Identity and Attribution in Data

Two technical problems sit underneath every population metric, and both are easy to underestimate: knowing which patient a record belongs to, and knowing whether that patient is yours. Patient matching and attribution are the identity layer of analytics, and getting them wrong quietly corrupts everything built on top.

Patient matching

The same patient appears across claims, multiple EHRs, ADT feeds, and lab systems, and there is no universal patient identifier tying those records together.

Worth remembering: matching records to the right patient is a foundational, permanent, imperfect chore. Names change, addresses vary, dates of birth get mistyped, and the same person appears with different identifiers in different systems. Match too loosely and you merge two people into one, blending their histories; match too tightly and you split one person into two, fragmenting theirs. Both errors distort every downstream metric. Identity matching is never fully solved; it is monitored and improved continuously, like data quality.

Attribution in the data

Once you know who a patient is, you have to know whether you are accountable for them. This is attribution, covered as policy in earlier courses, implemented here as data logic:

• The rules from the ACO and contracting courses, claims-based, voluntary, enrollment-based, prospective or retrospective, have to be computed from the data.
• The analytics function maintains the attributed list, applies the contract’s attribution logic, and tracks how it changes over time.
• Getting attribution logic wrong means investing analytics and care management in the wrong patients, or missing the right ones.

Why this layer is decisive

Identity and attribution are the join keys of the whole system. Every metric, total cost of care, risk score, care gap, is computed per patient and per attributed population. If the patient records are mismatched or the attributed list is wrong, the metrics are wrong no matter how sophisticated the analysis on top. This is why mature analytics functions invest heavily in getting identity and attribution right before trusting anything computed from them.

Key takeaways

• With no universal patient identifier, the same patient appears inconsistently across sources; matching them is foundational, imperfect, and continuous.
• Attribution logic implements the contract’s rules in data, maintaining the accountable list and tracking its changes.
• Identity and attribution are the join keys of all analytics; errors here corrupt every downstream metric, so they are worth getting right first.