Step rhythm on a wristband nearly matched a walking test on death risk

One measure takes seconds in a clinic; the other samples walking across days of ordinary life. On 23 October 2026, researchers from Norway, France and the UK published in JAMA Network Open a comparison of the two in 4,067 participants of the British Whitehall II study, former civil servants with a mean age of 69. In 2012–2013 they did a short walking test at a clinic visit and wore a wrist accelerometer; over a median follow-up of 11.8 years, 739 of them (18.2%) died, and the team asked which measurement tracked that risk better.

A stopwatch and a wristband see different walking

The clinic test was the classic one: walk 2.44 metres at your usual pace, three times, and take the average. The group's mean was 1.2 metres per second.

Participants wore a research-grade GENEActiv accelerometer on the non-dominant wrist for nine consecutive days; on average 7.9 days of recording were valid. The OxWearables stepcount software detected walking in 10-second windows and calculated cadence, the number of steps per minute. The authors used two summaries: the median cadence across all detected walking (75.3 steps per minute on average) and the 95th percentile, which reflects the faster end of a person's walking (112.2). Participants averaged 9,524 steps a day.

The two kinds of measurement barely agreed with each other: the correlation between clinic speed and cadence was only 0.27–0.28. A short corridor walk under observation and more than a week of free-living recording capture different things, and the authors state plainly that the measures should not be treated as interchangeable.

Both pointed the same way, beyond step count

After adjustment for age, sex, education, ethnicity, marital status, smoking, alcohol, fruit and vegetable intake, body mass index, hypertension, raised cholesterol, chronic diseases and limitations in daily activities, all three measures were associated with lower all-cause mortality. Per one standard deviation, the hazard ratio was 0.77 (95% CI 0.71–0.84) for clinic speed, 0.83 (0.77–0.90) for median cadence and 0.82 (0.75–0.89) for the 95th percentile.

In more familiar units: every 0.1 m/s of faster clinic walking went with a roughly 10% lower risk of death over the follow-up (HR 0.90; 0.87–0.93), and every extra 10 steps per minute of everyday cadence with roughly 17–19% lower risk (HR 0.83 and 0.81). The associations remained after the models accounted for daily step count, which supports, but cannot prove, that cadence carries information beyond the sheer amount of walking. The links with mortality did not differ significantly between women and men. The clinic-speed figure is consistent with a 2011 pooled analysis of nine cohorts with 34,485 adults aged 65 and over, which found a hazard ratio of 0.88 per 0.1 m/s.

Where the old test still holds an edge

The harder question was whether any of these numbers improves the forecast that can already be made from known risk factors. The authors compared models with Harrell's C index, a measure of how well a model ranks people by risk: 0.5 is no better than chance, 1.0 is perfect.

On its own, each walking measure scored modestly, 0.61–0.63. A model built only from the usual risk factors scored 0.742. Adding clinic walking speed raised it to 0.747, and that was the only statistically significant improvement. Wrist cadence reached 0.744 and 0.745, a gain the study could not distinguish from zero. Combining the stopwatch and the wristband did not add significantly either once health factors were in the model.

In the fully adjusted model, then, the wristband and the stopwatch had similar C indices, but only the stopwatch measurably sharpened a forecast already built on age, illnesses and habits. This is a comparison of population models, not a way to make a personal prognosis from one walking number.

What the study does not show

This is an observational study: nobody was asked to walk faster, so it shows an association, not an effect. Slowing down may itself be an early sign of illnesses that later shorten life, and adjustment cannot fully remove that. The sample was 74% men and 92.5% White, and healthier than the full 2012–2013 study wave. There was no separate cohort to validate the models, and there are still no standardised methods for deriving cadence from accelerometers. The study used a research-grade device and specific processing software, so it does not validate how consumer watches report steps or cadence. The work was funded by public and charitable bodies, including Horizon Europe, the US National Institute on Aging, the UK Medical Research Council, Wellcome and the European Research Council; according to the authors, the funders had no role in the analysis.

What to take from it

Lonevi has already covered what gait speed predicts and what counts as a meaningful change. The new data add that cadence recorded on the wrist carried a mortality association similar to clinic speed in this older cohort, but it did not replace the short test. A personal reading is best discussed with your doctor alongside symptoms, illnesses and other measurements, rather than treated as a forecast.

This material is for information only and does not replace a consultation with a doctor.

Sources

• Haddadj R, Danilevicz IM, Rowlands AV, et al. Clinical Walking Speed, Free-Living Walking Cadence, and All-Cause Mortality. JAMA Netw Open. 2026;9(10):e2640122. PMID 42912394 · doi:10.1001/jamanetworkopen.2026.40122

• Studenski S, Perera S, Patel K, et al. Gait Speed and Survival in Older Adults. JAMA. 2011;305(1):50–58. PMID 21205966