Wearable data EHR integration: what belongs in the chart

Most of what happens to a patient happens between visits. This post is about getting that data into the chart, next to the labs, where it can change a decision.

Most of what happens to a patient happens between visits. How they sleep. What their heart rate variability does after a hard week. What their glucose does after breakfast. Whether they moved at all on Tuesday.

For most of the history of medicine, none of that was measurable. Now nearly all of it is. Rock Health's 2025 consumer survey found that 46% of US adults own a wearable, up from 13% in 2015, and 59% of those owners have already discussed their wearable data with a healthcare provider.

Continuous glucose monitors stopped being a diabetes-only device in March 2024, when the FDA cleared the first over-the-counter CGM for adults who do not use insulin, including "those without diabetes who want to better understand how diet and exercise may impact blood sugar levels." In the same Rock Health survey, 47% of CGM owners report no diabetes diagnosis.

So the patient walks in wearing a ring, a watch, and a sensor on the back of their arm. They have more data about their own body than any chart a clinician could pull up. And almost none of it is in the chart.

Wearable data EHR integration means the patient's device data (Oura, Apple Health, CGM, smart scale) lands in the same chart as their labs and medications, trended on one timeline, instead of arriving as screenshots. This post covers which of those signals change a clinical decision, which do not, and how the integration works in Ultralight.

Wearable data is everywhere, except the EHR

The AMA and Medscape surveyed 2,222 physicians across six countries this year. In the US, 86% of physicians said they review wearable data at least sometimes, and nearly 30% said they take clinical action at least weekly after looking at it. Only 6% said that data was integrated into their clinical workflow, such as being able to import it into the EHR (Healthcare Dive's coverage has the US cut).

That gap is the whole problem. The data lives in the Oura app, the Dexcom app, Apple Health, Garmin Connect. It reaches the clinician as a screenshot in a portal message, a phone held up to the camera on a telehealth call, or a verbal summary from a patient doing their best to remember what the app said last Thursday.

A CGM trace in a standalone app is a curiosity. The patient sees a spike and does not know what to do with it. The clinician sees it a week later, out of context, with no fasting insulin next to it, no medication list, no lab history.

The same trace sitting next to fasting insulin, an A1c trend, a body composition trend, and the current supplement and medication list is a plan. Same data. Different action.

The chart was never the full context

An EHR was built to hold what happens inside the clinic: the visit note, the labs we ordered, the prescriptions we wrote. That was the full context of a patient in 1995. It is not the full context of a patient in 2026.

In functional and longevity medicine especially, the between-visit data is often the point. We are titrating on response. We are asking whether the sleep protocol moved HRV, whether the dietary change flattened the post-meal glucose curve, whether body composition is heading the right way on a peptide or a GLP-1.

The lab draw every three months is one frame. The wearable is the film. If the record holds only the frames, the clinician is doing the hardest part of the job, connecting the film to the frames, in their head, from memory, or from a screenshot.

Ask the wearable the same questions you ask a lab

Before any lab goes on a requisition, a good clinician asks three questions:

1. Does this result change a medical decision?
2. Does it help me make a better decision than I would make without it?
3. Does it change the treatment plan?

A wearable stream has to clear the same bar. Most of the friction clinicians feel with device data comes from skipping this step: treating every metric the app displays as equally clinical, when a handful of signals carry almost all of the decision value and several carry none.

The fourth question is the one the device companies would rather you not ask: what is the sensor actually measuring? Almost every consumer metric is a proxy. Skin temperature is not a progesterone level. HRV is not a cortisol level. A "readiness score" is not a physiological measurement at all.

Measuring a proxy of a hormone is a different clinical act from measuring the hormone's metabolites, and a clinician who reads the proxy as the thing it stands in for makes worse decisions.

Here is how the common signals sort out. For each one: what the sensor actually measures, whether it is reliable enough to act on, and the decision it can change.

