Sleep, activity and heart rate data can help care teams spot risk between visits and after discharge. Here is what the evidence supports, how HIPAA and Part 2 apply, and how to roll it out without alert fatigue.
Key takeaways
Biometric data in behavioral health has moved from research labs to patients' wrists and fingers. Many people entering residential, IOP, PHP or outpatient care already wear a device that records sleep, activity and heart rate. For executives and clinical leaders, the question is no longer whether this data exists. It is whether it can help a care team notice that a high-risk patient is struggling days before a missed appointment, a positive screen or a crisis call.
The answer is a qualified yes. Wearables will not replace clinical judgment, toxicology or validated assessments, but they add a continuous signal between those touchpoints, especially in the weeks after discharge. This guide covers the market, the evidence, HIPAA, Part 2 and billing rules, and how to roll out a program without burying staff in alerts.
Verity for Patients brings wearable data into the same platform programs already use for engagement and outcomes, with a focus on high-risk populations such as residential discharges, IOP and PHP step-downs, and substance use treatment.
Verity also offers the platform for other organizations, including health plans and provider networks, to deploy in their own programs. See how it fits residential treatment, IOP and PHP programs, substance use treatment programs and health plans, or start with the Verity for Patients platform overview.
Consumer platforms are adding mental health features, wearable companies are moving into clinical care, and researchers are building sensors for substance use risks.
With iOS 17 and watchOS 10, Apple added mood logging and "the same depression and anxiety assessments often used in clinics" to the Health app (Apple). Becker's identified those assessments as the PHQ-9 and GAD-7. Google's Fitbit Body Response feature on the Sense 2 uses electrodermal activity, heart rate, heart rate variability (HRV) and skin temperature to flag possible stress (Google).
Oura, the smart ring company, has grown through acquisitions, including Proxy in 2023 and metabolic health startup Veri and Sparta Science in 2024 (TechCrunch). In October 2025 it launched Cumulative Stress, a score built from a month of sleep, heart stress-response, temperature and activity data (TechCrunch). In April 2026 it acquired Galen AI, which links wearable data with medical records (the5krunner).
The biggest move came on November 10, 2026, when Oura announced it would acquire Counsel Health, an AI-native, physician-supervised care platform. Oura said the deal moves it "from helping members understand their health to enabling them to act on it," with care delivered inside the Oura app (Oura). Bloomberg reported the price at $625 million (via Let's Data Science). Oura has not framed the deal around behavioral health, but wearable data is clearly becoming a front door to clinical care.
A review in Current Addiction Reports found wearable sensors were the most commonly studied mobile technology for substance use disorder, with uses including stress and craving detection, opioid intake detection and overdose monitoring (Goldfine et al., 2020). In Australia, a study at two supervised injecting facilities is training a chest-worn sensor to detect opioid-induced respiratory depression (JMIR Research Protocols, 2024). Continuous alcohol monitoring is already established in justice settings: the SCRAM CAM ankle bracelet tests sweat for alcohol every 30 minutes (SCRAM Systems).
Some signals have decades of research behind them. Others are promising but immature.
| Signal | What the research shows | How to use it today |
|---|---|---|
| Sleep (onset, duration, continuity) | Sleep problems, especially difficulty falling asleep, have long been linked to relapse risk in alcohol use disorder (Brower, 2001). | Strong candidate for early-warning rules and check-in prompts. |
| Heart rate variability | A 2024 review of 14 studies found lower resting HRV in people who use substances, associated with higher stress, craving and symptom severity (Moon et al., 2024). | Trend each patient against their own baseline. Do not compare across patients or devices. |
| Activity and steps | Activity data is a common input to wearable depression models, and sharp drops can reflect low mood, withdrawal or disengagement. | Use with check-ins. Injury, a new job or weather can explain a drop. |
| Algorithm-based depression detection | A 2023 meta-analysis of 54 studies found accuracy between 0.70 and 0.89, with small samples and very high heterogeneity. The authors said wearable AI should be combined with questionnaires or interviews (Abd-Alrazaq et al., npj Digital Medicine). | Never treat as a diagnosis. Pair with the PHQ-9 and clinical review. |
| Overdose and withdrawal detection | Early-stage sensor and algorithm studies (see above). | Watch the field. Not yet standard of care. |
The most reliable use is change detection within one patient. Someone whose sleep falls from seven hours to four for several nights, whose steps collapse, and who skips two check-ins is worth a call.
