Healthcare has been organized around appointments for as long as there have been appointments. A patient notices a problem, comes in, gets treated, and goes home until the next one. For a sprained ankle that is a reasonable design. For someone managing heart failure, diabetes or a recovery from surgery, it means the care team sees a few minutes out of every ninety days and infers the rest.
Wearables close part of that gap, and patients have already closed it without waiting for anyone’s permission. They arrive at visits with a year of resting heart rate, sleep stages and step counts on a phone, and they expect it to mean something.
The useful question for a practice is not whether wearable data has value. It is narrower and more awkward: which of this data can change a decision, who is going to look at it, and which of it can be billed. Those are three different answers, and the third one surprises people.
The Line That Decides Almost Everything
Before any workflow discussion, there is a classification that determines what you can do with a device and whether the time spent on it is fundable.
The remote physiologic monitoring codes, covering setup and patient education, the supply of the device with daily recordings, and the treatment management time that follows, all turn on one requirement: the device collecting and transmitting the data must meet the FDA’s definition of a medical device. That is a different and often misunderstood test. A device does not have to be FDA cleared or approved. It has to fall within the statutory definition of a medical device in the first place.
Most consumer wearables are deliberately built and marketed to sit outside that definition, as general wellness products. That positioning is what keeps them out of a regulatory pathway, and it is also what keeps them out of the RPM codes. There is a second requirement that trips practices up just as often: the data has to upload automatically from the device. A patient reading numbers off a screen and typing them into a portal is self reported data, and it does not satisfy the device supply requirement.
So a patient’s smartwatch trend can be genuinely clinically useful and simultaneously not billable under remote monitoring. Both things are true at once, and a monitoring program built without knowing which category each device falls into will fail its first audit or its first margin review.
The practical consequence is that a practice needs two lists, not one. Devices it supplies and bills against, which must clear the medical device and automatic upload requirements. And patient owned consumer devices, which can inform a conversation, get referenced in a note, and prompt a proper measurement, but which sit outside the monitoring program. Verify the current code descriptors and the device status case by case against the fee schedule and your payer policies, because both the codes and the device landscape move.
Continuous Monitoring for Chronic Disease
For patients with heart disease, diabetes, hypertension or respiratory illness, the clinically interesting changes happen between visits by definition. A blood pressure taken in your exam room is one reading, taken in an unusual setting, at a time of day chosen by your scheduler. A month of home readings is a distribution.
What changes management is rarely a single alarming value. It is a trend: resting heart rate drifting up over three weeks, weight climbing five pounds in a week in a heart failure patient, activity dropping off after a medication change. Those are the patterns worth building a program around, because each has a defined response.
A systematic review in The Permanente Journal looking at how practitioners themselves experience remote monitoring found the pattern that anyone who has run one of these programs will recognise. Clinicians see the clinical value clearly, and their reservations are almost entirely operational: data volume, unclear responsibility for review, and the fit with existing workflow. The technology is not usually what fails.
Activity and Recovery Data
Fitness metrics get dismissed as consumer noise, which undersells them in one specific situation: post procedure recovery and rehabilitation, where the question is whether the patient is actually moving.
Patients reliably overestimate their own activity, not dishonestly but because effort feels like volume. A patient who says they are “getting up and about” and whose device shows 800 steps a day is not lying, and the gap between those two accounts is clinically useful. It tells you whether to escalate, reassure, or look for a reason the patient is not moving, such as pain control that is not working.
This is also the clearest case of data that changes care without being billable under monitoring codes, which is exactly why the two lists matter.
Smart Eyewear and Accessibility
Wearables are not only wrist worn, and the eyewear category has moved faster than most practices realize.
For patients who are blind or have low vision, consumer smart glasses now do things that used to require a dedicated assistive device or another person. Meta’s Ray-Ban and Oakley lines run an assistant that reads text aloud, describes surroundings and identifies objects, and they carry a Be My Eyes integration that connects the wearer hands free to a sighted volunteer, a trusted contact, or a company’s support team on a voice command.
What matters clinically is not the feature list, it is the form factor. A patient is far more likely to use assistance that lives in the glasses they already put on than assistance that requires carrying, unlocking and pointing a separate device. These are consumer products, ordinary sunglasses with camera and microphone built in and sold at retail, and that is the source of both their value and their limits. The value is adoption. The limit is that a general wellness product is not an assistive medical device, it carries no clinical validation, and it belongs in a conversation about independence and quality of life rather than in a treatment plan or a monitoring program.
Which is worth saying plainly to patients who ask, because they will ask, and the honest answer is that this technology can make daily life meaningfully easier without being part of their medical care.
Mental Health and Stress Signals
Mental health is the hardest thing to track between appointments, because symptoms fluctuate and patients do not always recognise a decline while it is happening, let alone report it.
Wearables capture proxies rather than symptoms: sleep duration and continuity, activity volume, heart rate variability. None of that diagnoses anything. What it can do is flag a change in pattern worth a phone call. A patient whose sleep has fragmented and whose activity has halved over two weeks is worth contacting, and that contact is the intervention, not the data.
The failure mode here is treating a metric as a symptom. Heart rate variability is not a mood score, and a practice that starts responding to it as though it were will generate a great deal of alarm and very little benefit.
Making It Work: the Integration Problem
Collecting more data does not improve outcomes on its own. Most wearable programs that stall do so for the same three reasons, and none of them are technical.
Nobody defined who reviews it
If review is everybody’s job it is nobody’s. It needs a named role, a protected window in the schedule, and an escalation path for when the reviewer is not the prescriber. This is the work that chronic care management is built to structure and fund, and it is where a monitoring program either becomes routine or quietly lapses after the first busy month.
The data lands somewhere nobody works
A vendor dashboard that requires a separate login gets checked for about three weeks. The values that matter have to arrive in the electronic health record as structured data, next to the labs and the medication list, where they can be trended, flagged and pulled into a note without anyone retyping them. Ask any prospective vendor whether values arrive as discrete fields or as an attached PDF, because that single answer determines whether the program is sustainable.
Everything alerts, so nothing does
Thresholds set by the device manufacturer are set for liability, not for your panel. They need tuning to the population, and the output has to be a short worklist rather than a stream. If a nurse opens the queue and sees forty names, the queue is broken and it will be ignored within two weeks.
Two things sit alongside those. Patient selection: not everyone benefits from monitoring, and enrolling broadly is the fastest way to bury the signal. And patient education, delivered through a patient portal or whatever channel they already use, because a patient who understands what is being watched and who is watching it wears the device, and a patient who does not stops charging it in week three.
What to Settle Before You Enrol Anyone
- Which devices you supply and bill against, confirmed as meeting the medical device definition with automatic upload.
- Which patient owned devices you will accept data from, and the note language for referencing it without implying it was validated.
- Who reviews, in what window, with what escalation path.
- Which specific trends trigger contact, written down, with the response for each.
- How review time gets captured, as a byproduct of the review rather than a second data entry task.
- Who you enrol, and just as importantly who you do not.
- What you will measure to decide in six months whether the program is working.
Conclusion
Wearables have already changed what a patient knows about themselves between appointments. What they have not yet changed, in most practices, is what the care team does with any of it.
The gap is not device availability and it is not clinician interest. It is the unglamorous middle: knowing which devices are billable and which are merely informative, getting values into the record as structured data, naming who reviews and when, tuning thresholds to a real panel, and capturing the time so the work funds itself. Practices that build that middle get a proactive care model. Practices that skip it get a drawer of devices, a dashboard nobody opens, and a reasonable conclusion that remote monitoring does not work.
