As the care continuum expands, patient-matching remains a problem

Matching patients with their medical information sounds like a simple concept, but it’s not. The practice is plagued by such issues as typos, missing data, similar names and new addresses—resulting in match rates as low as 80% within the same facility, according to the College of Healthcare Information Management Executives. 

That means 1 in every 5 patients may not be linked with the correct record.

It also leads to higher costs. Patient-matching remains a multi-
billion-dollar problem in the U.S., with inaccurate patient identification accounting for $1,950 in duplicative medical care costs per inpatient and $1.5 million in denied claims per hospital each year, according to a survey by Black Book.

Ensuring patients are linked with the correct medical information has become a particular challenge in the era of value-based care, when health systems and insurers are tasked with not only matching patients within their own organizations, but also across the continuum of care. That makes patient-matching integral for interoperability, said Mariann Yeager, CEO of interoperability not-for-profit the Sequoia Project.

Patient-matching is a “fundamental function of being able to get the right records, for the right person, at the right time, so that timely decisions can be made about his or her health,” she said. “There has to be a mechanism to ensure that you’re actually getting a copy of the records for the right person.” 

Part of the challenge with patient-matching is that it tends to rely on staff filling in a few key fields—such as name, address and date of birth—when a patient presents at a hospital. That leaves room for human error from both staff and patients. A few seemingly small, but consequential, problems could involve inconsistencies in the way addresses are written, if a patient has recently moved to a new home, or if a name or date of birth is entered with a typo. 

“If there’s some kind of error in entering those fields, either when the patient’s coming in or in a previous entry, the matching can go awry,” said Brendan Watkins, administrative director of enterprise analytics at Palo Alto, Calif.-based Stanford Children’s Health.

To ease patient-matching struggles, many countries have issued a national ID number for patients, said Dr. John Halamka, chief information officer at Beth Israel Deaconess Medical Center and international healthcare innovation professor at Harvard Medical School.

The U.S. is not one of them. “The United States has no national healthcare identifier and has no nationwide patient-matching strategy,” Halamka said. “That means every single organization figures it out on their own.”

In fact, Congress for decades has prohibited HHS from putting funds toward developing a unique patient identifier, citing issues related to privacy and security. To work around this, the CMS in February issued a request for information seeking industry input on how it can best support private-sector efforts to improve patient-matching. The RFI was part of a broader proposed rule on interoperability.

Two of the approaches involve computer program validation—either at the software level or the algorithm level—and build on the traditional way healthcare organizations have matched patients with their records. 

This provides some benefits, according to Watkins. “There’s usually different weights associated with each field” in an algorithm, as opposed to when an employee checks an entry form manually, he said. “For example, if you’ve got the Social Security number, then it’s close to 100%” certain that the program has the right match. “But, if you’re missing it—which can be the case—there’s other weights associated with telephone number, address.”

Healthcare organizations vary in their use of patient-matching algorithms or software. But the federal government doesn’t have a program to validate such tools, a feature desired by some. A lack of consistency in the tools used at different organizations might contribute to problems with mismatched records.

Even if two organizations are using two high-quality patient-matching solutions, there’s the potential for patients to slip through the cracks.

“If I’m using an algorithm that gets 96% right, and (my colleague) is using an algorithm that’s 97% correct … it doesn’t mean the 4% and the 3% (inaccuracies) match up,” explained Ray Deiotte, chief data officer at Centura Health in Centennial, Colo. For Deiotte, that means a national patient-matching strategy would need to establish a consistent way to match patients across organizations.

But healthcare IT groups took issue with the request for information’s language of requiring a “particular” algorithm or software, arguing that it would hinder development of new solutions. That type of mandate would be a mistake, and likely “have the chilling effect of stifling innovation,” said Eric Heflin, the Sequoia Project’s chief technology officer. Instead, he suggested the CMS help the industry establish principles and best practices to inform patient-matching strategies.

The College of Healthcare Information Management Executives, too, suggested that a better way for the federal government to intervene would be by requiring vendors to share patient-matching accuracy rates with providers, so they can make informed decisions. “Our (members) have found that’s not easy to come by,” said Mari Savickis, CHIME’s vice president of federal affairs.

Another approach, establishing a set of consistent data elements for healthcare organizations to use in patient-matching programs, seemed like a useful and realistic path for the CMS to take, according to healthcare leaders—at least as a first step.

Setting standard formats for demographic data proved the “biggest opportunity to immediately impact matching rates,” according to a white paper the Sequoia Project released last year. It’s “one of the easy things we can do to dramatically increase patient-matching quality, without having a national identifier or without ripping out technology and replacing it,” said Heflin, who served as a member on the federal Health IT Advisory Committee’s U.S. Core Data for Interoperability Task Force.

The Sequoia Project also found adding supplemental identifiers, such as driver’s license numbers, helped to boost patient-matching rates to more than 95%, according to the white paper.

One of the reasons many healthcare experts support standardized data is because it provides a foundation for other patient-matching approaches to build upon—for example, establishing a core set of data for algorithms to use. NextGate, a vendor of patient-matching solutions such as an enterprise master patient index, agreed that national standards would prove helpful for its services.

“That, to us, is huge,” said Dan Cidon, NextGate’s chief technology officer. He said consistently documenting a patient’s cellphone number and current address would prove helpful for many programs. “That kind of consistency would really improve the match rate that we see today, and that’s with existing technology,” he said. “It’s really more like a process change.”

