Mastering the Patient Index
Featured in For the Record
Ideas on Optimal Approaches and Best Practice Considerations
Patient matching challenges have wreaked havoc on the health care industry for years. In 2020, a BlackBook survey found that 1 in 5 patient records is a duplicate, limiting the ability of health care organizations to establish accurate case histories.
Despite concerted focus by stakeholders to improve the outlook, a single solution to the problem has remained elusive. Karen Proffitt, MHIIM, RHIA, CHP, vice president of data integrity solutions with Just Associates, believes progress has been slow, despite the attention the issue has received over the last decade.
“For example, third-party data have been beneficial for validating or resolving possible duplicate records and overlays primarily after the fact, but upfront matching/linking rates continue to be suboptimal. And inaccurate, incomplete demographic information continues to pose challenges to accurate patient identification and matching,” Proffitt says.
Examples include the dramatic decline in the capture of Social Security numbers, which are a key matching field in most health care algorithms. And the issue will only get worse as the concerns resulting from poor patient matching/identification are further amplified by the increasing amounts of data being exchanged.
“Basic and deterministic algorithms like those used in most EHRs today aren’t sensitive enough to match patients when there are multiple data discrepancies in key demographic fields,” Proffitt says. “It’s a problem magnified by recent public health emergencies—for example, temporary testing sites where minimal patient demographic data are collected by people who are not typically trained for proper registration and scheduling procedures.”
The dysfunction of health care’s duplicate record challenge was on full display during the pandemic, elevating the discussion as providers sought solutions. Todd Goughnour, MBA, RHIA IT/HIM, a consultant with e4 services, says duplicates stemming from patients registering for tests and vaccines—both for themselves and family members—at multiple points along the continuum exacerbated the challenge.
“We had probably every single client call us for additional support with their enterprise master patient index [EMPI] and for merging duplicates in general,” he notes. “I think that message is getting out there from the patient identity integrity, data integrity standpoint.”
Health care organizations are beginning to understand the bigger picture of patient matching, Goughnour adds, noting that it is much more than just duplicate records and understanding what downstream systems are tied to merged records. The pandemic illuminated the need to peel back the layers and understand why duplicates were happening.
Joaquim Neto, chief product officer with Verato, agrees, pointing out that the pandemic has accelerated a shift in emphasis on patient matching. “It used to be that health care data management primarily involved reconciling patient records across a few EHRs and practice management systems, along with connected ancillary systems, from facilities where patients received care,” he says, noting that health systems are much more complicated now. “In the past few years, the level of innovation occurring in the health care IT space as well as the continued move toward value-based care and the increasing role of the consumer in health care has made the health care data landscape much more complex. Now, there are many sources of data besides the EHR and many more applications that rely on a 360-degree view of the patient.”
Approaches to Managing the EMPI
Because modern health systems are more complex, Proffitt suggests a centralized model for managing EMPIs is optimal. “Technically speaking, the ideal is managing EMPIs centrally by sharing one core patient database, which is the case for most Epic, Cerner, and other popular EHR clients,” she explains. “Numerous problems can arise when multiple core MPIs are used, such as one for inpatient and another for ambulatory clinics. For example, HL7 messaging or updates overwrite data on one MPI but not the other.”
Another option, Proffitt says, is to implement an EMPI that is “bolted” on to other core EHR or clinical systems so all patients can be managed at the enterprise level—an approach that she says is particularly advantageous when there are several key disparate systems in use.
Goughnour finds overlaying an EMPI will produce more optimal results than trying to manage duplicates with an EHR alone. “These solutions will identify almost 50% more potential duplicates, and they’re going to weigh them,” he notes. “There’s a much better analysis of the EMPI itself using these more advanced algorithms because while the EMRs have their own built-in logic to identify potential duplicates, there is a lot that is not taken into account in that situation.”
For example, take the name “Danielle,” which can be spelled several ways. Goughnour says most EHRs are not equipped to detect nuanced differences such as spellings, but a separate EMPI solution typically offers robust functionality that can achieve this level of granular analysis.
The value of using EMPI software as an overlay to an EHR increases exponentially in line with the size of an organization and the number of facilities and systems that must be managed, Goughnour says. In today’s merger and acquisition climate, these complexities are even more pronounced.
“When it’s a really large organization, having a centralized EMPI is extremely valuable. We still want to attack it, where you clean up the individual systems themselves, and then you bring them all together,” Goughnour explains.
Goughnour points to a large provider in central Florida that was acquiring hospitals rapidly. At one point, the health system was looking at 460,000 potential duplicates. Patient matching challenges had to be addressed such that an accurate patient record could be accessed from any clinic or hospital within the organization’s umbrella. This required addressing duplicate record issues across upward of 20 downstream systems.
