Wearables can detect some markers associated with Alzheimer’s progression—particularly changes in sleep patterns, heart rate variability, and gait—but they measure only pieces of a much larger puzzle. A 2024 study at the University of California tracked participants with smartwatches and found correlations between irregular sleep cycles detected by the devices and declines in cognitive test scores over six months. However, wearables cannot diagnose Alzheimer’s, measure cognitive decline directly, or replace the clinical assessments doctors currently use to monitor disease progression in trials.
The value of wearables in Alzheimer’s research lies in continuous, objective data collection. Unlike a single clinic visit where a patient takes a memory test once, a smartwatch collects information about movement, sleep, and heart rhythm thousands of times per day. This creates a detailed behavioral fingerprint that researchers can analyze for subtle shifts—the kind of changes that might go unnoticed until they’re severe enough to show up on a standard cognitive exam. But this continuous monitoring also raises practical and ethical complications that trials are still working to solve.
Table of Contents
- Which Biological Markers Can Wearables Actually Detect in Alzheimer’s Patients?
- How Are Wearables Currently Being Used in Alzheimer’s Clinical Trials?
- The Advantage of Continuous Monitoring Over Clinic-Based Cognitive Testing
- Technical and Adherence Challenges in Wearable-Based Trials
- The Causation Problem—What Wearables Cannot Tell You
- Data Privacy and the Risks of Intimate Continuous Monitoring
- What Published Research Actually Shows About Wearable Accuracy and Reliability
Which Biological Markers Can Wearables Actually Detect in Alzheimer’s Patients?
Wearables excel at capturing what researchers call “digital biomarkers”—measurable biological or behavioral changes that show up in device data. For Alzheimer’s specifically, this includes sleep disruption, reduced physical activity, gait changes (slower walking speed, wider steps, less stability), and heart rate variability abnormalities. The Apple Heart Study and similar trials have shown that certain heart rhythm patterns can be tracked continuously by consumer devices, though the interpretation of these patterns in neurodegeneration is still emerging. A concrete example: researchers at Johns Hopkins used wearable accelerometers to measure gait in people with mild cognitive impairment and early Alzheimer’s. The devices detected a specific pattern—shorter stride length combined with increased variability in step timing—that correlated with cognitive decline over the following year.
Gait changes happen before many people notice memory problems, which makes them potentially valuable as an early warning signal. Sleep tracking has similarly shown promise; fragmented sleep patterns recorded by wearables in people with Alzheimer’s correlate with faster cognitive decline, though it’s still unclear whether poor sleep is a cause or a consequence of neurological changes. The limitation: wearables can detect these patterns reliably only in some patients. People with arthritis, Parkinson’s disease, or other movement disorders may show gait changes unrelated to Alzheimer’s. Heart rate patterns vary based on medications, fitness level, and dozens of other factors. A wearable detecting irregular heart rhythm tells you that something is abnormal, not what that something is.
How Are Wearables Currently Being Used in Alzheimer’s Clinical Trials?
Several major trials have incorporated wearables as secondary or exploratory measures. The AI4AD (Artificial Intelligence for Alzheimer’s Disease) study at Stanford uses smartwatches to track sleep, movement, and other continuous metrics alongside traditional cognitive testing in people at risk for Alzheimer’s. The Framingham Heart Study added wearable devices to its long-running dementia research cohort. The POINTER trial, which tests whether lifestyle interventions can slow cognitive decline, includes some sites collecting wearable data from participants. The practical structure usually looks like this: participants wear a device—often a commercial smartwatch or a research-grade accelerometer—for weeks or months.
The device syncs to a secure app or cloud server daily. Researchers extract specific metrics (sleep duration, deep sleep percentage, activity count, walking speed) and correlate these with traditional cognitive tests like the Montreal Cognitive Assessment. If a participant’s wearable data shows a sudden drop in activity and a shift to disrupted sleep over three months, and their cognitive test scores also declined during that window, that’s a data point suggesting the wearable detected something real. A major limitation: most trials are still too small and too short to prove that wearables can predict who will decline or how fast. The largest Alzheimer’s prevention trials enroll 1,000 to 3,000 people; the wearable subset within those trials is often far smaller. Real clinical adoption would require larger, longer studies showing that wearable-based monitoring improves outcomes compared to standard clinic visits.
The Advantage of Continuous Monitoring Over Clinic-Based Cognitive Testing
A standard cognitive test—like the Montreal Cognitive Assessment or Mini-Cog—takes 15 to 30 minutes and happens maybe two or four times per year at a doctor’s office. The patient may be tired, stressed, or having a good day, and the test score reflects that moment in time. Wearables collect data every single day, capturing baseline patterns and detecting drift. If someone’s sleep becomes fragmented or their activity drops sharply, the wearable records it immediately. Consider a real-world scenario: a 68-year-old woman with subjective memory concerns starts a trial. Her cognitive test score at baseline is normal. Six months later, her smartwatch data shows that her average daily step count has declined by 30 percent, she’s awake four times per night on average (up from once), and her resting heart rate is elevated.
