iht2 health it chicago summit

Patient Matching Decoded A Framework for Cross-Organizational Patient Identity Management

1 ©Copyright The Sequoia Project. All rights reserved.

Mariann Yeager, MBA CEO

www.sequoiaproject.org

http://www.sequoiaproject.org/

The Sequoia Project is the trusted, independent convener of industry and government

Works to address the challenges of secure, interoperable nationwide health data sharing

2

The Sequoia Project’s Role

NATIONWIDE SECURE INTEROPERABLE

© 2015 The Sequoia Project. All Rights Reserved.

Acting in the Public Interest

As a nonprofit 501 (c) 3 organization operating in the public interest, our public-private governance process insures transparent oversight of this work. The Sequoia Project serves as a neutral, third party convener.

The practical application of our work:

• Enables consensus agreement on the policies and standards required to reduce barriers to data exchange

• Advances development and continued support for health data exchange governance frameworks

• Focuses on real-world implementation issues to advance interoperability

3 © 2015 The Sequoia Project. All Rights Reserved.

Current Sequoia Project Initiatives

The eHealth Exchange is the largest and fastest growing health data sharing network in the US

Carequality is a public-private collaborative building consensus on a nationwide interoperability framework to inter-connect networks


The Sequoia Project and Care Connectivity Consortium (CCC) Strategic Alliance

• CCC is a collaborative of 5 prominent healthcare organizations: – Geisinger (PA)

– Intermountain Health (UT)

– Kaiser Permanente (CA, OR, WA,

VA, MD, HI, GA, CO)

– Mayo Clinic (MN, FL, AZ, GA, WI)

– OCHIN (17 states)

• CCC enhances capabilities of current HIE technologies and allows for sharing between organizations and health IT systems

• The CCC aids eHealth Exchange growth by: – Serving as a test bed for

new technologies

– Contributing innovations to the

eHealth Exchange community

• The CCC participates in Carequality and serves on its: – Steering Committee

– Trust Framework Work Group

– Query Work Group

– Operations Work Group


Interoperability is a Journey

Successes

• Growing pockets of interoperability

• Enhanced care coordination is saving lives and money

• Accountable care and customer demand fuel data sharing

eHealth Exchange and other networks are growing

Carequality will bridge networks

Challenges

• Still striving for comprehensive nationwide footprint

• Exchanges need to get faster

• Need to improve format and usefulness of data

• Difficulties in patient identity matching and accuracy


Patient Identity Management


Identifying and correctly matching patient data across disparate systems and points of care

- Without a shared unique patient identifier

- Comparing demographic information about an individual

- Within an enterprise

- Between organizations

- Precedence in other industries (e.g. financial services, credit bureaus, etc.)

- Very different than matching inside an organization

What is it?

Why is Patient Matching Such a Challenge?

• Technical – Heterogeneous technologies and vendors – Data exchange latency (e.g. timeouts, lack of consent/authorization)

• Data – Data quality, completeness, consistency – Different default values (“John Doe”)

• Policy and legal – Different legal jurisdictions and requirements (such as for minors) – Different patient matching rules – Consent, security, sensitive data sharing

• Process – Different QA and human and system workflows (latency, variations, definitions, etc.) – Organizational size, resource allocation, project timelines, commitment, skill levels – Change management

• The result: – Match rates ACROSS organizations are frequently unacceptable – 10% to 30% in several cases

Exacting Change: Intermountain Case Study


Intermountain Healthcare Case Study

• Not-for-profit health system serving Utah and southeast Idaho

• 22 hospitals

• 1,400 employed physicians at more than 185 clinics

• 750,000 SelectHealth insurance plan members

• Highly innovative and progressive

• Member of the CCC


Establishing a Baseline

• Sample trial to establish baseline

• 10,000 patients known to have been treated by Intermountain and an exchange partner

• High match rate expected

• Patient analysis demonstrated only 10% true match rate

• Why?

– Data quality / lack of normalization

Initial Cross-Organizational Patient Match Error Rate


Benchmark Trial

Offline Line Performance Measurement and Refinement


Completeness: At what rate is this trait captured and available? Validity: Is this trait known to be correct as compared to the known true values?

Distinctiveness: Is the trait able to uniquely identify a person? A trait such as administrative gender, for example, is not associated to a single individual. A trait such as an enterprise master patient index (EMPI) number is distinctive.

