The Clinic on Your Wrist: Wearables and the Turn Toward Prevention

For most of medical history, your health was a series of snapshots. You felt unwell, you made an appointment, and a professional measured you for six minutes in a room that smelled of disinfectant. Those six minutes had to represent the other 87,654 hours of your year. Everything the doctor did not see, they had to infer from what you could remember and were willing to say.

The wrist changed that. Not dramatically, and not for every condition — but for the handful of signals a sensor can watch continuously, medicine now has something it never had: the boring days. The nights before the episode. The eleven weeks of gradually rising resting heart rate that nobody, including the patient, would have described as a symptom.

What continuous actually buys you

The value of a wearable is almost never in any single reading. A blood-oxygen figure at 3 p.m. on a Thursday is close to meaningless. The value is in the baseline — the shape of your own normal, established over months, against which a deviation becomes legible.

This is why the most clinically useful features are the least exciting ones. Irregular rhythm notifications, which quietly identify atrial fibrillation in people with no symptoms at all, arguably justify the entire category on their own; untreated AF is a leading cause of stroke, and it is often silent until it is not. Continuous glucose monitoring turned diabetes management from a set of finger-prick guesses into a feedback loop tight enough that people can see a meal's effect while they are still at the table.

A snapshot tells you where you are. A baseline tells you which way you are going — and direction is the part that matters.

The shift from cure to prevention

Health systems have argued for prevention for fifty years and struggled to deliver it, because prevention requires acting on people who do not currently feel ill. That is expensive at population scale and unwelcome at individual scale. Continuous measurement changes the economics: instead of screening everyone periodically, you can watch everyone continuously at nearly zero marginal cost and screen only those whose own baseline has moved.

The clearest wins so far are in cardiac rhythm, sleep-disordered breathing, glucose control, post-operative recovery monitoring, and fall detection for older adults. The common thread is that each has a measurable physiological signal, a meaningful intervention, and a window in which acting early changes the outcome. Where any one of those three is missing, the wearable produces data and nothing else.

What it replaced

Chiefly, memory. The patient history — "how have you been sleeping?", "how often does this happen?" — is one of the least reliable instruments in medicine, not because people lie but because human recall of routine physiology is genuinely poor. Replacing recall with a record removes a whole class of diagnostic error.

It also replaced, for many people, the diagnostic hospital stay. Cardiac monitoring that once required a night on a ward with adhesive electrodes now happens at home, over two weeks, with far better data — because a person in their own bed behaves like themselves.

What it cost

The costs of this category are unusually well documented, largely because clinicians have been vocal about them.

• False positives are not free. A notification with a modest positive predictive value, delivered to millions of healthy people, generates a flood of anxious patients and normal test results. Each of those tests has its own risk, cost and waiting list, and the person ahead of them in the queue pays for it in time.
• Measurement can become the illness. A minority of users develop a genuine preoccupation with their own metrics — checking sleep scores obsessively, then sleeping worse for it. "Orthosomnia" was coined for exactly this, and it is a real clinical presentation.
• The data is not neutral. Health signals derived from consumer devices sit outside most medical-privacy regimes. A heart-rate record is not protected the way a chart is, and it is commercially interesting to insurers, employers and advertisers in a way that almost nothing else about you is.
• Accuracy is uneven. Optical sensors perform differently across skin tones, tattoos, wrist sizes and movement. A device validated mostly on one population will be quietly worse for everyone else, and users have no way to know.

Who it actually helps

Here is the uncomfortable pattern in the adoption data: the people wearing these devices skew younger, wealthier, more urban and already healthier than average — while the greatest clinical benefit would accrue to older, poorer, more rural and sicker populations. A technology that finds silent atrial fibrillation is most valuable in the demographic least likely to be wearing it.

The programmes that have worked against this gradient share a design: the device is supplied rather than purchased, the data goes somewhere a clinician actually looks, and someone follows up by telephone. That last element is the expensive one, and it is the one that gets cut. A monitoring programme with no human at the other end is not a health intervention. It is a notification service.