The youngest group in the study comprised 500 university students with a mean age of 19.9 years; the middle group included 437 working-age adults averaging 41.1 years; and the oldest consisted of 1,030 community-dwelling adults with a mean age of 74.0 years. Led by sleep researchers Ju Lynn Ong and Shuo Qin under senior author Michael W. L. Chee, the team pooled the three Singaporean cohorts into a single analytical framework totalling 1,967 adults. Every participant wore an Oura Ring for four consecutive weeks, providing a consistent, device-based measure of time spent in physical activity, sedentary behaviour and sleep. Alongside the movement data, the researchers measured body mass index, body roundness index, pulse wave velocity and a composite vascular health score, producing a multidimensional portrait of cardiometabolic status. The design matters because much of the existing literature leans on questionnaires, which notoriously understate sitting time, or on mismatched devices and protocols that blur comparisons between age groups. Here, each person contributed the same four weeks of free-living recordings, and the protocols for the two younger cohorts were registered on ClinicalTrials.gov, part of a broader push to bring laboratory-grade rigour to real-world wearable data.

The headline numbers are striking. University students accumulated an average of 633.89 minutes of sedentary behaviour per day — just under ten and a half hours — while logging only 320.28 minutes of physical activity and 431.75 minutes of sleep, the lowest of all three groups on both counts. Older adults did precisely the opposite: they recorded the least sedentary time of anyone, 500.86 minutes per day, along with the most physical activity, 420.28 minutes, and the longest sleep, 455.62 minutes. Working-age adults fell in between on every measure, with 581.28 minutes of sitting, 346.55 minutes of activity and 445.77 minutes of sleep. Put differently, the average student sat roughly two hours and 13 minutes more per day than the average 74-year-old and moved about 100 minutes less. Even sleep, the one behaviour young adults are routinely assumed to win, went to the elders, who out-slept the students by nearly 24 minutes a night. Across all three behaviours, the gradient ran cleanly in the direction opposite to what expectations about ageing would predict.

The reversal becomes less puzzling once the texture of contemporary young adulthood is considered. Undergraduates spend their days shuttling between lectures, laboratories, libraries and screens, in environments engineered for long, unbroken sitting, and much of their social lives now unfolds through devices. Older adults in the study, by contrast, were community-dwelling and largely retired, holding far more discretionary control over their hours — time that, in Singapore’s dense and walkable urban environment, readily converts into walking to markets, running errands and taking part in social activities that keep people on their feet. The findings arrive amid mounting international evidence that physical activity plummets during adolescence and early adulthood, precisely the years when lifelong habits are cemented. What makes the new analysis forceful is the scale of the gap and the quality of the measurement: four unbroken weeks of objective data rather than the self-reports that dominate the field. As the authors put it, the findings “challenge conventional expectations about age-related inactivity.”

To link these daily time budgets to health, the researchers used compositional data analysis, a statistical framework built around a simple constraint: a day contains exactly 24 hours, so time spent moving, sitting and sleeping are not independent quantities but competing shares of a fixed whole. Every minute gained by one behaviour must be surrendered by another, and conventional regression, which treats each behaviour as a separate predictor, can therefore yield distorted estimates when the variables are mathematically interdependent. Compositional methods instead work with the relative geometry of the data, using log-ratio transformations to express each behaviour as a proportion of the 24-hour day and then modelling how a health outcome changes when minutes are reallocated among them. This allowed the team to pose questions with direct everyday meaning: if a person traded 30 minutes of sitting for 30 minutes of activity, or surrendered 30 minutes of waking time for more sleep, what would happen to their cardiometabolic profile? The method cannot prove causation, but it produces estimates that map far more naturally onto real decisions than coefficients from standard regression models.

The substitution results pointed consistently in one direction. Reallocating 30 minutes per day into physical activity from any combination of the other behaviours was associated with lower body mass index, down 0.13 kg/m²; a lower body roundness index, down 0.04 units; slower pulse wave velocity, down 0.05 m/s; and a slightly lower composite vascular health score, in keeping with reduced overall risk. Carving 30 minutes out of sedentary behaviour was linked to even larger effects on body composition: body mass index dropped by 0.28 kg/m² and body roundness index by 0.07 units. Redirecting 30 minutes into sleep at the expense of waking behaviours was associated with a body mass index lower by 0.21 kg/m² and a body roundness index lower by 0.05. In isolation these are modest figures — fractions of a unit for half an hour of daily change. But they are population-level associations, and small per-person differences sustained across years and multiplied across millions of young adults translate into meaningful shifts in obesity prevalence and cardiovascular burden.

