Healthcare has spent decades connecting applications. The next challenge is harder: creating a trusted clinical data foundation capable of supporting intelligent, longitudinal and increasingly AI-assisted care.

The instinct is often to start with the model. Which AI system is most capable? Which copilot should sit inside the workflow? Which algorithm can predict deterioration first?

But every one of those questions comes later.

The $265.6 billion number is a warning, not a diagnosis

A widely cited JAMA review estimated that administrative complexity accounted for approximately $265.6 billion in annual waste in the U.S. healthcare system. That figure should not be described as the cost of fragmented healthcare data alone. It is broader than that.

What the number does show is the scale of the burden created when healthcare depends on repeated reconciliation, manual intervention, duplicated documentation and disconnected administrative processes.

Fragmentation is one contributor to that environment. A patient's story may be distributed across an electronic health record, laboratory system, imaging archive, pharmacy, insurer, specialty platform, home-monitoring device, patient portal and wearable. Each system can be useful on its own. The problem begins when every new application becomes another destination that must be separately integrated, interpreted and reconciled.

The patient story is larger than the EHR

Michael's complete PatientTrac Insights build makes that point visible from seven different directions.

These are not seven unrelated editorial subjects. They are seven views of the same problem: a clinically useful patient story has multiple authors and multiple kinds of evidence.

Moving data is not the same as creating a data foundation

Interoperability is essential, but exchange alone is not the destination.

HL7 describes FHIR as a standard for exchanging healthcare information electronically and notes that automated clinical decision support and other machine-based processing depend on data being structured and standardized. CMS is also requiring specified impacted payers to implement and maintain certain FHIR APIs under its Interoperability and Prior Authorization Final Rule, with major API requirements generally beginning January 1, 2027, depending on payer type and provision.

Those developments matter because they make structured exchange part of healthcare's operating infrastructure. But a message successfully transmitted from one system to another is not automatically clinical intelligence.

For information to support decisions, it also needs context. Identity has to be resolved. Terminology has to be normalized. Provenance has to be retained. Time matters. Who observed something matters. Whether a value came from a clinic device, a patient's home, a caregiver report or a claims transaction matters. What happened after the information arrived matters.

What belongs in a trusted clinical foundation?

If the goal is a longitudinal patient story rather than a collection of interfaces, the foundation has to accommodate more than traditional encounter data.

AI makes the quality of the foundation more important, not less

Artificial intelligence can summarize, classify, predict and assist. It does not eliminate the consequences of incomplete or poorly contextualized inputs.

If the medication list is stale, a recent laboratory result is missing, a device reading is disconnected from the clinical record, a symptom report never reached the care team, or the provenance of a value is unclear, a sophisticated model still begins with an incomplete representation of the patient.

That is why the data foundation matters before the model does.

Trustworthy clinical AI requires more than access to large amounts of information. It requires information that is appropriately structured, attributable, current, normalized and connected to the clinical context in which it will be used.

The hardest question is not whether the alert fired. It is where the alert lands.

Several PatientTrac Insights pieces already circle this operational truth. Where the Alert Lands asks whether a signal actually reaches the person who can act. Every Access Leaves a Line focuses on accountability. One Record, Every Door asks whether continuity survives organizational growth. The Story That Survives asks whether the patient's history remains intact over time.

Together they point to the same conclusion: clinical intelligence is not created by aggregation alone. It is created when information remains usable, governed and actionable.

A blood-pressure reading that no one sees is storage. A symptom report that cannot be routed is a message. A longitudinal record that cannot explain who changed what is a liability. Data becomes intelligence only when the surrounding system can understand it and support an appropriate next action.

From interoperability to clinical intelligence

PatientTrac's view is that the next phase of digital health will be defined by what happens after information is connected.

The important design principle is reuse. A home blood-pressure reading should not have value only inside a remote-monitoring dashboard. A symptom report should not disappear inside a single disease-specific application. A laboratory result should not require a new integration every time another clinical workflow needs it.

When data becomes part of a reusable longitudinal foundation, the same information can support multiple authorized purposes without recreating the patient record for every new application.

Clinical intelligence is not surveillance

More data is not automatically better care. A trustworthy architecture also has to know what not to collect, who is authorized to see information, how consent and access rules apply, and when a human decision must remain a human decision.

This is another reason provenance and auditability belong inside the foundation rather than being added after the fact. Trust depends not only on whether the data is accurate, but also on whether its use is explainable and governed.

One patient. One longitudinal story. Many authorized uses.

This is the larger idea behind the PatientTrac Clinical Intelligence Network.

The goal is not to create another destination for healthcare data. It is to create continuity across the systems, people and workflows that already participate in care.

That distinction becomes more important as care moves beyond the encounter. Home monitoring, wearables, patient-reported symptoms, mobile applications, caregiver observations and new treatment pathways are creating an unprecedented flow of information between visits.

Those signals become valuable when they join the broader patient context. Otherwise, digital health risks solving fragmentation by creating more fragmentation.

The next infrastructure decision

Healthcare organizations evaluating AI frequently begin with models, copilots and applications. Those tools matter. But the more durable infrastructure decision may be underneath them.

Can the organization create a trusted longitudinal foundation that supports the applications it has today, the care models it is building now and the clinical intelligence workflows it has not yet imagined?

That is a different architecture from integrating every new point solution independently.

The next generation of healthcare infrastructure will not be defined simply by whether systems can exchange information. It will be defined by whether that information can become reliable clinical intelligence at the moment it is needed.