Digital biomarkers for mobility and gait
Walking is frequent, structured and mechanically distinctive, which makes it the health domain wearable sensors read most reliably.
.png)
Mobility is the health domain where wearable sensors have made the most progress, because walking is frequent, structured and mechanically distinctive enough for an algorithm to detect reliably. A single accelerometer worn at the wrist, waist or lower back can record every step a person takes over weeks, which turns walking from something observed once in a clinic corridor into something described continuously in the places people actually live.
The measures collected here range from simple counts to structural descriptions of movement. Step count and moderate to vigorous physical activity summarise how much someone moves. Gait speed and walking bout structure describe how they move. Sedentary time and sit to stand transitions capture the parts of the day that involve no walking at all, which in impaired populations is most of it.
What makes this domain worth treating carefully is the gap between capacity and performance. A supervised walk test measures what a participant can do when asked. A wearable measures what they actually do, and the two frequently disagree. That gap is not measurement error. It is often the clinically interesting finding, and it is the reason mobility measures are usually reported alongside a clinic test rather than instead of one.
Mobility and gait measures in this library
How these measures are used
In clinical research these measures appear in three roles. As descriptive endpoints they characterise how a cohort actually moves at baseline. As response endpoints they test whether a treatment or rehabilitation programme changes daily function, which is often the question participants care about most. As monitoring endpoints they track deterioration or recovery over months without requiring anyone to attend a site.
Outside research, the same measures support remote care in conditions where declining mobility predicts poor outcomes. In practice most protocols pair a wearable measure with the clinic test it mirrors, such as gait speed with the Short Physical Performance Battery, or walking bouts with the six minute walk, so that capacity and real world performance can be interpreted together.
That pairing also guards against a common failure. Real world mobility moves for reasons unrelated to the condition under study, including weather, work schedules, holidays and changes in care setting. Without a supervised anchor, a seasonal drop in walking can be mistaken for disease progression.
What the evidence supports today
Mobility is the best evidenced domain in this library, though the strength varies sharply between measures. Step count and gait speed have decades of epidemiological support linking them to survival and functional decline, and both have established minimal important differences in several conditions. Bout level measures are newer, and their technical validation rests largely on the Mobilise-D consortium, which validated a shared processing pipeline against laboratory reference systems across several chronic conditions.
The strongest regulatory precedent in this library also comes from mobility. In 2023 the European Medicines Agency issued a qualification opinion on stride velocity 95th centile as a primary endpoint in ambulatory Duchenne muscular dystrophy, the first time a wearable derived measure reached primary endpoint status in Europe.
What remains weak is standardisation. Two studies can report the same measure from the same device and arrive at incomparable numbers because they applied different wear time rules, bout definitions or processing pipelines. Comparability across studies, not detection, is the open problem.
Common questions
How is a digital mobility measure different from a walk test?
A walk test measures capacity under supervision: what someone can do when asked, once, in a controlled setting. A digital mobility measure records performance: what they actually do across ordinary days. The two answer different questions and often diverge, which is why studies increasingly report both rather than treating one as a proxy for the other.
How many days of wear are needed before these measures are stable?
Accelerometry studies conventionally collect seven days and require a minimum number of valid days, commonly four, with a minimum daily wear time. Fewer days makes estimates sensitive to whether a weekend or an unusual day happened to be included. Measures of structure, such as bout distribution, generally need more days than simple totals.
Where on the body should the sensor be worn?
Lower back placement gives the best accuracy for gait and bout detection because it sits close to the body's centre of mass. The wrist is far more tolerable for long wear and is what most participants will accept, at the cost of noisier estimates. The right choice depends on whether the study needs precision on a few days or coverage across many weeks.
Can consumer wearables be used, or is a research grade device required?
Both are used. Research grade devices give access to raw acceleration data and documented algorithms, which matters when a measure will support a regulatory claim. Consumer devices give better adherence and lower cost, but their algorithms are proprietary and can change with a firmware update, which is a real threat to a longitudinal study. Many protocols now record device and firmware version alongside the data for this reason.
Run a study on these measures
WeGuide captures wearable data and patient reported outcomes in one platform, from screening through to analysis.
