In the Age of AI, Interoperability Isn’t Enough: Why Healthcare Needs Shared Understanding, Not Just Shared Data

Healthcare has made real progress on interoperability, but we are still solving the wrong problem. We have APIs, standards, and more data moving between systems than ever before, yet the industry continues to struggle with something more fundamental: what that data actually means.   

The issue for AI is interpretation and context, and not connectivity alone. This gap is highly evident and measurable in medical coding, the point where clinical interpretation is translated into financial outcome. A recent BlueCross BlueShield Association report on rising coding intensity claims it identified $663 million in additional inpatient spending, raising concerns that AI-driven coding may be inflating reimbursement. The reaction has been predictable: are these technologies introducing inaccuracies, or finally capturing previously underrepresented patient complexity?  The answer lies upstream of both arguments.

Currently, this gap between context and interoperability mitigates against automation that lacks grounding. In medical coding, for example, the missing context protects patients from overbilling, protects payers from inaccurate claims and costly audits, and relieves clinicians from the burden of triangulating care documentation across multiple, often conflicting payer-specific requirements.

Long before data is exchanged, it is shaped. Health systems configure their EHRs differently,  including documentation templates, order sets, problem lists, coding workflows, and data mappings. Further, providers document care differently. These are common operational behaviors, but they introduce variation across the ecosystem. Clinical documentation is narrative and contextual, while revenue cycle data is abstracted and optimized for payer-specific reimbursement. By the time clinical data is transformed for reimbursement, the result coming from the same patient story is often semantically very different from each other.

That gap has real consequences. It fuels tension between payers and providers around accuracy, “coding intensity,” and raises suspicion of over-coding for higher reimbursement. This gap also slows automation and limits the ability of AI to scale in a way that is both trusted and durable.

Consider a common scenario: The disparity with which a patient with diabetes and complications may be coded. One system can code to reflect a patient’s diabetes to meet minimum payer medical necessity requirements, while another codes the condition with the specificity documented by the provider, Type 2 diabetes with chronic kidney disease and neuropathy. Both can be defended, but only one reflects the full clinical picture and drives the right outcomes across reimbursement, care decisions, and analytics. Without a shared standard, both can coexist, and both can be labeled as accurate. 

This variability in specificity hinders AI from reaching its potential. We must train AI with the additional context and specificity, to allow it to produce the minimal subset to meet payer-specific constraints, while ensuring accurate interpretation for patient care and research use cases. In fact, as payers increase their use of AI to evolve their requirements for approving claims and prior authorization requests, they too ask for more specific coding to approve claims.

The recent wave of adoption of AI in clinical workflow outside of billing amplifies this gap. Applications that automate the clinical documentation process will generate code sets without longitudinal context from patients’ established medical history, leading to outputs that may not align with actual patient complexity. By trying to automate coding using disparate applications without shared context across the healthcare enterprise, AI will represent conflicting conclusions and potentially share erroneous information with the larger healthcare ecosystem.

Today, we rely on localized definitions of quality and accuracy shaped by providers’ current manual coding guidelines, individual workflows, and historical practices. The result is predictable: subjectivity. Agreement on accuracy across coders hovers around 50%, even among experienced and fully certified coders. In that environment, interoperability alone cannot deliver alignment.

What is needed is a layer above interoperability, an objective framework for context and quality that establishes shared understanding. This framework does not eliminate variation; it normalizes it, creating a consistent and trusted output across clinical, operational, and financial use cases.

Also, the framework must be uniformly administered. Acting as a compliance engine, this layer ensures that codes are not only technically correct but appropriate across use cases, clinical, financial, and analytical. When that standard is met, codes become more than billing artifacts; they become a reliable representation of patient history and a consistent entry point into the broader clinical record, regardless of where the data originated. Over time, this reduces friction across the system, including audits, denials that are ultimately reversed, and the burden of prior authorization.

We are beginning to see movement across the stack. Documentation platforms are embedding guidance, knowledge engines are moving closer to workflow, and systems are beginning to automatically generate relevant codes. Without an objective framework governing how context is interpreted and quality is measured, however, these advances risk exasperating fragmentation rather than resolving it.

The opportunity, and responsibility, is to align these layers through a shared definition of quality and accuracy. Much like the Rosetta Stone enabled translation across languages, the industry needs a framework that can translate clinical nuance into a consistent, trusted representation across systems and use cases.

When that happens, interoperability becomes more than data exchange; it becomes alignment. More data and faster pipes will not solve this problem on their own. Without an objective framework for context, quality, and compliance, we will continue to move information more efficiently, while critical detail leaks in the translation, and without agreeing on what it represents.

That is the shift healthcare needs next: shared understanding grounded in objective, compliant accuracy.

Photo: eichinger julien, Getty Images

This post appears through the MedCity Influencers program. Anyone can publish their perspective on business and innovation in healthcare on MedCity News through MedCity Influencers. Click here to find out how.