting that two drugs interacted badly.

This is the latest in a series of monthly discussions with authors of articles in the Harvard Business Review on topics of interest to IT managers.

It seems that the key to this system is that it pushes information to doctors, rather than having them search for it. The general notion is "just in time." The computer knows that you're trying to enter an order. It knows a lot about the patient—test results, other drugs, diagnosis; and [it knows] something about medicine—what goes together. It applies all this and makes a judgment, and about 400 times a day, a physician changes his mind on an order based on the computer. That's only 3% [of all orders], but it's catching something in the workflow when the physician may not be aware he needs the knowledge.

How readily do doctors accept this unsolicited advice? We were smart in starting with advice on no-brainers: Telling that a patient is allergic to a drug or that two drugs together is a bad combination.

Would they rather we didn't tell them? Of course not. But we didn't go into a class of guidance that is debatable.