Want early warning signs routed to the right staff member?See how Verity connects wearables to your programs.
A consumer wearable's data is generally outside HIPAA while it sits only in the patient's own app. That changes when your organization receives it: once identifiable data from a patient's device lands in a covered entity's systems, it is protected health information. HHS guidance also says that when a covered entity provides or contracts for a remote monitoring app, the developer is a business associate (summary of HHS OCR FAQs). In practice:
Sometimes, and the rules are tightening. Medicare pays for two families of codes: remote patient monitoring (RPM), for physiologic data such as heart rate, and remote therapeutic monitoring (RTM), for therapeutic data. For 2026, CMS added codes for 2 to 15 days of readings and a 10-minute management tier, including CPT 99445 and 99470 for RPM (ChronicCareIQ).
The CY 2027 final rule, issued November 4, 2026, limits RTM to established patients and requires a face-to-face initiating visit, in person or by telehealth, before RPM or RTM begins, starting January 1, 2027. A requirement that billable monitoring time come from clinical staff who are direct employees of the billing practice was delayed to January 1, 2029, and several device and data codes face practice expense cuts phased in over at least two years (Nixon Peabody). Our 2027 fee schedule guide covers the wider impact.
Device and documentation requirements are specific, so confirm with your billing team whether a consumer device qualifies, and remember that Medicaid and commercial coverage vary. For most programs, the stronger case is clinical and operational: earlier outreach and better post-discharge outcome data.
Illustration, not a statistic. Consider a program that discharges 30 patients a month. Assume 6 of them show a sustained drop in sleep or activity during their first month home. If earlier outreach helped even one of those six avoid a readmission each month, multiply your own average cost of a readmission or crisis episode by 12 to estimate the annual effect. Then subtract staff time: at 15 minutes to review and act on each alert, six alerts a month is about 1.5 staff hours. Use your own volumes and costs. The point is to compare a small, predictable staff cost with a large, unpredictable one.
Biometric data in behavioral health is a practical tool for high-risk patients when it is used as an early-warning layer, not a diagnosis. Sleep and HRV have the strongest support; pair algorithmic detection with validated assessments. Treat device data as PHI once you receive it, apply Part 2 where it applies, and check the 2027 RPM and RTM rules before counting on reimbursement. Start with one population, a few signals and clear owners, and review results every month.
To see how wearable signals, check-ins, rewards and assessments work together in one platform, book a demo of Verity.
Verity works with any EHR and connects to 200+ systems through API or secure file transfer, so appointments, attendance, and assessment results flow automatically.
Sleep, activity and heart rate are the most practical starting points. Sleep disruption has long been linked to relapse risk, and lower heart rate variability is associated with stress and craving in people with substance use disorders. Watch for changes from each patient's own baseline, alongside check-ins and validated assessments, rather than fixed thresholds.
Not while it sits only in a patient's personal app. Once your organization receives identifiable wearable data, it is protected health information, and vendors that store or process it for you need a business associate agreement. SUD programs subject to 42 CFR Part 2 should also treat it as part of the patient's SUD record.
Not on their own. Research links poor sleep and lower heart rate variability to craving and relapse risk, and machine learning models show promise, but accuracy varies and most studies are small. Treat wearable changes as a prompt for outreach and clinical review, not as proof that a patient has relapsed.
Possibly, through remote patient monitoring or remote therapeutic monitoring codes, but the requirements are specific. From January 2027, RTM is limited to established patients, and both require an initiating visit before monitoring starts. Device rules, documentation and payer policies vary, so confirm eligibility with your billing team before building a program around reimbursement.
Start with a few signals, wait for a baseline before alerting, and alert on meaningful change rather than single readings. Give every alert type a named owner and response time. Review alert volume and false alarms monthly, and retire rules that rarely lead to action. Fewer, better alerts are the ones that get answered.