CGM (Dexcom, Libre, Stelo)

• Measures: interstitial glucose, roughly 5 to 15 minutes behind blood. Published MARD is about 8 to 9% for Dexcom G7 and Libre 3 in manufacturer trials. No standalone MARD has been published for Stelo.
• Act on it? Yes, for patterns and post-prandial response. Not for a single reading.
• Decision it changes: meal composition and timing. GLP-1, metformin, or berberine response. Dysglycemia phenotype, when read against fasting insulin and A1c.

Resting heart rate

• Measures: nocturnal heart rate from optical (PPG) sensing. Against ECG, consumer devices land within about 2 to 3% (Dial et al., 2025, 13 adults, 536 nights).
• Act on it? Yes. One of the most robust consumer signals.
• Decision it changes: illness onset, overtraining, thyroid or anemia flags, medication effects.

HRV

• Measures: RMSSD from nocturnal PPG. A proxy for parasympathetic tone, not a stress hormone.
• Act on it? Within-person trends only. Never compare across devices or across patients.
• Decision it changes: training load. Recovery protocol response. Whether the sleep intervention did anything.

Sleep

• Measures: total sleep time from movement and heart rate, which holds up well. Four-stage sleep staging agreed with polysomnography on about 70 to 76% of epochs in a 2024 lab study of Oura, Apple Watch, and Fitbit, and deep sleep was the weak spot: one device caught about half of deep-sleep epochs.
• Act on it? Duration and regularity, yes. Stages, no.
• Decision it changes: sleep duration and timing targets. Not "your deep sleep is low."

Skin temperature (Oura, Apple Watch)

• Measures: distal skin temperature deviation from baseline. A proxy for progesterone's thermogenic effect, not a hormone level.
• Act on it? Trend only.
• Decision it changes: cycle mapping in perimenopause. Illness onset. Does not replace hormone testing.

Steps and activity

• Measures: accelerometry.
• Act on it? Yes.
• Decision it changes: activity prescription. Sarcopenia risk.

Body composition (smart scales)

• Measures: bioimpedance, which shifts with hydration. Not DEXA: against DEXA, individual body fat readings can be off by several kilograms in either direction.
• Act on it? Within-person trend only.
• Decision it changes: lean mass on a GLP-1. Protein targets.

Overnight SpO2

• Measures: pulse oximetry.
• Act on it? Screening-grade, not diagnostic. In a 2024 study against polysomnography, a consumer smartwatch reached an AUC of 0.89 for severe apnea (n=51).
• Decision it changes: sleep apnea referral.

Cuffless blood pressure (Hilo)

• Measures: calibrated optical estimate.
• Act on it? Trend, with monthly cuff calibration.
• Decision it changes: hypertension screening. Treatment response.

"Readiness," "strain," and "stress" scores

• Measures: proprietary composites of the above.
• Act on it? No.
• Decision it changes: none. A composite is not a chartable finding.

Read down the "decision it changes" lines and a pattern shows up. The signals that change decisions are the ones that can be placed next to a lab: glucose next to insulin, lean mass next to a GLP-1 dose, HRV next to the sleep protocol start date. The signals that do not change decisions are the ones with no lab to sit next to.

That is not a coincidence. It is the argument for putting device data in the chart rather than in a separate dashboard, and it is one of the things to weigh when comparing EHRs for functional and longevity medicine.

More data does not help until something makes sense of it

There is an honest counterargument here, and clinicians make it to us often. More data is not automatically better. A patient wearing two devices that disagree on resting heart rate makes for a harder conversation. A patient checking a CGM 40 times a day can end up with a worse relationship with food.

There is a subtler version of the same problem. A wearable is useful when it trains a patient's interoception, when they learn what a 40-point post-meal spike feels like and start noticing it before the app does. It is harmful when it replaces that awareness, when the patient checks the ring instead of checking in with their body.

The research on self-tracking and body awareness is still mixed. A 2023 mixed-methods study found that wearing a Fitbit increased awareness of bodily sensations, especially in women, without changing overall body perception. The best recent framing, from an ethnographic study presented at EPIC 2025, is that wearables should reengage bodily awareness rather than replace it. Some of the best clinicians we work with tell certain patients to take every device off for a month. That is a legitimate prescription.

