Dr. Carol Diamond is managing director of the Markle Foundation's Healthcare program. The foundation is the principal sponsor of Connecting for Health, a public/private program to advance connectivity in health care in order to improve its quality, safety and efficiency of health care. Government Health IT interviewed her about the work Connecting for Health has done on how to match electronic patient records from different systems to get a complete, accurate picture of the patient's medical history and condition.

Q: Some people think the most elegant solution to accessing and matching patient records online is a national health care ID number, which the government would issue. At one time, Congress directed the Department of Health and Human Services to issue such numbers. Why has the notion met resistance?

A: There is a long history in this country of political resistance to any form of a national identifier. That's always run high in the United States. In the current climate, post-[Sept. 11, 2001], any discussion of increased government inspection of personal data has been met with some considerable concerns for privacy and civil liberties. The record of success for linking information in the face of such opposition is extremely poor.

Q: In view of the interest in a National Health Information Network, do you think the idea might be revived?

A: We don't see that on the horizon. Also, in Connecting for Health, which is a broad, multistakeholder collaborative, because of the importance of linking information for patient safety and quality, waiting for [a national network] wasn't practical. Even if one were on the horizon, it would take years to deploy and adopt and implement.

It was clear to us in our thinking around three-year time frames that we needed to be thinking about how to address the problem without holding out hope that something like that would magically appear and solve everything.

Q: In the absence of a national health care ID number, what is the most promising or likely alternative plan for matching patient data via a national network?

A. We spent a year looking at this issue, and our working group had some of the country's leading experts on these issues. Their report is called "Linking Health Care Information: Proposed Methods for Improving Care and Protecting Privacy," and it can be found at www.connectingforhealth.org/assets/reports/linking_report_2_2005.pdf.

Their recommendations were that the disadvantages of a national health identifier outweighed the advantages. Because we are thinking about what can be done in the near term and because we wanted to have some of the privacy and security issues shape the way we thought about architecture and technology, we concluded that the linking of patient records and matching of patient records can be accomplished using methods that are already being deployed today.
It's called probabilistic matching or algorithmic matching. That is something that the industry has broadly deployed for years, and actually health care is not unique in that. Many industries deploy this [technology].

We are suggesting that as an approach going forward, not only because it avoids the need for a national identifier, but also because it has the capability to link data accurately. In other words, we need to get the value accomplished that a national identifier would accomplish, and we think this can do it without the risks of the national identifier. We call this an index or a record-locator service.

Q: Aren't some clearinghouses now performing this function?

A: Absolutely. And it's not only clearinghouses, it's many large integrated delivery systems. There's a term of art in the industry called a master patient index. It's the only way an organization where care for patients is delivered at multiple sites or where multiple systems are in place can reconcile records and make sure identity is properly matched within that system.

Q: Will it add another layer of complexity and expense to the system?

A. The only new piece of infrastructure that Connecting for Health has recommended is this index. You can't address a problem without having to do something new. This is something that organizations today that are delivering health care are doing already within their enterprise.

And it doesn't require all of the new things that would have to be done if there were a national identifier ? issuing it, administering that issuance process, adopting it. You know, most ID systems today don't have a free space called "national identifier" waiting for something to be put there.? It would require a massive adoption and implementation effort, not to mention the political resistance and the overall administering and development. That would come with many more delays and expenses.

Q: Is the accuracy of probabilistic matching good enough for health care?

A: It depends on how probabilistic matching is deployed, and we're working on this problem right now. The level of accuracy is very good. And even if a national identifier existed, we would still have some errors ? missing numbers, transposed digits, shared numbers. Even the Social Security number is not without its flaws. What we learned in this discussion was that for this to be done on a national scale, you could never rely on any one identifier anyway. There is always an error rate that you have to take into account.

Q: Is this issue a major hurdle for development of a National Health Information Network?

A: No question about it. We identified this as one of the issues that needed to be solved in order to achieve connectivity in this country. If you couldn't accurately identify a patient who has medical information in two different systems ? forget national, they could be across the street from each other or down the hall from each other ? if you couldn't accurately match the information for that patient, our goals of improving quality and safety, reducing medical errors, and giving both the patient and the clinician the information that they need at the point of care to make good decisions, couldn't be achieved. It is absolutely essential that this be one of the aspects of interoperability that we focus on.