Comparability: Is the trait readily, programmatically, and accurately matched with the same trait at another organization? An address is an example of a relatively difficult to compare trait, whereas a social security number (SSN) can be easier to compare.

Stability: How much does the trait remain constant over a patient’s lifetime? Although examples exist to the contrary, traits such as gender, birth date and SSN are generally consistent over time.

Identifying Best Patient Match Attribute

Patient Attributes Analysis


Attribute Name Completeness Validity Distinctiveness Comparability Stability

EMPI # 100% -- 100% Very High Very High

Last Name 99.85% 99.84% 5.1% Medium High

First Name 99.85% 99.33% 3.1% Medium High

Middle Name 60.54% 60.54% 2.6% Medium High

Suffix Name 0.08% 0.08% 0.08% Medium Medium

SSN 61.40% 60.92% 98.0% High High

Sex (Admin Gender) 99.98% 99.98 0.00008% High High

Date of Birth 98.18% 97.38% 0.8% High Very High

Date of Death 3.36% 3.36% 3.4% High Very High

Street Address

(1 or 2)

95.00% 94.61% 44.4% Low Low

City 94.84% 94.83% 0.8% High Low

State 94.81% 94.39% 0.8% High Low

Facility MRN 99.90% 99.90% 99.90% High Low

Postal Code 92.31% 92.0% 0.6% High Low

Primary Phone

Number

90.68% 87.26% 51.6% High Medium

Work Phone

Number

20.28% 19.79% 51.6% High Low

Ethnicity 25.25% 25.25% 0.0003% High Very High

Race 76.25% 76.25% 0.0001% High Very High

Applying Updated Algorithms

Rules Created from Analysis Results

1. First name, last name, date of birth, gender & telephone number

2. First name, last name, date of birth, gender & zip code

3. First name, last name, date of birth, gender & last 4 digits of SSN

4. First name, last name, date of birth, administrative gender & middle name

5. First name, last name, date of birth, administrative gender & first character of middle name

Updated Algorithm Expected Result: Cross-Organizational Patient

Match Error Rate


This was the expected match rate for the benchmark trial

Examining the Remaining Error Rate

Detailed Analysis of Error Rate Actual Result: Updated Algorithm and Data Quality

Cross-Organizational Patient Match Error Rate


Analysis of error rate uncovers actually 15% errors; other issues

triggered failures but were unrelated to patient matching

errors

Optimizing Patient Identity Management

Additional Strategies for Raising the Bar Nationwide

• Re-use knowledge

• Apply results of prior work

• Standardize formats is biggest and fastest opportunity

• Determine minimal acceptable match rate

• Proactively manage fragile identities (e.g. with partial info)

• Improve the human workflow

• Involve patients in managing their identities

• Data integrity (99.99%) requires supplemental identifier

• Leverage CCC Shared Services

Final Cross-Organizational Patient Match Error Rate


Is Your Patient Identity Management Working?

Top Questions to Ask Your Organization

• Are our staff trained and actually capturing high-quality patient identity data?

• Are all our patient demographics data as complete as possible?

• Are we capturing the telephone type as well as the number itself?

• How do we handle patient consent with respect to patient matching?

• More topics covered in the white paper


Cross-Organizational Minimal Acceptable Principles

Overview of Proposed Framework

Traits & Identifiers


Matching Algorithms Exception Handling

Cross-Organizational Maturity Model

Overview of Proposed Framework

Level 0 • Ad hoc

• No oversight

• Unpredictable matching results


Level 1 • Basic

processes

• Limited oversight

• Becoming predictable

Level 2 • Increasing

algorithm use

• Quality metrics gathered

• Consistent quality

Level 3 • Advanced

technologies

• Management controls quality metrics

• Highly optimized

Level 4 • Ongoing

optimization

• Active management

• Innovating

• Leading industry

Next Steps

• Review patient identity management white paper with study findings, principles and maturity model

• Provide public comment

• Register for the webinar and learn the top 10 things you should be asking your organization to make strides in patient matching

• Be part of the Sequoia community!

• For more information, or to receive a copy of the paper, email us at:

– admin at sequoiaproject dot org


Questions?

For more information: www.sequoiaproject.org


http://www.sequoiaproject.org/