The outcomes the team tracked deserve attention because together they measure far more than a bathroom scale can. Body mass index, the familiar ratio of weight to height squared, cannot separate muscle from fat. The body roundness index, calculated from height and waist circumference, is better tuned to central adiposity — the abdominal fat most tightly entwined with metabolic disease. Pulse wave velocity may be the most revealing measure of all: it records the speed at which a pressure wave travels along the large arteries, and arteries that have stiffened carry that wave faster. Arterial stiffness is a well-validated precursor of hypertension, heart attack and stroke, and it can advance silently for decades. The composite vascular health score bundles several such vascular measures into a single standardized index. That sedentary time and physical activity were associated not only with body composition but with the behaviour of the arteries themselves hints that the costs of a chair-bound young adulthood may be inscribed in the vasculature long before the first cardiac event.

The public health implications reach well beyond the campus gates. Non-communicable diseases are propelled by risk factors that accumulate over decades, and the architecture of adult habits is typically erected in late adolescence and early adulthood. If the most sedentary segment of society is also the youngest, prevention programmes aimed mainly at older people may be arriving long after the trajectory is set. The authors conclude that young adulthood is a critical and under-recognised target for interventions to promote physical activity, curb sedentary time and protect sleep for long-term health. In practical terms, that could mean redesigning university timetables to break up prolonged sitting, building standing desks and movement-friendly spaces into libraries and lecture halls, rewarding active commuting on campuses, and treating sleep as a pillar of student health rather than a casualty of academic pressure. The research was supported by centre funding from the National University of Singapore’s Centre for Sleep and Cognition, the Lee Foundation, and Singapore’s National Medical Research Council — an institutional wager that mapping daily movement across the life course is central to long-term health.

Caveats nonetheless apply. The data capture each cohort at a single moment, so the differences across age groups could partly reflect generational effects rather than ageing itself, and the cross-sectional design cannot exclude reverse causation — healthier people may simply find it easier to move more, sit less and sleep well. All three cohorts were Singaporean, and activity patterns are deeply shaped by local urban design, climate, work culture and transport, so the absolute numbers may not transplant neatly to other countries. The Oura Ring, though worn identically by every participant, is a consumer device whose algorithms differ from research-grade instruments, and volunteers for wearable studies are not necessarily representative of the wider population. The authors also disclosed potential competing interests: Chee serves on the medical advisory board of Ouraring Inc., and three co-authors belong to the Oura–National University of Singapore Joint Lab as of August 2026, although the study was designed, funded and executed independently by the university. None of this dissolves the central finding, but it argues against reading the associations as ready-made prescriptions.

Even with those qualifications, the study arrives at a moment when sedentary living among the young is increasingly described as a slow-motion public health emergency, and its most lasting contribution may be the assumption it dismantles. One tightly controlled, four-week comparison cannot settle what decades of campus sitting do to a young body, and the reported associations, while statistically grounded, describe patterns rather than proven cures. But the research gives universities, employers and policymakers a concrete target — 30 reallocated minutes a day — and a pointed reminder that cardiometabolic trajectories are not fixed at retirement but assembled in the lecture hall. For the millions of young adults spending their most physically capable years behind desks and screens, the message embedded in the data is blunt: the stage of life widely assumed to be the healthiest may, in the arithmetic of the 24-hour day, be the most sedentary of all, and the cheapest opportunity to change that is now, not decades downstream.

The concept at the heart of the research is “digital phenotyping”—the use of data generated by personal digital devices to characterize an individual’s behavior, physiology, and health over time. Unlike a traditional clinical test, which offers a snapshot, a wearable can collect repeated measurements across weeks, months, or even years. Accelerometers can register movement and rest; optical sensors can estimate heart rate; sleep algorithms can infer periods of inactivity; and connected platforms can organize these signals into detailed time series. The study focuses particularly on activity rhythms, meaning the regular cycles in movement and rest that reflect the interaction of the body’s internal clock, sleep-wake behavior, environment, and daily routines. These rhythms can reveal information that a simple daily step total may miss.

A person who walks 8,000 steps every day may appear stable when judged by a weekly average, yet the timing, intensity, and fragmentation of those steps could be changing. Movement might be increasingly concentrated in short bursts, shifted toward later hours, or interrupted by longer sedentary periods. Such changes may indicate alterations in sleep, circadian regulation, physical capacity, mood, or recovery. To capture these patterns, researchers can analyze wearable signals using methods from time-series analysis, signal processing, and computational biology. Measures such as rhythm strength, regularity, amplitude, timing, and day-to-day variability transform raw sensor readings into interpretable features. Together, they create a behavioral signature—one that may change gradually as the body moves through the aging process.