Neto suggests that cloud infrastructures can help health care organizations of all sizes capitalize on EMPI solutions. “That’s because cloud solutions are making it possible for rural organizations to stand up EMPI programs quickly, easily, and reliably with accurate results. Cloud solutions also are a good fit for organizations with limited IT resources,” he says. “Organizations must have the internet bandwidth to support a cloud-based solution, but these days that is less of a factor.”
Industry experts suggest weighing the pros and cons of the cloud. For example, Proffitt notes that “on the plus side, small community facilities may not have a platform or need for a larger EHR system but would still benefit from the functionality one offers.” While cloud-based systems address this conflict and provide secure connectivity during data transmissions, she warns that there are always privacy concerns with cloud-based solutions.
“It can also be challenging to conduct security risk assessment due diligence on hosting vendors,” Proffitt says.
Neto believes that at this point, the industry consensus seems to be that the pluses outweigh the minuses of the cloud as long as an organization is working with a HIPAA-compliant and HITRUST-certified vendor. “Some advantages of the move to the cloud include simpler and faster deployments, fewer demands on your already stretched workforce, improved agility and scalability, and increased levels of reliability,” he says. “With true software as a service [SaaS] solutions, you usually lose some ability to control what the underlying technology components are, and you likely have less ability to customize the software itself, but often, it’s those very constraints that help ensure the SaaS solution is available to you 24/7 with minimal maintenance.”
Getting Accuracy Right
To check an EMPI’s accuracy, Proffitt suggests conducting manual reviews of matches and possible duplicate records. In some instances, performing sampling audits to determine false-positive rates is necessary. “Another indicator for missed matches is understanding and analyzing potential MPI errors that are identified by other sources, such as when clinical or registration staff alert HIM to errors that were not identified by existing matching algorithms,” Proffitt says.
Patient matching is a data-hungry process by nature. Consequently, the accuracy of an EMPI is only as good as the data going into it, Goughnour emphasizes, noting that “typically the responsibility for good data falls back to the folks who are registering patients and entering the demographic data.”
He says organizations should determine whether registration is looking up patients correctly and adopt fundamentally sound policies and procedures to ensure data integrity.
Goughnour, who recalls an instance at a rural hospital in which a security guard was registering patients, says health care organizations must be proactive and vigilant that policies and procedures are being adhered to. While that is an important first step, they should also create a level of accountability. Are there metrics in place? Can the origins of a duplicate record be tracked?
Unfortunately, in many cases, key data that can inform process improvement initiatives are missing, Goughnour says. “You know, if we get that list of potential duplicates, or even the EMPI itself, we can say, ‘OK, out of your 1 million patient EMPI, 20,000 of these flag duplicates because you’re all nines on the Social Security [number].’ Tackle that first—you want to capitalize on the low-hanging fruit and go from there,” he says. “There are ways to extract data out of EMPI to get the full picture.”
For example, Goughnour recommends utilizing gender as a matching criterion, a best practice used in many organizations. “Standardizing those policies and procedures will help with accuracy down the line. But to check on how accurate your EMPI is now, you need to look at that data and really break them down,” he says.
According to Neto, benchmarking exercises that compare patient matching against a reputable alternative are some of the best ways to ensure accuracy. “No two EMPI algorithms are identical, so there will always be some variation. However, it’s fairly straightforward to look at the subset of cases where the existing EMPI gave a different answer than the benchmark and then assess how well your EMPI is performing when it comes to false-negatives, missed matches, and false-positives, or ‘over matches,’” he notes. “These types of checks offer a gut check for organizations that have some uncertainty about their current accuracy levels, those that suspect they have an issue with patient matching, or those that are actively looking to make a technology change.”
Neto says significant shifts are occurring in that health care organizations are no longer implementing EMPIs for the sole purpose of keeping EHR data clean—though he is quick to emphasize that these tools still serve that purpose. “Now, it’s often being treated as a strategic program with a goal to create and manage a complete view of the patient that extends well beyond clinical information and well outside your facility’s four walls,” Neto says.
For example, organizations are using the EMPI to manage direct-from-consumer information, from digital front door initiatives to data from new customer relationship management platforms. Data from health care delivery partners or through collaborations with health plans also raise the level of external data being integrated within the health system.
Because these data did not originate in the EHR, Neto says that they often require different governance strategies than those that were originally established for a traditional EMPI program. “That’s why it’s important to take a fresh look at your existing EMPI program as a whole, not just your technology,” he suggests.
Proffitt believes current movements will continue to push data standardization, which will, in turn, move the needle for improved patient matching. One example of these efforts is the Office of the National Coordinator for Health Information Technology’s Project US@, an initiative in collaboration with HL7 and others for unified specification for address standardization.
“In the final analysis, a national patient identification strategy will need to be comprised of several solutions to be fully effective. There are also numerous technical, legal, political, and ownership/governance considerations,” Proffitt says. “Ultimately, there is no silver bullet that can resolve patient matching and identification on its own.”