Skip to content Skip to navigation

Live From eHealth Initiative: Why You Should Care About Patient-Matching Algorithms

February 4, 2015
| Reprints
Will current algorithms for patient matching scale up?

How important is accurate patient matching to the interoperability of health data? Here’s the problem in a nutshell: In Texas, the Harris County Hospital District’s database has medical records on almost 2,500 people named Maria Garcia and 231 of them have the same birth date. 

At the eHealth Initiative annual meeting in Washington, D.C., on Feb. 4, Shaun Grannis, M.D., an associate professor of family medicine in the Indiana University School of Medicine and a medical informatics research scientist at the Regenstrief Institute, talked about the increasing importance of the algorithms HIEs and enterprise master patient index vendors rely upon. 

The Indiana Health Information Exchange (IHIE) has researched how fragmented data is in the Hoosier state, he said. Twenty-five percent of labs in the IHIE are performed outside the home health system of the patient, and 40 percent of emergency department patients have data residing in other hospital emergency departments. Each health system has created its own identifier for the patient. So how do HIEs match up these patient records?

They rely on algorithms that have about 92 percent accuracy, Grannis said. Since the country currently lacks the political will to adopt unique patient identifiers, it is important to understand whether the kind of algorithms used today will scale up to handle the needs of ACOs, larger HIEs and research networks. “Billion-dollar bets are being made that if we invest in integrated medical data, good things will happen,” he said. “Will current algorithms for patient matching scale?” he asked. “It is an unanswered question today. If they do, we need to make sure best practices are crystal clear because we are betting big. If it doesn’t work, then we need to begin strategizing about weaknesses in the system.”

Lee Stevens, policy director of ONC’s State HIE program, talked about some of the things ONC is working on to help with patient-matching algorithms, including work on standardizing the attributes used such as date of birth, names and historical addresses, and through certification how EHRs capture this data. ONC has innovators in residence working on an algorithm that can be made available on for health systems to use or test against.

Congress has prohibited HHS from using funding to look at unique patient identifiers. It is a social and political problem defined as a technology problem, Grannis said, adding that in countries that use unique identifiers, such as Canada, they work well (although they are not perfect). He noted that Indiana studied the possibility of assigning a unique state identifier to all patients in the state. They estimated it would cost $250 million, and that would only raise the accuracy another five to seven percent. It was a non-starter, he said. “Show me the politician that is going to invest in that,” he said.






And yet the health industry thinks the ONC interop plan is a great idea even though ONC is sticking the providers and vendors with a problem Washington does not really want to deal with. The ONC deadline is provider implementation by 2017. Now here’s yet another date that will get pushed out and I am sure several times while providers will take the brunt of the beatings for the many delays.

I predict all the delays will be tied to one impossible hurdle buried deep in the plan that could be eliminated with the wave of a Congressional wand. The big, very big hurdle is creating a real unique patient identifier.

So there you have it, ONC wants providers to accomplish what Congress refused to deal with and do it through a Rube Goldberg process of virtual and probabilistic tools. And by the way ONC says they will measure via the MU program “provider performance”, e.g., the percent accuracy rate for proper patient identification with MU penalties doled out if you miss the measure.

Sad story is if our government was willing to deal with the ID issue head on we could have true interop in six months. After forty years in this industry one thing I have learned is that politics and systems design make terrible bed fellows.

My suggestion is: If a person, such as the real Maria Garcia, wants true health record portability coast to coast they agree to subscribe to a unique ID, if they want to maintain ‘privacy’ they do not. Opt in or Opt out, your choice.
Frank Poggio
The Kelzon Group


Thanks for your observations. If it weren't so sad, it would be funny to hear the folks from ONC describe how they were by statute not allowed even to mention the idea of unique patient identifiers as they crafted their interoperability roadmap for the next 10 years. Yet at the eHI conference, someone mentioned that the feds have given grant money to researchers who were helping other countries work on developing identifiers.

David Raths