Can Your Phone Predict a Fall? Walking Steadiness, Gait Speed and What iPhone Actually Measures

Short answer: an iPhone can estimate fall risk from how someone walks, using gait measurements that are reasonably accurate, but Apple has not published how well its Walking Steadiness rating actually predicts falls.

That gap matters because the stakes are high. More than one in four adults over 65 falls each year [1], and the speed at which older adults walk is one of the most powerful predictors of how long they will live [2]. A phone in a pocket now measures gait continuously, at no cost, for millions of people. For Medicare Advantage plans, home care providers, physiotherapy programs and anyone serving older adults, that is a new signal. This post covers what the phone measures, how accurate it is, what is and is not known about fall prediction, and how to use the data responsibly.

Why do gait and falls matter?

Falls are common and costly. According to the CDC, more than one in four adults aged 65 and over falls each year. Falls lead to about 4.5 million emergency department visits and 1.4 million hospitalizations a year in this age group, and about 319,000 hospitalizations for hip fractures, the large majority caused by falls. About 37% of people who fall are injured badly enough to need care or restrict activity, and falling once doubles the chance of falling again [1].

Gait speed is a vital sign in all but name. In a pooled analysis of nine cohorts with 34,485 adults aged 65 and over, followed for 6 to 21 years, each 0.1 meter per second faster gait was associated with a 12% lower risk of death. At age 75, predicted 10-year survival ranged from 19% to 87% in men and 35% to 91% in women across the range of gait speeds. Age, sex and gait speed predicted survival as well as models using chronic conditions, hospital use and blood pressure [2].

So a measure that tracks walking speed continuously, without a clinic visit, is potentially one of the most valuable signals a phone can produce for older adults. The population is covered in more detail in wearables and adults over 65.

Fall detection or fall prediction?

These are different jobs, and products often blur them.

• Fall detection notices a fall after it happens, from a sudden impact followed by little movement, and can alert contacts or emergency services. Apple Watch, Pixel Watch and Galaxy Watch all offer it. It reduces the harm of a fall, especially for people who live alone.
• Fall prediction estimates the risk of a fall before it happens, from how someone walks. It is what makes prevention possible: exercise, physiotherapy, medication review, home changes.

Walking Steadiness on iPhone is a prediction feature. That is why it is more interesting to care programs, and why the evidence behind it deserves a careful look.

What does iPhone measure, and how accurately?

Since iOS 14, iPhone 8 and later have estimated a set of mobility metrics from the motion sensors whenever the phone is carried near the waist during steady walking on flat ground [3]:

• Walking speed, from a model of the body’s center of mass.
• Step length, from height, cadence and speed.
• Double support time, the share of each step with both feet on the ground. