Congratulations to @RiverSpring Living, winner of the Early Detection & Prevention category in Newsweek's inaugural AI Impact Health Awards.

But Companion goes beyond detection. It turns what a patient reports at home into an alert with an owner.

An alert appears in the care team's view with the patient's own words attached.

A recovery check-in answered at home reported a symptom worse than yesterday. On its own it's one data point. But the team set a threshold before discharge, and this answer crossed it.

Whoever opens the alert can see the procedure, the discharge plan and every earlier check-in without leaving the record.

Is the symptom a problem? Someone who can find out is already looking.

That's early detection at the scale of one person. It's worth noticing, because most talk about early detection happens at the scale of millions.

Much of the work in that space scores whole populations and flags who might deteriorate. It's valuable, and it has one stubborn gap: somebody still has to notice the score, decide it matters and reach the patient.

Clinical policy has been moving toward the individual version. The Centers for Medicare & Medicaid Services' CY 2026 fee schedule now recognizes remote monitoring over as few as 2 to 15 days of data in a 30-day period, the short windows when many post-procedure problems appear.

Companion starts from the patient rather than the population. Its check-ins at thirty, sixty and ninety days ask about what the care team chose to watch for this patient and this procedure. When a reported value crosses the team's threshold, the alert lands in the record the team already works in, so the people who planned the treatment see the signal first.

Detection also depends on the patient still answering on day twelve. Companion gives patients reasons to keep opening it: a daily care plan, medication schedules, secure messages to the care team, and a guardrailed assistant that helps them understand their instructions without ever diagnosing.

There's a quieter point here for anyone thinking about where lasting value sits. A risk score nobody acts on leaves no trace. An alert in the patient's own record leaves a complete one: what the patient reported, when the threshold was crossed, who opened it and what they did next. That chain turns early detection from a statistic into accountable care, and it's the same documentation that shows monitored care took place.

No technology can promise a better outcome. What this design can do is make sure every warning a patient gives has someone responsible for it.

So when you compare early-detection approaches, ask whose action the signal triggers, and whether the record shows what happened next.