Value-based care can work.
The data infrastructure was
never built to match it.

The premise of value-based care is straightforward: align the financial incentives of the provider with the long-term health of the patient. When it works, the provider earns its return by keeping people healthier for longer, not by billing more encounters. The bet is right. The infrastructure was never built to match it.

Managing a population's cost of care with data that arrives 90 days late is like navigating by the last known position of a ship. The position is accurate. It's just not now. By the time claims data shows a patient's health trajectory declining, they've already been admitted. The cost is incurred. The intervention window is gone.

This is the lesson of the 2022–2024 value-based reset. Bright Health, Clover, Haven, Agilon: each unwound for variations of the same root cause. Not enough real-time visibility into what was happening to their patients between visits.

A patient with congestive heart failure doesn't deteriorate in front of a clinician. They deteriorate on a Tuesday afternoon, at home, when their resting heart rate climbs and their step count drops. That signal exists. It's flowing now, from the wrist of roughly a third of American adults.

The question is whether the entity financially accountable for what happens next is looking at it.

Until recently, interpreting continuous longitudinal biosignal required specialists reviewing data manually. Clinical AI changes this. Models trained on patient-wearable data can now identify precursors to an acute event, days before symptoms emerge. What required a cardiologist reviewing hours of recordings now runs at scale across an entire enrolled population. That is what makes the between-visit layer economically viable.

For the first time, three conditions are simultaneously true. Clinical AI can interpret longitudinal biosignal with meaningful accuracy. Wearable adoption has reached consumer scale. And a regulatory mandate is standardizing the FHIR data substrate that makes biosignal ingestible alongside the clinical record. This platform was not buildable before 2024.

From 2024 to 2026, each condition crossed its threshold. Clinical AI models reached accuracy levels that made biosignal interpretation clinically defensible. Wearable penetration crossed one third of American adults, enough for population-level programs to be viable. And the ONC's information blocking rules took effect, forcing the first real wave of FHIR-compliant APIs into production. None of these were true simultaneously in 2024. All three are true now. The window opened in 2026. It won't stay open long.

We're building the infrastructure to make continuous between-visit signal legible: fused with the clinical record, ranked by care plan context, surfaced in the workflows of providers that carry real financial accountability for what the signal is telling them. Not a better dashboard on the same old data. A different data substrate that makes value-based care's original promise workable.