Health plans have invested heavily in building robust data infrastructure over the past decade, enabling them to store, exchange, and model data. However, this infrastructure often lacks a critical component: the ability to accurately connect data records to the same individual. This connection, known as identity, is the foundation upon which all other data analysis is built.
A plan can run any model, but if it can’t determine whether the member in the enrollment system is the same member in the care management system, the results can’t be trusted. According to the report, poor data quality costs organizations an average of $12.9 million per year.
Getting Identity Right
Resolving identity issues comes down to three key capabilities: recognizing the same person over time, seeing who and what is connected to them, and standing behind every answer. Recognizing a returning member as the same person, with their history intact, is the first capability. This is essential, as every model a plan runs, from risk stratification to fraud detection, inherits the quality of the data beneath it.
When records are fragmented or duplicated, errors compound across every downstream decision. Seeing the member whole is what lets a plan document conditions completely, across providers and pharmacies, and back through prior enrollment. This, in turn, enables the plan to identify members who would benefit from care management.
Seeing Connections
The second capability is seeing the people and organizations around the member, such as the caregiver who manages appointments, the household the member belongs to, and the provider. Linking members who share a household is a challenging task, but financial systems have successfully implemented similar capabilities.
The payoff is that members stop doing the work the plan was supposed to do. Today, a member repeats their medical history at every new visit and carries results from one office to the next. A plan never wants to hear that it lacks information about a member.
Connection also shapes the moments that decide how a member feels about the plan. A prior authorization is decided on the provider’s actual network status and the member’s own history, so care isn’t delayed by a wrong denial. Or the member isn’t charged out-of-network rates for a doctor the directory listed as in-network.
Standing Behind Every Answer
The third capability is that every output should trace back to the data that produced it, the transformations applied along the way, and the rules that shaped it, with someone accountable for the result. This is critical, as the slowest part of deploying AI inside a health plan is getting compliance and clinical reviewers to trust the output enough to sign off.
In June 2024, the largest Medicare Advantage organizations overturned 95% of appealed prior authorization denials for skilled nursing facility admission, a rate federal reviewers said raises concerns about the initial denials themselves. A reversal rate that high means the plan could not stand behind its decisions.
A plan that skips this and goes straight to deployment simply gets confident wrong answers at scale. Instead, plans should start where the data enters, validating and standardizing identity in real-time, at onboarding and at live call intake.
The full effort is a multiyear project that takes real capital, and leaders serve their organizations better by saying so, including how much time it takes and what it costs in budget and staffing. The industry has shifted from careful stewardship of data toward something closer to careful freedom.
Thomasina Anane is the associate vice president, enterprise analytics at the Alliance of Community Health Plans; Vinay Kulkarni is the chief information officer of SCAN Health Plan; and Martin Hougaard is general manager, product marketing, at Verato.
The importance of getting the identity layer right cannot be overstated. It is the foundation upon which all other data analysis is built, and it has a direct impact on patient outcomes, business success, and regulatory compliance.
In the end, it all comes down to one simple question: Who is this person, and who are they connected to? By answering this question accurately and consistently, organizations can unlock the full potential of their data, improve patient outcomes, and achieve business success.
