Digital Phenotyping: How Wearables Are Changing Cognitive Health Screening

By the time most people find out they're losing their memory, the window to do something about it may have already closed. For decades, we've relied on pen-and-paper tests administered once a year to catch a disease that develops over decades. This is starting to change. Since the 1970s, doctors have used standardized cognitive assessments to detect decline. Patients complete tasks evaluating memory, attention, and visuospatial skills, recalling words, drawing shapes, and performing serial calculations. Tools like the Mini-Mental State Examination (MMSE) and Montreal Cognitive Assessment (MoCA) were revolutionary when introduced, providing valuable diagnostic snapshots. But that’s all these tests provide, snapshots, only providing insight into a single moment under artificial conditions, usually administered once per year (Wang et al., 2017)

By age 90, 40% of people will receive a dementia diagnosis (Alzheimer’s Society, 2021). Yet research shows that up to 40% of all dementia cases could be prevented or delayed through lifestyle interventions (Livingston et al., 2020). The gap between these statistics represents an enormous opportunity for prevention through earlier intervention. 

The challenge is timing. By the time cognitive decline registers on standardised tests, years of brain changes have already occurred. The optimal window for intervention, when the brain still has plasticity and treatments can meaningfully alter trajectories, may have already closed. Moreover, these clinical assessments are vulnerable to confounding factors: poor sleep the night before, caffeine intake, and test anxiety can all skew results (Moshe et al., 2021). 

Digital phenotyping offers a different approach. Rather than relying on annual snapshots in artificial settings, it continuously monitors behavioural and physiological patterns through the smartphones and wearables people already own. Early signs of cognitive decline don't first appear during memory tests; they show up in how someone sleeps, moves through their day, types on their phone, and navigates their neighbourhood.

By tracking changes across five key domains of daily life, digital biomarkers can detect subtle shifts years before they become apparent on traditional screening. 

Sleep and Circadian Rhythms

Sleep stands out as one of the most reliable early warning signs of cognitive decline. How quickly someone falls asleep, how many times they wake up during the night, and how much actual sleep they get. These patterns reveal a lot about brain health trajectories.

As it turns out, disrupted sleep-wake cycles can show up before someone is diagnosed with Alzheimer's, Parkinson's, or Huntington's. Research showed that people who spend more time in bed (not necessarily sleeping) are more likely to show signs of depression. Those who wake up frequently during the night consequently have higher anxiety levels. What is most telling is frequent waking during the night. The strongest predictor of cognitive decline isn't how long someone sleeps, but how often they wake during the night, what researchers call "wake after sleep onset." In a five-year study of 759 older adults, researchers used wrist actigraphs to objectively measure sleep patterns and compared them against cognitive assessments. They found that participants with more fragmented sleep showed measurably worse cognitive function over time, even when they reported feeling well-rested. Someone might believe they're sleeping fine, logging 8 hours in bed each night, but their wearable reveals brief awakenings scattered throughout the night, reducing actual consolidated sleep. These disrupted patterns can signal cognitive decline years before memory problems register on standard tests, and most people remain completely unaware (Lauderdale et al., 2019).

What makes sleep disruption particularly alarming is what's happening beneath the surface. Chronic insomnia is physically linked to volume loss in specific regions of the hippocampus, the brain's memory center. As these structural changes accumulate, they create a cascade effect: worse sleep leads to hippocampal atrophy, which impairs memory consolidation and emotional regulation, which increases vulnerability to further cognitive decline. Consumer wearables like Fitbit and Oura Ring can track these sleep-wake patterns continuously, providing early indicators of this deterioration years before it becomes clinically apparent (Moshe et al., 2021, Menassa et al., 2025)

Mobility and Spatial Navigation

Smartphone GPS data reveals something fascinating. How far people travel from home and how varied their locations can signal cognitive decline.

Researchers found a clear pattern. Healthy older adults have the widest travel radius and visit the most diverse locations. People with mild cognitive impairment travel less. But people with dementia tend to move in the tightest, most repetitive circles. The same grocery stores, coffee shops and routes on the way home with little to no variation. One study quantified this and found that people with dementia had only a 5% chance of visiting a random new location compared to 8% for healthy people. A person might stop going to the library on Tuesdays or start taking the same route to the grocery store. The "exploring" part of their routine slowly disappears. But their phone's GPS data shows the contraction clearly; a life space that used to span 15 miles is now contained within 6 blocks. When an older adult's world begins to subtly shrink, passive monitoring can pick up these signals and flag them as potential early markers of cognitive decline (Cullen et. al, 2022). 