The chart-side review, clinician and patient looking at the same trend together, is where the first outcome gets taught and the second gets caught. It is also why raw streams need three things before they are clinically useful, and why each one prevents a specific error.

• They need to sit next to the labs. A glucose curve read without fasting insulin invites over-treating a normal response, or missing an abnormal one.
• They need to be trended on the same axis as everything else. Without a shared timeline, "did the change we made in June do anything?" gets answered by attributing the result to the wrong intervention.
• They need a layer that reads them fast. Nobody has 20 minutes per patient to scroll 90 days of sleep staging. Reading a long series against the rest of the chart and surfacing what changed is the job AI is actually good at, and the change that mattered is exactly what scrolling misses.

The modern EHR should be the place where all of this comes together: labs, notes, imaging, uploaded records, and the continuous data from the devices patients already wear, on one surface, with help making sense of it. That is what we built, and it is rolling out now.

How patient wearable data gets into the EHR: Trends with connected devices

A 2-hour post-prandial glucose of 165 in a standalone app is a number. The same reading in Trends, next to a fasting insulin of 18, a triglyceride-to-HDL ratio of 3.2, and a lean-mass trend heading down on a GLP-1, is an insulin-resistant phenotype losing muscle, and a plan you can write today.

That is the job Ultralight does with device data: turn streams into signals. It co-locates the wearable data with the labs it has to be read against, trends both on one axis, and lets Copilot read 90 days in the time it takes you to scroll three.

How the data gets in. Patients connect their devices from their phone. Apple Health (HealthKit) or Android Health Connect pulls data straight from the phone, and anything a patient already routes through Apple Health, including a Stelo sensor or an Apple Watch, comes through this path.

Direct connections today include Oura, Fitbit, Garmin, Dexcom, Freestyle Libre, Withings, Eight Sleep, Polar, Strava, Peloton, Ultrahuman, Cronometer, Google Fit, Omron, Kardia, Beurer, Hammerhead, and Wahoo. New devices are added over time, and the in-app list is the current one.

The patient controls the scope. Before anything is shared, the app lists exactly what will be: steps and activity, heart rate, sleep, heart rate variability, workouts, and body measurements such as weight and body fat percentage. The patient turns on the categories they want to share, category by category, and can disconnect any device at any time from the same screen.

Where it lands. Once connected, the data goes into Trends, the same view that holds the patient's lab biomarkers. A clinician sees a glucose series or a resting heart rate series on the same surface as fasting insulin, hs-CRP, or vitamin D. A clinician can pin the trends that matter for a given patient, and the patient sees the pinned set in their own app. One picture, shared by both people in the room.

Where the reading happens. From there, the clinician can ask Copilot about the trend in the context of the whole chart: what changed since the last visit, whether the post-meal curve flattened after the dietary change, whether HRV moved after the sleep protocol started.

If wearable data is one of the reasons you are considering a move, our guide on how to switch EHRs covers how to test it against your own workflow before you commit.

Wearable data in a functional medicine practice

Crystal Brust, PA-C, runs Farm to Functional Medicine, a functional medicine practice that works largely with firefighters and first responders, a population where shift work makes sleep, recovery, and glucose patterns both harder to manage and more important to see. She has been on Ultralight through the whole wearables rollout, and she has been using CGMs with patients for a while.

"I use a lot of CGMs with my patients," she told a group of clinicians on one of our onboarding calls in July. Her patients wear the sensor, and the data comes into the chart through Apple Health.

Here is how she described what changes once it is in the chart, on a call in August:

"I have patients that upload their CGM activity and we can go over that together, and then AI can actually build that in and we can talk about what that means for their metabolic syndrome. It gives us insight into the CGM data and lets us discuss it with the patient in real time. And that's been a lot of fun."

The phrase that matters is real time. The CGM data used to be something the patient looked at alone, at home, between visits. Now it is something the two of them look at together, next to the labs, in the visit.

On why this belongs in the chart rather than in the patient's phone:

"A lot of times patients get all of this data but they don't really know how to connect all of the dots. Being able to upload that into Ultralight, and then having AI also be able to help connect these dots on the recovery, heart rate variability, all of that."