The approach, however, would require getting organizations across the U.S. to agree on a standard to use. In its request for information, the CMS suggested using the U.S. Core Data for Interoperability—a standardized set of data elements proposed by the Office of the National Coordinator for Health Information Technology.

But figuring out what data elements are best for patient-matching might be an ongoing conversation. According to a study published in the Journal of the American Medical Informatics Association, setting standards for many elements—such as date of birth, telephone number and Social Security number—didn’t improve match rates. However, standardizing addresses specifically to the format used by the U.S. Postal Service did prove helpful.

Beginning last year, the CMS kicked off its effort to replace the Social Security numbers on Medicare ID cards with separate beneficiary identifiers.

Most healthcare leaders who spoke with Modern Healthcare agreed that expanding these Medicare beneficiary identifiers agencywide would be helpful for patient-matching, but they stressed it’s just another data point.

Tom Leary, the Healthcare Information and Management Systems Society’s vice president of government relations, noted that Medicare ID cards were created with an express purpose—and it wasn’t patient-matching. The randomly generated 11-character Medicare beneficiary identifiers are meant to cut down on identity theft and fraud risks associated with sharing Social Security numbers.

“The Medicare number is a step in the right direction for security reasons, but it’s also only one of many data points about an individual,” he said, adding that although a single identifier sounds appealing, programs that integrate multiple data points will likely be more useful. “The idea of a single number being a solution isn’t something that’s right around the corner.”

That’s particularly true given the shift to value-based care and a growing interest in the role social determinants of health play in patient care. Jaime Bland, CEO of the Nebraska Health Information Initiative, said a CMS-wide identifier would likely be helpful when matching out-of-state patients with their medical information, but highlighted the work the health information exchange would still have to do to match patients with a range of organizations in the region.

“It would definitely be helpful,” Bland said of an expanded Medicare beneficiary identifier program. “But we’re trying to match across not only what healthcare is generating, but also all of the community resources, which are not traditional healthcare settings. They have identifiers of their own.”

That’s why data standards remain a more sustainable solution, from her view.

“We don’t necessarily need a unique identifier if we could just be consistent in the data elements that are collected,” she said.

Referential matching, or using external data sources to support patient-matching decisions, is a popular approach among IT vendors—it’s part of how NextGate and the Nebraska Health Information Initiative match patients, for example. The method involves using publicly available third-party data, such as information from credit bureaus and the U.S. Postal Service, to verify a patient’s identity.
 

Verato, a patient-matching startup that claims to have pioneered referential matching, uses external information from public records to create a more comprehensive view of a patient’s demographics, including their previous names and addresses. The company subsequently uses this data to corroborate its matches and reduce duplicate records held within the same organization, Verato CEO Mark LaRow said.

However, LaRow cautioned that referential matching, again, is only a piece of the puzzle.

“External data by itself is not the answer,” he said, 
noting that Verato combines third-party data with referential-matching algorithms to fuel its decisions. As a result, he—like CHIME’s Savickis—advocated for the CMS to create a way to validate the accuracy of patient-matching technologies across the board, so that vendors applying different types of approaches could be compared head-to-head.

Providers, however, voiced some concerns about using data not owned by their own organization when matching patients, particularly when it comes to connecting this information to the EHR, as suggested by the CMS.

“I think it’s intriguing, but it’s fraught with some challenges,” Marc Probst, CIO at Salt Lake City-based Intermountain Healthcare, said of referential matching. “If you don’t control those outside data sources, then you also don’t control the format of that data. When you bring it in, it may not be in the format that you think it is.”

The CMS ended its request for information with an open-ended question: “To what extent should patient-generated data complement the patient-matching efforts?” For healthcare leaders, it’s an interesting question—but still, just a question. And much of the answer depends on what type of patient-generated data the agency is referencing.

Patient-input demographic data, for example, might not prove helpful.

Healthcare organizations have found that information patients enter about themselves via patient portal often has to be revalidated, Savickis said. “People could be mis-keying information. That happens all the time,” she said. “I think there’s some work to do before that can get to a place where, from a matching standpoint, it can be used.”

Beyond standard demographic data, an emerging form of patient information that might prove useful for matching is using biometrics, or physical characteristics, to identify patients. For example, recognizing patients based on fingerprints or facial scans, similar to how many consumers unlock smartphones today.

In a recent report, the Pew Charitable Trusts found patients consistently expressed interest in using biometrics for patient-matching.

“Innovative approaches like the use of biometrics … should be examined, though there’s likely a longer implementation time,” said Ben Moscovitch, the Pew Charitable Trusts’ project director for health IT. One of the considerations healthcare organizations would have to consider with biometrics is the potential trade-off with privacy risks, depending on how data is stored and transferred.
 

Deiotte at Centura Health said as organizations begin to discuss consolidating patients’ demographic, medical, social determinants, consumer and now biometric data to fuel patient-matching decisions, there are two priorities to balance: having enough data to match the patient appropriately, while also reducing the risk of data being exposed in a cybersecurity incident.

“Our potential for risk and spillage becomes that much higher, because it’s kind of a one-stop shop for all of that information,” he said. “We have to balance the risk with the reward.”