Wearable and Phone Data in Dementia Care

Wearables and smartphones can show a care team how someone's sleep, activity and walking are changing over time. Research suggests some of those patterns may be linked to early cognitive change in groups of people. What the data can't do is diagnose dementia, explain why a pattern changed, or reliably predict what will happen to one person. Used well, these signals are a prompt for a conversation. Used badly, they create false alarms and false reassurance.

This guide covers the kinds of signals people are studying, what the research shows, what the data can't tell you, and how care teams and program leads can use it responsibly. It is written for provider teams, clinical leaders and technology buyers. It is general information and not clinical or legal advice.

What Kinds of Signals Are We Talking About?

A 2026 systematic review of wearables for early detection of cognitive impairment and dementia describes the opportunity. Traditional cognitive screening relies on episodic clinical assessments and may miss early changes. Wearable and mobile health technologies can monitor sleep, physical activity and circadian rhythms continuously, generating what researchers call digital biomarkers that may support early detection at scale.

The main categories are:

• Sleep and daily rhythm. When someone sleeps, wakes and is active.

• Physical activity. Steps, movement and time spent sedentary.

• Walking and gait. Speed and how steady or variable a person's walk is.

• Navigation. How someone moves through and finds their way around spaces.

• Phone and device use. Routine use of a phone or watch, sometimes paired with brief cognitive tasks.

What Does the Research Show?

Reviews find promise, with caveats. Two systematic reviews of navigation and gait measures in Alzheimer's & Dementia included 27 navigation studies and 83 gait studies. The authors highlight several measures that may identify people at risk, and they note that different devices can capture them. Their eligibility criteria were deliberately strict. They included only cross-sectional studies of people with biomarker-confirmed early disease or higher genetic risk, which means the findings describe differences between groups in research settings, not predictions for individuals in routine care.

The evidence is still scattered. The 2026 systematic review says current evidence remains fragmented across devices, analytic approaches and cognitive outcomes. That is a polite way of saying the studies use different tools and methods, which makes them hard to compare or apply.

Large studies are testing feasibility. The Intuition study enrolled 23,004 U.S. adults who used iPhones and Apple Watches, according to a paper in Nature Medicine. It collected routine device use, self-reported health information and cognitive assessments over an extended period, with the aim of classifying mild cognitive impairment and tracking cognitive trajectories. The paper describes the design and the participants at the start. Nature's summary calls the results a demonstration of feasibility and a route to methods for early detection, and it doesn't present the data as a clinical test.

Gait work in people already living with dementia calls for standard methods. A systematic review of wearable gait assessment in people living with dementia concludes that more studies using standardized protocols are needed to explore usefulness for things like fall prognosis and early diagnosis.

Acceptability matters. In the CODEC-II study, participants used four kinds of wearable technology for two weeks every three months over a year, and researchers interviewed a subgroup about their experience. A device that people won't wear or tolerate can't produce useful data.

What the Data Can Tell a Care Team

The most defensible use is noticing change against a person's own baseline. Someone who is usually active by mid-morning and now isn't, or whose nights have become restless, has given the team a reason to ask a question. Signals like these can help a care team:

• Decide who to contact sooner

• Ask more specific questions on a call

• Notice changes that a family member hasn't mentioned

• Bring a clearer picture to an appointment

The data adds context. It works best alongside what the person and their caregiver say.

What the Data Can't Tell You

It can't diagnose. A pattern in sleep or steps isn't a diagnosis. Diagnosis requires a clinical assessment.

It can't explain a change. Less walking could mean pain, a cold, low mood, bad weather, a broken device or a new medication. The data shows that something changed, not why.

It can't predict one person's future. The research describes groups. Group differences don't tell you what will happen to an individual, and consumer devices weren't built or validated as dementia tests.

It can be wrong in both directions. Missing data, a watch left on the charger or a different wrist can look like a change. A real change can go unseen if the person isn't wearing the device.

It reflects who is in the studies. The Intuition study lists limited representativeness in current cognitive health research as a problem it aims to address. Results from one population may not apply to another, and people with dementia may be less likely to wear or manage devices.

Regulatory and Privacy Questions

Software that turns sensor signals into a diagnosis or recommendation can be regulated as a medical device. According to a summary of FDA's January 2026 clinical decision support guidance, software that processes sensor signals to generate a diagnosis is generally treated as a device. If a vendor says its product detects or predicts dementia from wearable data, ask about its FDA status and the evidence behind the claim.

Privacy and consent matter too. Continuous data about someone's movement and habits is sensitive, and a person living with dementia may not be able to weigh the trade-offs alone. Involve your compliance and legal advisers, get informed consent from the person and, where appropriate, their legal representative, and collect only what you'll use.

How to Use These Signals Responsibly

1. Decide what you'll do with a signal before you switch it on. If nobody knows who acts on a change, the data becomes noise.

2. Compare people with themselves. A person's own baseline is more meaningful than a population average.

3. Pair data with a human check. A short call to ask "how has the week been?" tells you more than a graph.

4. Keep it simple. A few signals that your team understands beat a dashboard nobody trusts.

5. Watch for alert fatigue. Too many flags teach staff to ignore them.

6. Say what it isn't. Tell families and staff that the data is a prompt, not a diagnosis.

7. Review what happens. Track how often a signal led to a useful conversation, and adjust.

Questions to Ask a Vendor

• What does the product actually measure, and how was that validated in people with dementia?

• What evidence exists, who was studied and how were results measured?

• What happens when data is missing or wrong?

• Has the vendor assessed whether any feature falls under FDA's definition of a device?

• Who owns the data, who can see it and how is consent handled?

• How does the product avoid overwhelming staff with alerts?

Our guide to choosing dementia care technology for a value-based program covers the wider checklist.

Where Elli Cares Fits

Elli Cares connects to Apple HealthKit and Android Health Connect, and it gives care teams a clinician dashboard, AI-generated summaries, daily wellbeing check-ins and family observations. The idea is to put what a device shows alongside what people report, so a care navigator can see who may need earlier outreach. Elli isn't a diagnostic tool, and it makes no claim that wearable data detects or predicts dementia. You can read more on our Value-Based Care page.

Frequently Asked Questions

Can a smartwatch detect dementia?

Not reliably for an individual. Research shows some wearable-derived measures, such as gait and activity patterns, differ between groups of people with early disease and others, but evidence remains fragmented and consumer devices aren't validated as dementia tests.

What can wearable data tell a care team?

Changes in a person's sleep, activity and walking compared with their own baseline. Those changes can prompt a conversation or earlier outreach. They don't diagnose or explain the cause.

Is gait a sign of dementia?

Gait is being studied as a possible early marker, and reviews highlight several gait and navigation measures that may identify people at risk in research settings. A change in walking has many possible causes, so it needs clinical assessment.

Is dementia monitoring software regulated by the FDA?

It can be. FDA's January 2026 clinical decision support guidance, as summarized by industry sources, treats software that processes sensor signals to generate a diagnosis as generally a device. Ask any vendor how it has assessed its features.

What should a program consider before using wearables?

Consent, privacy, acceptability for the person, how missing data is handled, who acts on a signal and how to avoid alert fatigue.