Wearables and continuous glucose monitors have changed quickly. In the US, CGMs are now sold without a prescription, smartwatches carry FDA-cleared hypertension notifications, and in January 2026 the FDA loosened its rules for wellness wearables. In hospitals, chest patches and wrist sensors stream vital signs to AI models that try to spot deterioration hours before it is obvious.
For hospitals and clinics, the question is practical: which of these data streams can improve care, and what does it take to use them without drowning staff in alerts? This guide covers what is new in 2025-2026, what the evidence says, and how to design a remote monitoring programme that works in India.
The short version: start with a clear clinical question, a defined patient group and a team that will act on the data. The device is the easy part.
What changed in 2024-2026
- Over-the-counter CGMs. In March 2024 the FDA cleared Dexcom's Stelo as the first glucose biosensor sold without a prescription, and in June 2024 Abbott received clearance for its Lingo and Libre Rio over-the-counter systems.
- AI features on CGM apps. Dexcom has added generative AI features to Stelo, including AI-assisted food logging and daily insights, with further updates announced in February 2026. Oura added glucose tracking with Stelo pairing in 2025.
- Smartwatch hypertension alerts. Apple's hypertension notification feature was cleared by the FDA in September 2025.
- FDA wellness and CDS updates. On 6 January 2026 the FDA revised its general wellness guidance to cover more non-invasive wearables that estimate values such as blood pressure, oxygen saturation or glucose, provided outputs are framed for wellness rather than clinical use, according to Foley's analysis. Commentators linked the change to a 2025 FDA warning letter to WHOOP over an uncleared blood pressure feature.
- AI on ward wearables. Northwell Health researchers reported in Nature Communications in November 2025 that an AI model using data from clinical wearables flagged at-risk inpatients an average of about 17 hours before deterioration.
Consumer versus clinical devices
This is the most important distinction for a hospital.
| Device type | Examples | Accuracy and claims | Hospital use |
|---|---|---|---|
| Consumer wellness wearables | Fitness bands, smartwatches (general features) | Wellness claims; not for diagnosis | Engagement, activity, sleep trends |
| Cleared consumer features | Smartwatch ECG or hypertension notifications | Specific cleared claims with known limits | Prompts to see a doctor, not diagnosis |
| OTC CGMs | Stelo, Lingo | Glucose data for non-insulin users | Lifestyle programmes, prediabetes |
| Prescription CGMs | Libre, Dexcom G-series | Clinical glucose management | Diabetes care, insulin titration |
| Clinical wearables | Chest patches, medical-grade wrist sensors | Cleared vital-sign monitoring | Ward monitoring, post-discharge, hospital at home |
Do not build clinical decisions on wellness data. Use it for engagement and prompts; use cleared clinical devices for monitoring that drives care.
Reading smartwatch alerts correctly
Apple's hypertension feature illustrates the limits well. According to the validation data summarised by the American Academy of Family Physicians, the feature had a sensitivity of 41.2% and a specificity of 92.3%. In plain terms, most people with undiagnosed hypertension will not get an alert, and a small share of people without it will.
For clinics this means:
- An alert is a reason to measure blood pressure properly, not a diagnosis.
- No alert does not mean normal blood pressure. Keep routine screening.
- Expect patients to arrive with smartwatch data. Have a simple protocol for it.
CGMs in clinical practice
CGMs give a continuous glucose profile instead of occasional finger-stick readings. Metrics such as time in range, glucose variability and patterns around meals or overnight give doctors far more to work with.
Practical uses for Indian hospitals and diabetes clinics:
- Diabetes clinics: 10-14 day sensor wear before a review visit, then adjust treatment based on the pattern.
- Insulin initiation and titration: fewer missed hypoglycaemia episodes.
- Gestational diabetes: closer monitoring where clinically indicated.
- Inpatients: some hospitals use CGMs on wards to reduce finger sticks; check device labelling and local protocols for inpatient use.
- Lifestyle and prediabetes programmes: OTC-style sensors combined with coaching.
The AI layer is mostly pattern detection and summarisation: highlighting nocturnal lows, post-meal spikes or adherence gaps, and drafting a summary for the doctor. Keep dosing decisions with clinicians.
Ward wearables and AI deterioration models
Ward patients typically get vital signs every four to eight hours. Continuous wearables fill the gaps, and AI models look for trends.
Evidence is promising but still maturing. The Northwell study is a model development and validation study, not a trial of patient outcomes. Randomised trials of continuous ward monitoring with alerting are under way, including a UK trial in post-operative patients that completed recruitment in December 2025. Meanwhile the lesson from sepsis AI applies: alerts only help if they reach someone who acts, and false alarms are a real cost. Our guide to AI sepsis early warning systems covers alert design in detail.
Designing a remote monitoring programme
| Element | Questions to answer |
|---|---|
| Clinical goal | What outcome are we trying to change: readmissions, hypoglycaemia, early deterioration? |
| Patient group | Who qualifies, and who does not? |
| Devices | Cleared for this use? Patient can wear and charge it? |
| Data flow | Device to app to vendor cloud to our system: how and how often? |
| Thresholds | What triggers review? Who set them? |
| Response team | Who watches, during which hours, and what do they do? |
| Escalation | When does a nurse call the patient, a doctor review, or the patient come in? |
| Documentation | Where are readings, alerts and actions recorded? |
| Consent and privacy | What has the patient agreed to? Where is data stored? |
| End point | When does monitoring stop? |
The response team is where most programmes struggle. A nurse-led monitoring desk working defined hours, with clear escalation to doctors, is more sustainable than sending alerts to busy ward staff.