Where wearable data in the EHR is still early

Two honest limits are worth naming, because the devices are moving faster than the evidence.

First, using wearable data to predict what happens next is still a research question. The signals are real. A March 2026 study in npj Digital Medicine of 4,244 people with complex chronic illness found that a morning rise in heart rate and drop in HRV predicted worse symptoms that evening, and adding those biometrics to the prior day's symptom reports improved model AUC from 0.73 to 0.83 up to 0.82 to 0.85, depending on the model.

A November 2025 study in Scientific Reports tracked 53 rheumatoid arthritis patients on Apple Watch, Fitbit, and Oura and found "all metrics were altered up to 4 weeks prior to inflammatory and symptomatic flare development."

Both are small or retrospective, both sets of authors flag confounders they could not control, and neither has been turned into a validated clinical rule. Reading a trend next to labs is useful today. Treating a trend as a forecast is not yet supported.

Second, what wearables can measure is still narrow. Sleep, heart rate, HRV, activity, glucose, weight. That is most of the list. The next layer is arriving but not here.

The FDA cleared the first cuffless, over-the-counter continuous blood pressure band in July 2025, with monthly calibration against a cuff still required. In-ear EEG for brain health and cognitive function is an active research area, and an April 2026 review in Frontiers in Human Neuroscience still lists "the lack of standardized validation protocols" as a gap limiting clinical use.

Continuous metabolic monitoring beyond glucose is coming through sweat. In May 2026, UC Irvine published a sweat sensor in Nature Biomedical Engineering that tracks cortisol, glucose, lactate, and urea at once, and the technology "is currently under further development." Expect all of these to show up on patients' wrists before the evidence base catches up. Run each one through the four questions above when it does.

And the most underrated wearable is not a device. Dr. David Lipman, a physician with degrees in medicine and exercise physiology who writes the Nexus Health & Performance newsletter, put it this way: the most underrated wearable is someone's calendar. Sleep, stress, and recovery track what a person's week looked like, and no sensor captures that on its own.

The point of getting continuous data into the chart is to hold as much context about a patient as possible in one place, next to the labs, where the signals that change decisions can be read against the labs that give them meaning. Making sense of it is still messy. It will get less messy as the research matures and the devices broaden. A chart that already holds the data will be ready when it does.

If you are a clinician whose patients keep showing up with data you cannot see in the chart:

‍Book a demo and we will show you Trends on a real chart.

Frequently asked questions

Can you connect a CGM to an EHR?
Yes, in Ultralight. Dexcom and Freestyle Libre connect directly through the patient's app. Stelo connects through Apple Health, which Dexcom feeds on a roughly 3-hour delay. The glucose series appears in Trends next to the patient's lab biomarkers, and Copilot can read it in the context of the patient's chart.

Does Ultralight integrate with Dexcom?
Yes. Dexcom CGMs are supported by direct connection. The over-the-counter Stelo sensor is a Dexcom product but routes through Apple Health rather than a direct connection.

Does Ultralight integrate with Freestyle Libre?
Yes. Freestyle Libre is a direct connection through Other Apps and Devices in the Ultralight app.

Does any EHR integrate with Oura Ring?
Ultralight does. Oura is a direct connection: the patient logs in with their Oura account from the Ultralight app and chooses what to share, and sleep, HRV, resting heart rate, and activity appear in Trends next to labs. Most EHRs do not connect to Oura; the usual workaround is a separate analytics tool or a screenshot.

Does Ultralight connect to Apple Watch, Fitbit, or Garmin?
Yes. Fitbit and Garmin are direct connections. Apple Watch data comes through Apple Health on an iPhone. Android wearables come through Health Connect.

Does Ultralight support Whoop?
Not yet. Whoop is in progress.

What wearable data shows up in the chart?
Through Apple Health or Health Connect: steps and activity, CGM glucose, heart rate, sleep, heart rate variability, workouts, and body measurements including weight and body fat percentage. Direct device connections bring the metrics that device reports. Everything lands in Trends alongside labs and is available to Copilot in each patient's chart.

