The patient-matching challenge

April 25, 2018
Mark LaRow

The well-publicized Trusted Exchange Framework and Common Agreement (TEFCA) proposal from the Office of the National Coordinator (ONC) contains some very good ideas. For example, providers should indeed have ready access to all health information for their patients. Patients should also have easy access to their health information wherever it resides. But one major criticism of the TEFCA proposal is that it doesn’t address how to enable the nationwide patient matching that is the cornerstone of the entire system. The current generation of matching technology will never be accurate enough to serve on a nationwide scale. I believe the ONC or the RCE in TEFCA should specify a “national identity matching resource” used by all Qualified Health Information Networks (QHINs), and that the resource should be based on Referential Matching.

I agree with TEFCA’s call for a centralized authority, the Recognized Coordinating Entity (RCE), and a top layer of interoperability, which are the QHINs. It is too optimistic to think that market forces alone can lead to a secure and interoperable health information exchange. While that process might work, it would take many years or even decades to achieve the goals set forth in TEFCA. That’s not an indictment of the healthcare stakeholders or healthcare IT, but it seems there is a natural requirement to have a single governing body take responsibility for providing a uniform foundation that many independent participating stakeholders can build upon.

There are important examples of this outside of healthcare. For instance, we would not have nationwide phone number portability if it were not for a central authority providing phone number translation services to all phone carriers. Similarly, we would not have a coordinated internet if we didn’t have the Internet Corporation for Assigned Names and Numbers (ICAAN) governing the standards and operations for a network of domain name servers (DNS). In fact, the parallel between the ICAAN and the RCE goes deeper. In the same way that the DNS network must serve a wide range of requirements from the largest enterprise servers to the smallest Internet of Things (IoT) devices regardless of vendor. So too does our healthcare network need to serve the largest providers/payers as well as the smallest community hospitals and individual patients, regardless of EHR product or patient app. The only way to do this is to have a powerful network layer that harmonizes differences in identity-matching, from the most technologically sophisticated providers to the least technologically sophisticated clinics.

TEFCA describes the need for a patient “Identity DNS” capability

The Communications Broker functionality described in TEFCA is, in essence, an Identity DNS. It says that a set of QHINs will collectively manage patient identity and record location information for hundreds of millions of patients, contributed by many thousands of participants. And, the QHINs must collaborate to create a virtual nationwide MPI and RLS so that all participants and individuals can access appropriate healthcare information without having to deal with the underlying complexity of discovery and consolidation. The reason we don’t have this capability today is NOT because we don’t have HIEs and IHE/FHIR protocols. It’s because our current patient-matching technology is not good enough.

“Probabilistic matching” is the algorithm used by all modern patient matching software, as exemplified in eMPI products and, to a lesser degree, in EHR products. It’s been around for over 30 years and we’ve squeezed every match we can from it. With a lot of manual effort and constant tuning, it serves the matching needs within the four walls of a hospital system acceptably well because hospitals can impose strict data standards, data governance, and tuning settings. They can also hire “data stewards” to resolve patient matches that the algorithm cannot. Even then, hospitals still see 8-15% patient duplication rates in their EHRs and MPIs. Backlogs of hundreds of thousands or millions of potential matches are generated by these eMPIs and require human intervention and years of effort to resolve them.

The weaknesses of probabilistic patient matching will get much worse when applied to HIE networks like the QHINs for two reasons. First, HIE networks cannot impose the same degree of data format standards, data capture standards, data governance rules, algorithm tuning rules, and manual adjudication that hospitals can. So, they cannot be expected to achieve the same match rates as their contributing members. Second, it is a fundamental law of data science that data quality degrades rapidly as more disparate sources contribute to a database. So, if just one hospital in an HIE network introduces an errored patient identity into the HIE, then every member hospital within the HIE will now be at risk of mismatching that patient on all future queries.

QHINs will need Referential Matching technology to deliver Identity DNS at scale

Referential Matching can solve the problem of patient matching accuracy at QHIN scale. It does so by using a reference database of identities as an authoritative answer key against which all patient identities are compared. The reference database contains demographic information for the US population and includes a wide range of identity information for each person—nicknames, maiden names, current and former addresses, current and former phone numbers, and emails.

By comparing every patient identity to its corresponding reference identity, Referential Matching avoids the chain reaction of errors that occur when patient records must all be matched to each other. For Referential Matching to work within the TEFCA framework, it only needs to be used within the Communication Broker at the QHINs, and does not require any matching changes at the Participant level.

A National Patient Matching Resource need not be provided by a single vendor. The RCE could organize two or more Referential Matching vendors to collaborate in providing such a service. They would have to first harmonize their massive reference databases, but once done, multiple vendors could serve as part of this national resource.

We cannot keep doing what we have been doing

One thing we all should agree upon is that we cannot keep doing what we have always done and expect it to suddenly work. We have tried probabilistic matching, and it barely works at the hospital level. Even the best intentions and policies will fail if the matching technology proves inadequate, and that ultimately will dictate the success or failure of TEFCA.

Sponsored Recommendations

Enhancing Remote Radiology: How Zero Trust Access Revolutionizes Healthcare Connectivity

This content details how a cloud-enabled zero trust architecture ensures high performance, compliance, and scalability, overcoming the limitations of traditional VPN solutions...

Spotlight on Artificial Intelligence

Unlock the potential of AI in our latest series. Discover how AI is revolutionizing clinical decision support, improving workflow efficiency, and transforming medical documentation...

Beyond the VPN: Zero Trust Access for a Healthcare Hybrid Work Environment

This whitepaper explores how a cloud-enabled zero trust architecture ensures secure, least privileged access to applications, meeting regulatory requirements and enhancing user...

Enhancing Remote Radiology: How Zero Trust Access Revolutionizes Healthcare Connectivity

This content details how a cloud-enabled zero trust architecture ensures high performance, compliance, and scalability, overcoming the limitations of traditional VPN solutions...