Sunteți pe pagina 1din 7

Unifying Patient Identity Across

Organizational Boundaries: A Proven Approach


Executive Summary
The multiple providers that care for a patient population often have separate electronic health
record (EHR) systems, which mean that patients’ health records are spread across multiple
databases. Unifying them into a cohesive record is necessary for patient safety, coordination of
care and population health management. Yet, accurately identifying a patient across disparate
EHR systems and linking his or her records together into a logical whole presents many
technical and operational challenges.
Many organizations use Enterprise Master Patient Index (EMPI) technology for matching patient
data across organizational boundaries, but this technology was primarily designed for use in
centralized, unified systems, not the complex landscape of hospitals, physician groups, state
exchanges and other entities that exist today. EMPI forces disparate organizations to adopt
a master identifier and requires slow and costly centralized processing of probable matches.
A different approach, based on data architectures that use graph structures,1 can match
distributed information and eliminate the need for independent entities to adopt a unified
identifier. This helps healthcare organizations of all types and sizes coordinate care to improve
outcomes, reduce costs and enhance the patient experience.

Author
Chris Ingersoll
VP Clinical Architecture at RelayHealth

1 An informative introduction to graph databases can be found at http://en.wikipedia.org/wiki/Graph_database

© 2015 RELAYHEALTH AND/OR ITS AFFILIATES. ALL RIGHTS RESERVED.


Unifying Patient Identity Across Organizational Boundaries: A Proven Approach

Patient Identity Is Key to Unlocking EHR Value


Over the past decade, the healthcare industry has made tremendous leaps in the digital quantification of patient health.
Provider adoption of EHRs has grown from about 20 percent in 2004 to nearly 80 percent in 2013.2 This growth has been
fueled by a belief shared by providers and legislators that digitization at the point of care is a foundational step toward
achieving what the Institute for Healthcare Improvement (IHI) calls the Triple Aim: improving the patient experience and
population health while lowering costs.3

Figure 1: The IHI Triple Aim

Population

Experience of Care Per Capita Cost

However, the EHR landscape is highly fragmented. In 2013, there were 556 unique EHR vendors with meaningful use (MU)
attestations—representing an annual increase of more than 25 percent.4 A given patient population might be represented
in dozens of different systems. At each point of care—physician’s office, hospital, labs and radiology, pharmacy, long-term
care, home health—providers are recording the clinical codes, measures, attributes and notes that summarize patient
health. However, unifying that data into a 360-degree view of patient health is proving to be a challenge.
Let’s take a typical example. Bob Nguyen sees his primary care provider a couple of times a year, and an associated
specialist in the same group once a year (silo 1). Bob needs to have a biopsy, so he visits a hospital where an affiliated
physician treats him (silo 2). However, the lab that performs testing is not part of the hospital (silo 3). Neither is the
pharmacy where he picks up some antibiotics (silo 4). Later that year, he sees a physician in another city (silo 5). Bob’s
medical record is now stored across five separate, possibly incompatible data silos.

2 http://www.cdc.gov/nchs/data/databriefs/db143.htm
3 http://www.ihi.org/Engage/Initiatives/TripleAim/pages/default.aspx
4 http://www.softwareadvice.com/medical/industryview/ehr-meaningful-use-market-share-2014

© 2015 RELAYHEALTH AND/OR ITS AFFILIATES. ALL RIGHTS RESERVED.


Unifying Patient Identity Across Organizational Boundaries: A Proven Approach
Figure 2: Lack of interoperability and a fragmented EHR landscape result in data silos across the care continuum.

Independent
Affiliate Practices

Regional Hospitals Physician


& Health Systems Associations

Accountable Care State Health


Organizations Exchanges

Pharmacies
& Labs

Efficiently matching patient identities across these barriers is critical to achieving better outcomes, reduced costs and
improved patient experiences. The more information his medical providers have about Bob’s healthcare history, the better
care they can deliver, whether it’s avoiding a medication allergy, enabling team-based care, or simply having test results
easily at hand. More broadly, population-level health data can be used to identify care gaps and advance
evidence-based medicine.
However, correctly matching a patient’s healthcare records is not a simple matter, even within one health system. Bob
Nguyen’s records must not be mixed with those of another patient with the same name, yet some of Bob’s records might
be stored under his full name, Robert S. Nguyen, Sr. The proliferation of multiple EHR approaches and products makes
it an even more challenging task. The only true national identifier, the social security number, cannot be used in most
healthcare contexts due to privacy concerns. Even if it could, it does not provide complete coverage of the US population,
especially in regions with large immigrant populations.
One way to solve the issue is for all parties to standardize on the same software and data structures. Large hospitals and
groups often have the leverage to unify EHR technology across hospital-based and ambulatory sites. However, acquiring
and deploying software on such a scale is costly, and can also be a financial and operational burden for acquired providers
who must switch from their already-deployed EHR solution. Then there is the proportion of care provided by independent
physician groups, who may not have the resources or the desire to adopt the systems and procedures of the
central organization.
Another way of addressing the issue is a Health Information Exchange (HIE). An HIE is an information clearinghouse that
translates information across EHR silos, making them an interoperable network. Governance of the exchange may be
managed by a health system, physician network, or an independent entity. Depending on the solution architecture, the
operational administration of the solution may be a centralized function or a responsibility that is shared by stakeholders.
HIEs are a promising solution, but they still must overcome the problem of matching patient identities across entities.

