A new population-based study has found that the way a person’s daily movement is patterned — not just how much they move — may carry meaningful information about their risk of dying. Using wrist-worn accelerometer data from thousands of American adults, researchers discovered that people whose active periods are more fragmented, switching frequently between activity and sedentary behavior, faced a higher risk of death from all causes during follow-up. The findings, published in BMC Public Health, add a nuanced dimension to the science of physical activity and mortality, while also sounding a note of caution about how far such measurements can be pushed.

The research team, led by Jie Li and Ling Yu of The Second Hospital of Jilin University in Changchun, China, analyzed data from 7,098 adults aged 20 years and older who participated in the National Health and Nutrition Examination Survey (NHANES) between 2011 and 2014. NHANES is a nationally representative survey program run by the US National Center for Health Statistics, and for these survey waves participants were asked to wear a wrist accelerometer for seven consecutive days. The researchers linked these movement records to National Death Index mortality data through 2019, creating a prospective cohort in which objective device measurements preceded the outcomes being studied.

At the heart of the analysis is a metric called the active-to-sedentary transition probability, or ASTP. Rather than measuring total activity volume — the familiar step counts or minutes of moderate-to-vigorous exercise — ASTP captures the probability that a minute of active movement is followed by a minute of sedentary time. A high ASTP means that active bouts tend to be short-lived and quickly interrupted by sitting or lying down; a low ASTP means that when someone is active, they tend to stay active for longer stretches. In other words, ASTP quantifies the continuity, or fragmentation, of everyday movement, a quality that wrist accelerometers are uniquely positioned to capture with minute-level precision.

During a median follow-up of 6.8 years, 522 deaths occurred in the cohort. After adjusting for a comprehensive set of clinical covariates — a sensitivity analysis that included 6,586 participants with complete data — each one-standard-deviation increase in ASTP was associated with a 16.5 percent higher hazard of death from all causes (hazard ratio 1.165, 95 percent confidence interval 1.042 to 1.304). The association was even stronger for deaths from causes other than cardiovascular disease and cancer, with a hazard ratio of 1.293 (95 percent CI 1.099 to 1.521). These estimates came from survey-weighted Cox proportional hazards models, the standard epidemiologic tool for estimating how a characteristic relates to the timing of an event such as death.

The cause-specific results were more complicated. For cardiovascular mortality, the estimates were statistically imprecise in both the cause-specific Cox models and the Fine–Gray subdistribution hazard models, which are designed to handle competing risks — the situation where one type of death precludes the possibility of another. For cancer mortality, the picture was mixed: the Fine–Gray model suggested an inverse subdistribution association, while the cause-specific estimate was imprecise. This divergence illustrates a technical but important point in survival analysis: when competing causes of death are common, different modeling frameworks can yield qualitatively different answers, and researchers must be careful not to overinterpret any single estimate.

The investigators went to considerable lengths to test the robustness of their findings. Sensitivity analyses addressed survey nonresponse by reweighting the sample, and the all-cause hazard ratio barely moved, from 1.165 to 1.168. A 24-month landmark analysis, which excluded deaths occurring in the first two years of follow-up to reduce the possibility of reverse causation — that is, pre-existing illness causing both fragmentation and early death — yielded a smaller and statistically nonsignificant estimate of 1.097 (95 percent CI 0.959 to 1.253). Most tellingly, when the researchers additionally adjusted for the total amount of movement, the association was attenuated or even reversed. This suggests that activity fragmentation is strongly correlated with how much people move overall, and that the apparent mortality signal may largely reflect movement volume rather than an independent effect of movement continuity.

The team also explored whether ASTP could improve mortality prediction beyond established clinical risk factors, using internal validation with 1,000 bootstrap refits and five-year calibration checks. The answer was technically yes but practically marginal: adding ASTP to a model changed the apparent C-index — a measure of predictive discrimination — by only 0.0034, with a bootstrap confidence interval of 0.0006 to 0.0067, and the optimism-corrected difference was 0.0032. The authors are explicit that this small prediction increment was internal and exploratory, with no established clinical utility. In plain terms, knowing how fragmented someone’s movement is does not yet meaningfully sharpen a doctor’s ability to predict their risk compared with knowing their age, blood pressure, metabolic markers and other standard variables.

These caveats matter because the study’s conclusions are deliberately restrained. The authors state that wrist-derived activity fragmentation, as measured by ASTP, is an epidemiologic marker of movement pattern that is correlated with movement amount, and that the observed associations with all-cause and other-cause mortality after clinical adjustment do not establish an independent causal effect of activity continuity. They also emphasize that ASTP is a distinct construct from accumulated brief moderate-to-vigorous physical activity, meaning that the fragmentation metric should not be conflated with the well-established benefits of exercise bouts. This distinction is important for a field that has increasingly turned to wearable devices to decompose physical activity into ever finer components.

Why does fragmentation matter at all? The concept has gained traction in gerontology and population health because the patterning of daily activity may reflect underlying physiological reserve, fatigue, frailty or subclinical disease. A person whose active periods dissolve quickly into sedentary time may be experiencing early signs of declining stamina or health that a single measure of total activity cannot fully capture. Wrist accelerometers, now ubiquitous in consumer fitness trackers, make it possible to measure this patterning objectively in free-living conditions, free from the recall bias that plagues questionnaire-based physical activity assessment. The NHANES accelerometry program has been a cornerstone of this research, providing nationally representative device data linked to long-term outcomes.

For the public, the takeaways are measured. The study does not claim that stitching together longer active bouts will independently extend life; the evidence for that specific causal claim remains unproven, and the association was sensitive to how much people moved overall. What it does establish is that movement fragmentation is a measurable, reproducible characteristic of daily behavior that tracks with mortality risk in a large, representative sample of US adults, and that it can be quantified with the same consumer-grade sensors millions of people already wear. As wearable technology continues to proliferate, metrics like ASTP may find their place not as standalone predictors but as descriptive markers that help researchers understand how the architecture of daily movement relates to long-term health — a reminder that in the science of physical activity, how we move may be as intriguing as how much.

Subject of Research: Association between accelerometer-measured activity fragmentation and cause-specific mortality in US adults

Article Title: Wrist-worn accelerometer-derived activity fragmentation and cause-specific mortality in US adults: a population-based prospective cohort study

Article References: Li, J., Sun, X., Zhang, J., Zhai, L., & Yu, L. (2026). Wrist-worn accelerometer-derived activity fragmentation and cause-specific mortality in US adults: a population-based prospective cohort study. BMC Public Health. https://doi.org/10.1186/s12889-026-29778-9

Image Credits: AI Generated

DOI: 10.1186/s12889-026-29778-9

Keywords: physical activity, accelerometry, sedentary behavior, activity fragmentation, mortality, NHANES, wearable technology, competing risks, population health, Cox model, risk factors, Wrist-worn