Fine Motor Skills and Gait Coordination

Another answer to early detection may be right at our fingertips. Researchers are now measuring "keystroke dynamics". This entails how long each key is held down, how much time passes between keystrokes, and the rhythm of typing. People with mild cognitive impairment type differently. Slower, more variable, less consistent. One study showed these typing patterns could identify cognitive impairment with 78% accuracy, and the patterns correlated directly with traditional test scores (Moshe et al., 2021).

Smartwatches track something similar during walking, but the mechanics are more elegant than you might expect. Researchers have developed a "Smartwatch Gait Coordination Index" that uses the gyro sensors already built into devices like the Apple Watch or Galaxy Watch. When we walk, our arms and legs move in a coordinated pendulum motion, a natural rhythm that smartwatches can capture as sinusoidal waves. By analyzing the peaks and valleys of your arm swings, the watch detects subtle asymmetries or timing shifts in your gait. These slight irregularities, invisible to the naked eye, can signal early motor dysfunction linked to Parkinson's and dementia. No expensive gait lab required, just the device already on your wrist (Han & Paul, 2023).

Language and Cognitive Output

Those struggling with dementia often show a reduced level of complexity in their speech. And these difficulties can reveal themselves in communication before they’re truly noticeable in day-to-day speech. 

Researchers can now analyse spontaneous writing to track vocabulary diversity, sentence complexity, and word choice patterns. Do people use fewer specific nouns and more vague pronouns? Are sentences getting simpler? Is vocabulary shrinking?  Something as simple as an email that once would have said, "Could you forward the Q3 analysis from the finance meeting?" might now read, "Can you send that document from last week?" The meaning is clear enough, but the linguistic precision has quietly eroded.

People with mild cognitive impairment show measurably less vocabulary richness and simpler sentence structures in their everyday writing. Combined with typing rhythm changes, digital communication creates a detailed cognitive fingerprint (Moshe et al., 2021). 

Physiological Stress Markers

Heart rate variability (HRV), the variation in time between consecutive heartbeats, provides a window into nervous system function and stress response.

Higher HRV usually means the body handles stress well. Lower HRV, on the other hand, is associated with chronic stress and anxiety (Moshe et al., 2021). But chronic stress, on top of affecting mood, can actually accelerate brain aging through sustained cortisol elevation, inflammation, and even shrinkage of the hippocampus, the brain's memory center (Joo Kim et. al, 2015). 

The same fitness trackers measuring sleep are tracking this too. Every heartbeat is logged passively, revealing patterns in stress response that could signal trouble ahead.

What Comes Next

Synthesizing this evidence base is just the beginning. The next step is integrating these digital biomarkers into Health Impact Alliance's Independence Scorecard, a comprehensive system for predicting nursing home admission risk and enabling early intervention.

Health scorecards distill complex health-related data sets into simple, robust numeric summaries to support diagnostic, evaluative, or prognostic decision-making (Baxter, 2023). HIA's Independence Scorecard applies this approach to the domains explored throughout this article: sleep quality, mobility patterns, fine motor coordination, language complexity, and physiological stress markers. Rather than relying on annual clinical assessments that capture only snapshots under artificial conditions, the scorecard continuously aggregates passive data from devices people already own, like smartphones, smartwatches, and fitness trackers, to generate a composite score reflecting overall independence risk.

This builds a future where cognitive decline is detected not when someone fails a memory test at age 75, but when their sleep patterns shift at age 68, when their daily walking routes become slightly more restricted at 70, and when their typing rhythm slows imperceptibly at 72. The goal is intervention during the window when cognitive plasticity still exists and treatment can alter trajectories, before crisis forces reactive decisions about care.

The loss of independence isn’t always dramatic. It happens gradually through decades, culminating through small losses across multiple domains, each individually insignificant but collectively predictive. By synthesising behavioural data into a single, interpretable score, HIA can provide families what they need most: advance warning. Time to arrange support services thoughtfully. Time to make housing decisions before emergencies force them. Time to maintain quality of life for both the older adult and their caregivers, measured not by clinical thresholds but by each person's own standards for living well.