© 2015 RELAYHEALTH AND/OR ITS AFFILIATES. ALL RIGHTS RESERVED.


Unifying Patient Identity Across Organizational Boundaries: A Proven Approach

Identity Matching Challenges in the HIE


When enabling HIE, identity matching may be handled by EMPI technology. EMPI establishes a new, community-based
identifier for each patient (the term refers to both the technical infrastructure and the identifier it establishes). The
managing entity’s identifiers and local identifiers (those at organizations other than the main entity managing the system)
are linked to the EMPI in a tree-based structure. Local identifiers are linked to each other through the root EMPI identifier.

Figure 3: EMPI’s tree structure requires centralized matching at the scope of all potential identifi­ers.

EMPI’s tree structure requires centralized matching at the scope of all potential identifiers.

EMPI is essentially answering the question “What is the enterprise identifier to use when sharing data?” Systems are
asked to resolve their internal, localized identifier to this master identifier before sharing data. This model works well for
systems within a given enterprise, but not so well across independent organizations. Take the example of an independent
physician group that serves multiple health systems, each with its own EMPI strategy. The group has to connect to
multiple EMPI systems and manage a unique set of identifiers for each one. If the patient population using a given EMPI is
small enough, a practice may not adopt the EMPI schema at all; leaving a gap in the patient’s records that is not
easily addressed.
Another significant patient identity challenge is that matches are frequently uncertain. For example, is Bob Nguyen the
same person as Robert Nguyen? To answer these questions, EMPI uses sophisticated algorithms to analyze demographic
information. If a match is within a given confidence threshold, it can be automatically performed. If it is not, it is sent to a
work list, where an administrator can use specialized tools to resolve it.
For the operational work of resolving a potential match, EMPI’s tree structure requires a centralized team to perform this
function. There is no way to show a local organization only the potential matches that are relevant to its operations. All
connections are always visible because they are rationalized to the common identifier. Organizations often have one or
many employees tasked with reviewing lists of these potential matches and determining if they are the same patient
through additional research. The centralized model complicates the issue of responsibility and costs in a
multi-stakeholder network.
Also, consider an independent physician associate (IPA) that connects to multiple other entities. A probable match
between a patient at the IPA and other entities must be separately resolved by each of those entities. The IPA has no
ability to resolve its own matches and must wait for the partner health systems before getting access to a unified health
record. However, the value of matching is shared by all organizations involved, not just the ones doing the matching.

© 2015 RELAYHEALTH AND/OR ITS AFFILIATES. ALL RIGHTS RESERVED.


Unifying Patient Identity Across Organizational Boundaries: A Proven Approach
Another problem with centralized match resolution is that it slows the care process as providers wait for a separate entity
to verify matches. Consider a situation in which a patient sees a provider affiliated with a health system. The patient
is flagged as a probable match to a patient record in that health system. The patient breaks her arm and goes to the
health system’s Emergency Department (ED) before the probable match has been resolved. The ED is a high-transaction
environment where throughput is key; not a scenario where it is feasible to wait for a separately operated function to
resolve the potential match.
The ED may identify the affiliated record through a Record Locator Service (RLS). However, it does not have the tools,
training or time to resolve the flagged probable matches in the EMPI. Following the ED encounter, the patient goes to
the affiliated provider for follow-up. Because the physician and hospital records were never linked during the ED visit, the
affiliated provider does not have efficient access to the clinical information generated by the ED clinicians.

