HIMSS14: HIEs attempt to solve patient matching puzzle

ORLANDO—Health information exchanges, as they develop, are discovering that patient matching across organizations is a challenge very different from internal matching, according to a presentation at the Health Information and Management Systems Society's annual conference.

“A lot of organizations could be setting themselves up for a shock in the future in that they may have a blind spot in terms of patient identity management,” said Eric Heflin, CTO for eHealth Exchange and HIETexas.

A heterogeneous technological landscape combined with varying workflows, different data characteristics and large latency increase the difficulty of matching between organizations.

“In the absence of a single national identifier, for example, we have two problems,” explained co-presenter Christopher Ross, CIO for the Mayo Clinic in Rochester, Minn. “One is the inability to detect a shared patient, which you might consider a false negative. And the second is matching an incorrect patient, which you might consider a false positive.”

Authorization management is another barrier, with some organizations encountering issues where there is a lack of patient authorization on file at a responding organization or an inability to collect authorization at the point of care.

As an example of the scope of the problem—and how to address it—Heflin and Ross described a trait sensitivity study being conducted by Healtheway, a nationwide eHealth exchange network, and the Care Connectivity Consortium (CCC), which works to spread the adoption of a national HIE network.

The study aims to evaluate the recommended minimum traits for successful patient matching and minimum processing models. A database of more than 4 million clinical patients was compared with a sample of 10,000 HIE patients using Intermountain Healthcare and local exchange partners including the Utah Health Information Network. It was based on a live production pilot using CCC services.

Results from an initial algorithm revealed that the match rate was just 10 percent. “Can you imagine being there right then?” said Heflin. “Here they are, two very sophisticated, world-class organizations that did the research, they invested heavily in IT technology and they have a match rate of 10 percent.”

However, the study reevaluated the usefulness of various traits of the matching algorithm—including patient first name, last name, date of birth and other information—and eventually boosted the matching rate to more than 90 percent. Additionally, the study recommended standard formatting of demographics among HIE participants and accounting for differences in workflow, data capture and employee training. Patients were also shown to be inconsistent themselves, sometimes offering different names for different organizations. Heflin suggested also ditching deterministic matching for probabilistic matching.

Addressing matching issues will be key to maintaining patient safety and satisfaction, said Ross. “Our patients expect that the physicians who care for them can collaborate. They expect that the data are going to be transportable.”