Integration and data
Most device vendors provide their own app and cloud, and many offer APIs. For a hospital, the goal is to bring the relevant data into the patient's record, not to make clinicians log into five portals.
A practical architecture:
- Vendor APIs or device gateways pull readings at agreed intervals.
- An integration service maps readings to the patient (UHID, ABHA where linked) and normalises units.
- Rules or AI models compute summaries and flag readings needing review.
- A monitoring dashboard shows the nurse desk a prioritised list.
- The HIS or EMR stores summaries, alerts and actions, with FHIR Observation resources a natural fit for vitals and glucose.
Our HL7 FHIR integration guide explains Observation resources, and our telemedicine software guide covers the video and messaging side. Algon builds this kind of healthcare integration and dashboard layer for hospitals.
Indian considerations
- Cost. Sensors are a recurring cost, often paid out of pocket. Programmes must show clear value to patients.
- Connectivity and phones. Many devices depend on a smartphone app. Check whether elderly patients have, and can use, a compatible phone, or whether a caregiver will.
- Language. Instructions and alerts in Malayalam, Hindi or other local languages improve adherence. See our guide to AI discharge instructions in Indian languages.
- Regulation. Wearables that measure or interpret clinical values may be medical devices under CDSCO rules. India's guidance excludes general wellness apps, but clinical claims change that.
- Privacy. Continuous health data is personal data under the DPDP Act and deserves extra care; consent should be specific and data flows documented.
Costs
Indicative cost drivers for a hospital programme:
- Devices or sensors, often recurring.
- Vendor platform or API fees.
- Integration and dashboard development.
- Monitoring staff time, the largest ongoing cost.
- Patient onboarding and support.
How these add up depends mostly on programme length and monitoring hours. A two-week CGM review before a diabetes clinic visit is a small, bounded cost. A 24x7 monitoring desk for post-discharge heart failure patients is a staffed service with salaries, shift cover and escalation rotas. Decide which model you are building before you price devices.
How to start
- Pick one programme with a clear outcome, such as post-discharge monitoring for heart failure patients or CGM-guided review for poorly controlled diabetes.
- Choose cleared devices suited to that use and your patients.
- Define thresholds and escalation with clinicians before enrolling anyone.
- Staff a monitoring desk with set hours and backup.
- Integrate summaries into the record so doctors see data at the next visit.
- Measure outcomes for 3-6 months against a comparison group, then decide whether to scale.
Our patient engagement work covers onboarding and reminders that keep patients wearing and charging devices.
Mistakes to avoid
- Buying devices before defining who will watch the data and when.
- Sending raw alerts to ward phones with no triage, creating alarm fatigue within weeks.
- Treating wellness readings as clinical measurements.
- Enrolling patients indefinitely with no end point or review.
- Ignoring device adherence: a sensor that is not worn or charged produces silence, which can look like stability.
- Keeping data in the vendor portal only, so doctors never see it at the next visit.
Conclusion
Wearables, CGMs and AI give hospitals more data about patients between visits and between nursing rounds than ever before. In 2026 the technology is accessible: OTC glucose sensors, cleared smartwatch features, clinical-grade ward wearables and AI models that read trends. The evidence is strongest for CGM in diabetes care and still building for AI-driven ward monitoring.
The programmes that work start with a clinical goal, use cleared devices for clinical decisions, bring data into the record, and staff a team to act on it. Start with one patient group and measure.
Frequently asked questions
What is remote patient monitoring with wearables?
Using devices such as CGMs, chest patches, blood pressure cuffs and smartwatches to collect health data outside regular visits or between nursing rounds, with a care team reviewing trends and alerts and acting when needed.
Are over-the-counter CGMs accurate enough for diabetes care?
OTC CGMs such as Dexcom Stelo and Abbott Lingo are intended mainly for people not using insulin. Patients on insulin or with complex diabetes should use prescription CGMs under clinical supervision.
Can a smartwatch detect high blood pressure?
Apple's FDA-cleared hypertension notification had 41.2% sensitivity and 92.3% specificity in validation. An alert should prompt proper blood pressure measurement; no alert does not rule out hypertension.
What changed in FDA rules for wearables in 2026?
In January 2026 the FDA updated its general wellness guidance to cover more non-invasive wearables that estimate values such as blood pressure or glucose, as long as outputs are framed for wellness and not for specific clinical use.
Can AI predict deterioration from ward wearables?
Research is promising. A Northwell Health study reported an AI model flagging at-risk inpatients about 17 hours before deterioration on average, but randomised trials of patient outcomes are still under way.
How do hospitals integrate wearable data into records?
Through vendor APIs or gateways feeding an integration service that maps data to the patient, computes summaries and alerts, and stores results in the HIS or EMR, often as FHIR Observation resources.
