Hope for patient ID dwindles

Connecting for Health Report

Some health information technology planners have waited for years for the federal government to issue every American a unique personal health identification code, similar to a Social Security number that would distinguish each of the 16,000 William Smiths in America from one another.

But it’s increasingly unlikely that will happen, despite its seemingly obvious appeal for the architects of the national health information network.

And now many of the leaders in the drive to modernize and interconnect the nation’s health information systems say they wouldn’t rely on a single health identifier if one were to be issued.

Instead, they are endorsing two basic strategies:

* Record locator services, which would maintain indexes of patients and where to obtain their health records.

* Probabilistic matching, which is a relatively new technology used in search engines.

For a doctor whose office wanted to find the records of a new patient, the record locator would point to the hospitals, pharmacies and other locations where the records could be found. The matching system would narrow down the possible matches so that only the records of the patient in question would be fetched.

Precisely how the system will work is still unclear. Some answers are likely to emerge from the National Health Information Network prototypes that will be built over the next year for the Department of Health and Human Services, and also from demonstrations such as the large-scale records exchange project launched in June by the Connecting for Health program. The program is sponsored by the Markle Foundation and the Robert Wood Johnson Foundation.

The issues include retrieval speed, accuracy, privacy protections and scalability. A system that works well for a region or state may not perform as well on a national scale. The system also should be able to be deployed rapidly.

Experts such as Dr. Carol Diamond, managing director of the Markle Foundation’s health care program, and Dr. Shaun Grannis, a physician and scientist with the Regenstrief Institute, Inc., in Indianapolis, say they are convinced that a system can be designed to do the job without relying on an official ID number issued by the government.

A task force created by Connecting for Health issued a report in February 2005 advocating a decentralized approach to linking patient records.

The report, “Linking Health Care Information: Proposed Methods for Improving Care and Protecting Privacy,” acknowledges that the idea of a single national health ID “is appealing in its simplicity: give everyone a unique number, to be used only for their health records, so that linking two records becomes a matter of comparing the numbers. If they match, the records refer to the same patient. If they don't match, they refer to different patients.”

But the task force said that view was too simplistic.

Probabilistic matching uses several pieces of data, such as first and last names, city and state of residence, date of birth and gender. Besides seeking true matches of strings of data, the algorithms can find words that sound much the same and can check for variants resulting from misspelling and data entry errors.

In addition, system designers can assign different weights to each piece of data that will be matched, increasing the accuracy of the matches.

As described by the task force and others, the record locator services would be new organizations or new services provided by existing organizations. Their job would be to maintain indexes of patient records, but they would not hold the records themselves. The records would remain in the institutions where they were created.