Which wearable metrics are actually clinically useful?
The ones that can be read against a lab or an intervention date: CGM glucose patterns, resting heart rate, within-person HRV trends, total sleep time, activity, body composition trends, overnight SpO2, and calibrated cuffless blood pressure. Proprietary composite scores ("readiness," "strain," "stress") are not chartable findings.

Can patients control what they share?
Yes. The patient chooses which categories to share when connecting, and can disconnect any device at any time from Connected Apps. Once disconnected, the care team stops receiving new data from that device.

How does a patient connect their wearable?
In the Ultralight app: Account, then Connected Apps. Choose Apple Health or Health Connect to share from the phone, or Other Apps and Devices to log in to a device brand directly. Repeat for each device.

How do I get patient wearable data into my EHR?
Three paths exist today. Patients send screenshots or PDF exports, which you file manually. A separate analytics platform aggregates devices and shows you a dashboard outside the chart. Or the EHR connects to the devices directly, as Ultralight does, so the data lands in the chart next to labs with no export step. Only the third puts the data where the clinical decision is made.

Is wearable data in the EHR the same as remote patient monitoring?
No. Remote patient monitoring (RPM) is an insurance-billed program with specific CPT codes, device requirements, and monthly review thresholds. Wearable data in the EHR is the patient's own consumer device data trended in the chart to inform visits. A cash-pay practice can use it without an RPM program, and it does not by itself generate RPM billing.

Sources

• Rock Health, 2025 Consumer Adoption Survey. 8,000 US adults, fielded December 2025, published June 1, 2026. Retrieved September 10, 2026.
• FDA, FDA Clears First Over-the-Counter Continuous Glucose Monitor, March 5, 2024. Retrieved September 10, 2026.
• AMA, Wearable data adoption stalled by system barriers, July 8, 2026, and Healthcare Dive coverage of the US figures. Retrieved September 10, 2026.
• Garg et al., Dexcom G7 accuracy, Diabetes Technology & Therapeutics, 2022 (Dexcom summary); Karinka et al., FreeStyle Libre 3 accuracy, ADA 2022 abstract 76-LB. Retrieved September 10, 2026.
• Dial et al., nocturnal heart rate and HRV validation of Oura, Whoop, and Polar against ECG, Physiological Reports, August 2025. Retrieved September 10, 2026.
• Sleep staging validation of Oura Gen3, Apple Watch Series 8, and Fitbit Sense 2 against polysomnography, Sensors, 2024. Retrieved September 10, 2026.
• Bioimpedance vs DXA body composition, Frontiers in Analytical Science, 2026. Retrieved September 10, 2026.
• Browne et al., smartwatch SpO2 and sleep apnea detection vs polysomnography, Journal of Clinical Sleep Medicine, 2024. Retrieved September 10, 2026.
• Boldi et al., Exploring the impact of commercial wearable activity trackers on body awareness and body representations, Computers in Human Behavior, vol. 151, 2024 (online 2023). Weller, Cassin and Maury, Somatic Intelligence: How Wearables Can Reengage, not Replace, Bodily Awareness in the Age of Generative AI, EPIC 2025 Proceedings. Retrieved September 10, 2026.
• Digital physiological biomarkers predict within-person symptom changes in complex chronic illness, npj Digital Medicine, March 24, 2026. Retrieved September 10, 2026.
• Wearable devices detect physiological changes that precede rheumatoid arthritis flares, Scientific Reports, November 30, 2025. Retrieved September 10, 2026.
• Fierce Biotech, FDA clears over-the-counter, cuffless blood pressure monitor, July 8, 2025. Retrieved September 10, 2026.
• In-ear EEG wearables for brain activity assessment and cognitive rehabilitation, Frontiers in Human Neuroscience, April 20, 2026. Retrieved September 10, 2026.
• UC Irvine News, Wearable sweat sensor for long-term health monitoring, May 13, 2026. Retrieved September 10, 2026.
• Crystal Brust quotes: Ultralight onboarding calls, July 29 and August 26, 2026; Ultralight and Parsley Health call, August 24, 2026. Transcripts on file.