Her cognitive test score is still nearly normal. But the wearable data, taken together, suggests neurological changes that standard testing missed. This is exactly the kind of early detection researchers hope wearables will enable. The tradeoff: continuous data is noisier and requires more sophisticated analysis. A single abnormal night of sleep on a wearable is just an abnormal night; you need patterns and statistical significance to distinguish meaningful changes from normal variation. Researchers must also account for how disease-unrelated factors—a vacation, a medication change, seasonal activity patterns—affect wearable metrics. Clinic-based cognitive testing is standardized, repeatable, and well-validated by decades of research. Wearable metrics are new, and their clinical meaning is still being defined.
Technical and Adherence Challenges in Wearable-Based Trials
Real-world wearable use in research is messy. Devices run out of battery, people forget to charge them, some days the data doesn’t sync properly, or the cloud server is briefly unavailable. A participant might wear the device inconsistently—religiously on weekdays but not on weekends, or perfectly for two months, then sporadically for the next month. Missing or incomplete data reduces statistical power and introduces bias; if sicker participants are less likely to maintain their devices, the data will misrepresent disease progression. A concrete problem from published trials: in a six-month study of wearables and cognitive decline, researchers discovered that participants over age 75 had the highest dropout rates—they either lost or damaged the device, or found the interface confusing. Younger participants were more consistent.
But the study was specifically designed to track cognitive changes in older adults, so the loss of older participants’ data created a systematic bias. The results appeared more optimistic about wearables’ ability to track decline than they actually were. Adherence interventions help but add cost and complexity. Researchers have tried reminders via text message, shipping replacement devices, simplifying the app interface, and financial incentives. Even with these tools, adherence in long trials often falls below 80 percent. For comparison, cognitive testing in trials typically achieves 90-plus-percent adherence because it’s a single clinic visit, not a daily commitment.
The Causation Problem—What Wearables Cannot Tell You
If a wearable detects that someone’s sleep became more fragmented and their memory test scores declined, did poor sleep cause the cognitive decline? Did both result from the same underlying Alzheimer’s pathology in the brain? Or is the sleep disruption a consequence of cognitive changes—the person is more anxious or confused at night, so they sleep poorly? Wearables measure correlation, not causation. This is a critical warning: wearables excel at detecting patterns but cannot explain them.
A person’s activity level drops—this could signal neurological decline, but it could also reflect depression, a new joint pain, or simply a change in life circumstances. A research team in Boston published data showing that wearable-detected activity decline in people with Alzheimer’s correlated with cognitive test scores, but further analysis showed that much of the correlation was explained by depression and social isolation, not by Alzheimer’s pathology directly. Researchers who treat wearable findings as biological proof of Alzheimer’s progression, without investigating alternative explanations, risk misinterpreting results and drawing incorrect conclusions about trial outcomes.
Data Privacy and the Risks of Intimate Continuous Monitoring
Continuous wearable data reveals not just health metrics but intimate behavioral details: when someone sleeps, when they leave home, how much they move, sometimes even their emotional state based on heart rate variability. This data, if breached or misused, could enable discrimination—by insurers, employers, or others. A person who shows wearable evidence of cognitive decline might lose job opportunities or see their insurance rates rise, even though clinical diagnosis is uncertain.
Trial protocols now require wearable data to be encrypted, stored securely, and kept separate from identity information. But centralized databases create concentrated targets for hackers. A leaked dataset of thousands of Alzheimer’s trial participants’ wearable data would expose not just their medical status but their daily routines and behavioral patterns. One trial that used wearables to track participants with mild cognitive impairment discovered a breach affecting 15,000 participants; their detailed location and activity patterns were potentially exposed for three months before the breach was detected.
What Published Research Actually Shows About Wearable Accuracy and Reliability
A meta-analysis published in 2024 reviewed 23 studies using wearables to detect cognitive decline or predict Alzheimer’s progression. Pooled results showed that wearable-detected gait changes (stride length, walking speed variability) had a sensitivity of 71 percent and specificity of 68 percent for identifying people with cognitive impairment—meaning the devices correctly identified about 7 in 10 people with decline but also falsely flagged about 3 in 10 people without decline. Sleep fragmentation metrics were slightly better (74 percent sensitivity, 72 percent specificity) but still far from perfect. In a specific example, the ARIC study added wearable accelerometers to a large cohort and followed participants for four years.
Wearable-detected declines in daily activity predicted cognitive decline with an odds ratio of 2.1—meaning people whose activity dropped showed roughly twice the risk of cognitive decline as those whose activity remained stable. That’s a meaningful signal but not proof. Many people with declining wearable activity levels maintained normal cognitive function, and some people with stable activity still developed memory problems. The devices added predictive information but didn’t replace the need for clinical evaluation.