This is where biological aging enters the picture. Chronological age is simply the number of years a person has lived, but biological age refers to the condition and functional state of the body. Two people of the same chronological age can have sharply different levels of cardiovascular fitness, metabolic health, immune function, and physical resilience. Researchers have developed biological-age estimates using molecular markers, clinical measurements, and physiological data. Wearable-based approaches offer a different perspective by observing what people actually do in daily life. Activity rhythms may act as a real-world indicator of functional aging because they reflect mobility, energy, recovery, circadian stability, and the ability to maintain consistent routines outside controlled laboratory settings.

The longitudinal aspect of the study is especially important. A single day of wearable data can be distorted by illness, travel, unusual work hours, weather, or a missed device charge. Long-term monitoring makes it possible to distinguish temporary disruptions from persistent trends. Statistically, researchers can model an individual’s activity trajectory and examine how rhythm-related features evolve over time. They may also compare those trajectories with established indicators of aging, allowing them to investigate whether changes in movement patterns correspond to broader biological decline or resilience. This approach shifts the scientific question from “How active is this person today?” to “How is this person’s daily activity system changing, and what might that change reveal about future health?”

Commercial wearables make such research unusually scalable. Instead of requiring participants to visit a laboratory for repeated assessments, investigators can potentially study large populations using devices already worn by millions of people. That creates an unprecedented volume of passive health data, gathered in natural environments rather than under artificial experimental conditions. It also opens the possibility of identifying early warning signals: a gradual weakening of daily rhythms, increasing irregularity, or a sustained reduction in movement could prompt closer clinical evaluation. In the future, algorithms might help distinguish a short-term response to stress from a longer-term change associated with frailty, chronic disease, or accelerated aging.

Yet the apparent simplicity of wearable data conceals substantial technical and scientific challenges. Commercial devices do not measure every variable directly. Step counts depend on proprietary algorithms; sleep is inferred rather than observed; heart-rate readings can be affected by skin contact, motion, and device placement; and different brands may produce non-equivalent measurements. User behavior also shapes the data. People may remove devices during exercise, charge them at inconsistent times, or wear them less regularly when they feel unwell. These gaps can introduce bias, particularly if the people most at risk of declining health are also the least likely to generate continuous records. Any biological-aging model built from wearables must therefore account for missing data, device changes, demographic differences, and the limits of consumer-grade sensors.

The research also raises an important question about what activity rhythms actually represent. A less regular movement pattern might reflect biological aging, but it could also reflect shift work, caregiving responsibilities, disability, depression, socioeconomic conditions, or an unpredictable living environment. The same wearable signature may have multiple explanations. For that reason, digital phenotyping is most powerful when combined with contextual information and validated against clinical outcomes. Machine-learning models can detect complex patterns that conventional statistics might overlook, but their predictions still require careful interpretation. A model may identify who is at higher risk without explaining why, and a correlation between irregular rhythms and aging does not by itself prove that one causes the other.

If the approach proves robust, its impact could extend well beyond research laboratories. Clinicians might eventually use longitudinal wearable profiles to monitor rehabilitation, detect functional decline, personalize exercise recommendations, or evaluate whether an intervention is improving daily-life recovery. Public-health researchers could study how work schedules, urban design, pollution, and social conditions shape activity rhythms across entire populations. Individuals could receive feedback based not only on how much they move, but also on whether their patterns are becoming more stable, adaptable, and compatible with healthy sleep. Such systems would need to avoid turning normal variation into a diagnosis or encouraging constant self-surveillance. The goal would be early insight and better care—not a new source of anxiety.

Shim and Onnela’s study arrives as wearable technology is transforming the definition of a health measurement. Blood tests and imaging remain indispensable, but they capture only selected moments and biological compartments. Commercial sensors offer a complementary view: behavior in context, repeated continuously, and connected to the rhythms of ordinary life. By examining activity patterns over time, the research explores whether these signals can serve as a digital window into biological aging. The most viral aspect of the idea is also the most scientifically provocative: aging may leave detectable traces not only in cells and organs, but in the timing, regularity, and texture of everyday movement. The challenge now is to determine how accurately those traces can be read—and how responsibly they can be used.

Subject of Research: Longitudinal digital phenotyping of activity rhythms and biological aging using commercial wearable devices

Article Title: Longitudinal digital phenotyping of activity rhythms and biological aging using commercial wearables

Article References: Shim, J., Onnela, JP. “Longitudinal digital phenotyping of activity rhythms and biological aging using commercial wearables.” Nature Communications (2026). https://doi.org/10.1038/s41467-026-76147-6

Image Credits: AI Generated

DOI: 10.1038/s41467-026-76147-6

Keywords: Digital phenotyping, commercial wearables, activity rhythms, biological aging, longitudinal health data, wearable sensors, circadian patterns, machine learning, health monitoring