BCBSA Patient-Matching Algorithm for Payer Community Hits 99.5% Accuracy

A new case-study report published by The Sequoia Project and the Blue Cross Blue Shield Association (BCBSA) reveals that the health insurance conglomerate has developed a patient-matching algorithm for the payer community with a 99.5 percent accuracy rate.

The nonprofit Sequoia Project is a public-private collaborative chartered to advance implementation of secure, interoperable nationwide health information exchange. The organization recently collaborated with BCBSA to apply A Framework for Cross-Organizational Patient Identity Management for the payer community and develop person matching strategies. This week, The Sequoia Project published Person Matching for Greater Interoperability: A Case Study for Payers, “which demonstrates high matching accuracy rates, and provides actionable insights for improving person identity matching across the payer community, a critical component of successful health information exchange and interoperability,” officials stated.

The case-study paper documents the journey of a specific payer entity to developing an algorithm with a 99.5 percent accuracy rate across 36 independent companies, the concepts and practices described can be applied to a broader payer population, officials said.

“The ability to match someone with their health data – regardless if they’ve changed insurers – is critical to ensuring people receive the care they need and deserve,” said Rich Cullen, vice president at BCBSA. “To address this health industry need, we developed a way to safely and securely match a person’s health data from one Blue Cross and Blue Shield company to another. We believe this will lay the foundation for larger health data-sharing efforts within the broader healthcare system.”

As detailed in the paper, the specific challenge addressed was the need to link healthcare experiences of an individual person across the entirety of the BCBS system and to prepare for future interoperability requirements that may mandate sharing with external stakeholders via national networks. For example, when a member (e.g., John Doe) is covered under one BCBS company and then transitions to another, the rich history from the previous BCBS experience is not easily available for the new BCBS company to draw from in order to improve care outcomes for the member, as explained in the paper.

As such, BCBSA developed its Member Matching Index Identifier (MMI) solution to benefit its individual BCBS companies. As officials noted in the paper, “While the focus of this paper is on the experience of the BCBS system, the concepts are useful across the payer ecosystem since all payers experience the need to link members within their own companies and when they take on new accounts.”

As explained in the report, two BCBS companies participated in the pilot phase of the project. These companies submitted specific data elements through a secure file transfer process and throughout this phase, multiple iterations of data elements were tested in combination to yield the best matching results. This continued as the companies provided different data elements to the point of diminishing returns—at which time the introduction of additional data elements did not create any positive impact on the results, and in some cases reduced the matching success rate.

One lesson learned during the pilot phase was that email addresses and phone numbers can cause “over matching.” For example, the report described, a mother may use her email address and phone number for her minor children. The consumer reference data vendor’s algorithm was tweaked to give a lower weight to these data elements, and it was also identified during the pilot that a consumer reference identifier would not address assignment of the MMI for minors. The outcome of the MMI process is the assignment of a unique number that is used to link a single member across any and all BCBS companies from which coverage has been provided.

Per the report, the MMI solution went through multiple iterations including a proof of concept phase and pilot testing phase before moving to production, according to the paper. Early on, the use of a referential consumer data vendor was determined to be a critical part of the solution. Before developing the MMI process fully, a referential consumer data vendor was selected through a request for proposal, and it was also determined that three years of history would be required as BCBS companies on-boarded their data. Some use cases, for example the Healthcare Effectiveness Health and Data Set (HEDIS), required a look back period spanning many years.

Another key factor that informed the development of the MMI solution was the decision that for security reasons the matching algorithm would not use a Social Security number (SSN), as explained in the paper. In addition, BCBSA found the SSN was sometimes unreliable and often unavailable for person matching purposes. For instance, member names and addresses were identified as being specifically necessary to assign the MMI, which was a variance from prior BCBSA data warehouse requirements that did not demand this data. In summation, the decision to link members across BCBS companies in lieu of utilizing the Golden Record approach was foundational to the formation of the MMI solution, officials stated.

The implementation of the MMI solution was also subject to an accuracy goal. Although BCBS companies informed BCBSA that a 95 percent accuracy rate was fairly easy to achieve—a rate that is typically sufficient for analytical purposes—a higher accuracy rate was ultimately preferred. BCBSA’s use cases, such as for care management, were member-facing, therefore a higher accuracy rate of 99.5 percent was established to ensure the most accurate data BCBSA could provide.

Ultimately, the MMI is a hybrid solution that uses a vendor’s probabilistic engine complemented by rules specific to payers, according to the paper. For example, no two members of a contract can have the same MMI and no data warehouse member identifier at its lowest grain (Plan/Member/Product) can have multiple MMIs. This solution and the complementary matching rules were developed because none of the existing market solutions met the intended need of this project, officials said.

In the paper, officials did acknowledge that data quality issues were encountered during the implementation of the MMI solution. Each BCBS company establishes its data quality strategies and metrics. Therefore, the quality of the patient matching data sent to BCBSA can vary from company to company. As a result, part of the onboarding process for this effort included a data quality assessment, and, if required, remediation by the individual BCBS companies, the paper stated.

“Since our provider-focused framework was published in 2016 and revised in 2018, we’ve seen tremendous interest in how we apply those principles to raise the floor for interoperability,” said Mariann Yeager, CEO of The Sequoia Project. “When the Blue Cross Blue Shield Association agreed to collaborate on the application of these principles to the unique needs of the payer community, we were thrilled for the opportunity to work together to expand our thinking from ‘patient matching among providers’ to ‘person matching in other settings.’”