Graphs, Not Trees: A Better Model for Patient Identification Across Data Silos
Instead of forcing providers to support multiple, complex EMPI schemas, it is possible to use a different
information sharing architecture that simplifies the problem of sharing data. EMPI keys all patients records back to
a “root” master identifier. An alternative data structure known as a “graph” treats certain entities as static and tracks
the changing connections among them. In this case, it links the local identifiers connected to a given patient.
When new identifiers are added, they are simply linked to the original entity and all the identifiers used by
other organizations.
A graph structure avoids the complexity inherent in EMPIs by avoiding the question of which identifier to use
entirely. Instead, it simply links one identifier to another through a static entity. The EHR systems involved do
not need to share, or indeed even know, the identifiers used by others. This model does not necessarily replace
existing EMPI systems. The EMPI identifiers are simply mapped into the graph structure, alleviating the challenges
of matching.

Figure 4: EMPI tree structure vs. XMPI (Cross-Community Matching Patient Identity) graph structure

Requests to EMPI tree structure requires all identifiers XMPI graph structure links identifiers for the
for the same patient be linked to a common root. same patient directly to each other.

© 2015 RELAYHEALTH AND/OR ITS AFFILIATES. ALL RIGHTS RESERVED.


Unifying Patient Identity Across Organizational Boundaries: A Proven Approach

Enabling Distributed Match Resolution


The graph structure also overcomes the problem of centralized match resolution. Instead, it enables probable
match resolution to be performed at the point of care. Because there is no master root identifier to resolve to,
each stakeholder can manage its own work queue, consisting of flagged matches between its records and those
of other entities. When a match is resolved by one entity, it is resolved for both. The two enterprises do not have to
duplicate the work of matching the patient to the master record.

Figure 5: XMPI graph enables each organization to resolve only its own potential matches.
XMPI MODEL

By embedding the work of resolving potential matches as a function of a RLS, a graph structure enables the match
to be performed at the point of care, such as in the ED described above, rather than forcing providers to wait
for the central administrator to verify the match. When the patient goes for follow-up, the link has already been
established and the attending physician has immediate access to the clinical data generated by the ED visit.

The Identity Graph Regionally and Nationally


There are many examples of multi-stakeholder communities that could benefit from the distributed match
resolution process enabled by the identity graph. For example, a regional care collaboration network could
connect a health system to independent affiliated physicians, linking patient identifiers to unify the patient clinical
record. A full picture of patient health emerges, enabling providers to identify gaps in care, giving researchers
more complete data for the development of evidence-based approaches, and providing patients with access
to their complete personal health records. The aggregated record also helps ambulatory providers increase the
efficiency of referrals and consultations, and enables a hospital’s ancillary businesses to connect to physician
practices with lab, radiology, and pathology orders and results.

© 2015 RELAYHEALTH AND/OR ITS AFFILIATES. ALL RIGHTS RESERVED.


Unifying Patient Identity Across Organizational Boundaries: A Proven Approach

At the national level, the identity graph has transformative potential, enabling a base-level linking of patient
records not possible with any other presently available method. Take the example of a patient who is on vacation
in a different state, gets injured, and goes to the local hospital ED. This patient has significant medical history
locked in his home health system’s EHR, including allergies and current medications. Now new significant medical
history is created as a result of this vacation ED visit.
It is not presently feasible to establish a shared identifier between these entities, or to provide a national central
patient matching resolution process. With the identity graph, however, it is possible for the ED administrator to
discover the patient’s “home identity” and link it to the individual’s “vacation identity.” With this base-level linking of
the patient records, data can flow efficiently among the members of this truly national care team.

Conclusion
Matching patient information across data silos is a critical component of EHR effectiveness. The EMPI approach
works best within an enterprise. It tends to break down when it is used to unify independent organizations
because of its centralized operational model and the difficulty of standardizing on a single patient identifier. A
graph structure works better in this context because it establishes pair-wise links between local identifiers rather
than linking them to a master identifier. This enables match resolution to be distributed across organizations and
performed at the point of care, and it eliminates the need to enforce a common identifier across
independent entities.

About RelayHealth
RelayHealth provides clinical connectivity to physicians, patients, hospitals and more using innovative health
information technology. RelayHealth’s HIE platform performs patient matching using an identity graph technology
called Cross-Community Matching Patient Identity, or XMPI. Probable match queues are automatically populated
and can be administered centrally or by designated resources at each participating facility. These queues are easy
to use, requiring minimal training, and are built into the collaborative workflows that drive patient care. XMPI also
provides national connectivity through the CommonWell collaborative platform that is used by major health IT
vendors (www.commonwellhealthalliance.org). Learn more about how RelayHealth is maximizing the value of EHR
for all healthcare stakeholders. For more information, get in touch at http://www.relayhealth.com/contact-us.

© 2015 RELAYHEALTH AND/OR ITS AFFILIATES. ALL RIGHTS RESERVED.

S-ar putea să vă placă și