A cardiologist's blunt verdict on patient shared smart band stats: roughly 70% is vendor defined noise, and the two genuinely useful numbers are buried inside it.
A patient walks into Dr. David Kao's Wednesday morning appointment with smart-band stats she is worried about. The cardiologist, an associate professor at the University of Colorado School of Medicine, glances at the numbers and lands the line that should organize this story: "Probably 70% of it, I just don't know what to do with clinically, because it's all been made up by the company. And then there were like two things that were incredibly useful that we would not have had if she wasn't wearing her [device]."
That two-part quote is the spine: the critique and the genuine value, held together in the same breath.
The wearable health boom has put a continuous fire hose of consumer health data into a system built around episodic, problem-driven visits. Heart rate, sleep, steps, stress scores, blood-oxygen dips, recovery metrics. Patients show up with screenshots, exports, and questions. Clinicians, on the other side of the desk, are supposed to do something with it. Most do not have a workflow for it, a billing code for it, or a way to tell which numbers are validated and which are marketing copy repackaged as a metric.
This tension is the spine of ZDNET's "How Data Can Improve Your Health and Wellness" feature, which documents the gap between what consumers can measure about themselves and what a clinical visit can actually absorb. The article frames a real, named structural mismatch: a continuous stream of vendor-defined metrics on one side, an episodic care system on the other, and no shared grammar between them.
Kao's verdict is not "wearables are snake oil." It is more specific and more useful than that. Most of the consumer metrics he sees are vendor-defined and lack the clinical validation a doctor needs before acting. But buried in the noise are occasional findings that change a visit: an arrhythmia caught early, a sleep pattern that explains a daytime symptom, a heart-rate trend that reframes a complaint. The two are not in conflict. They are the same data stream seen from two different jobs.
The "what happens next" question is not whether AI will save the clinic. The honest answer is that several pieces of plumbing would have to be built first, and each one has a party that owns it.
Some physicians, according to the same ZDNET feature, hope AI summarization tools will compress a week's worth of wearable data into a paragraph a doctor can read in thirty seconds. Vendors would have to publish validation evidence for the metrics they ship, with a clear line between a fitness signal and a clinical claim. Health systems would need reimbursement codes for asynchronous data review, or the visit stays at fifteen minutes and the wearable export goes unread. EHR integration would have to ingest continuous streams in a structured form, with provenance attached, so a number from a consumer device is not mistaken for a number from a clinical device. Regulators would need standards that say what a cleared claim means when a feature is added by software update rather than a new device.
None of this is inevitable. These are bets being placed in specific places, by specific partners, and the outcomes will vary. The work to watch is who builds the translator, and on what timeline.
The patient in Kao's exam room is not the problem, and the doctor is not the failure. They are two people trying to use tools that were not designed to talk to each other. The next chapter of wearable health will be written by whoever builds the bridge.