Do we really need universal patient identifiers?

A person’s health picture is a fragmented, broken mirror, when you look across our archaic paper-based landscape. According to a survey we did last year, patients recall seeing over 19 different physicians in their lifetime, underscoring the reality that each place where health data is stored (each physician’s office) represents a piece of the whole picture. As we move from paper to an electronic platform, the potential of linking those pieces together moves out of the realm of “science fiction” and into the realm of real, day-to-day experience.

Meaningful Use of Electronic Health Record technology has helped move us along. Stage 1 was focused on getting clinicians and hospitals used to having computer screens, rather than paper charts, in front of practitioners’ eyes all day long. Stage 2, under development now, focuses more on connectivity – linking the electronic systems together in ways that are feasible.

However, as we try to knit together the pieces of health data into a coherent whole, one of the conundrums we face is the fact that each location – each physician EHR, each hospital, each lab, each insurance plan – refers to the same patient using differing identifiers (patient account or record numbers). There is no standard, and no easy way to link them all together. Social Security numbers, for example, were never intended to serve as a universal identifiers, and several regulations even prevent the use of SSNs for this purpose.

So, how do we link it all together? Do we need to create some sort of national “universal patient identifier,” or is that too invasive and Big Brother-ish for our society to accept? Heated debate about this, involving divergent stakeholders, has yet to come up with a consensus. The federal government, in fact, has published guidelines for creating Universal Healthcare Identifiers.

“Big system” needs vs. local practitioner needs
The challenge is different when considering how “big systems” look at the question of interoperability, compared to how it impacts local practitioners. A “big system” looks at the need to automatically link multiple large stores of data, without the need for human intervention – how to distinguish between the “several hundred, perhaps even several thousand patients with the same name, with the exact-same birthdate in a given geography,” as mentioned by the CMIO of one large “big system” at a recent conference. Within a given “big system,” the problem is easy – simply assign some internally-derived identifier to each unique patient, and use it everywhere inside that system. This works for an integrated delivery network, but fails when you connect outside that “walled garden.”

Is a Universal Healthcare Identifier really necessary for exchange of health information between systems? Do we need to wait for such a system to be implemented (if it ever does)? No – “fuzzy matching” is currently done, and is acceptable at the local physician level.

One example of such “fuzzy matching” is in routing lab data, or refill requests for e-prescribing. Several elements of patient identification are sent (name, date of birth, gender, zip code) from the source (the lab, or the e-prescribing network). The receiving system (the physician’s EHR) then tries to match that data with its in-practice patient list, and suggest a “best fit” match. Depending on how the EHR is built, this can be pretty sophisticated, but will not always be a 1:1 match. If the EHR system finds anything other than a perfect, unique match, then a “short list” of likely patients is presented to the user (the clinician), with the option to search the whole practice if the “short list” doesn’t find the right one. To date, this method has worked well.

The difference between “big systems” and local docs illustrates this adage: all face-to-face medicine is local. Matching data from different systems involves linking things together, and machines can help assist that process. But individual, local clinicians make that linkage. “Big system” linkage on very large scales can’t do that. It must rely on the linking done at the local level, one-at-a-time. Therefore, the data accumulated is limited to (1) the realm within a “big system”, or (2) the manual linking record-by-record done locally between external data sources.

For the purposes of delivering healthcare to individuals, and accessing the needed data that may exist in different silos, the “fuzzy matching” method we have now will move us in the desired direction. For the purposes of data analysis across multiple systems, that is still a ways off. As a society, we may not have the desire for a truly Universal Healthcare Identifier system. But, short of that, we can still link things together in meaningful ways, and conduct data research in ways never before possible.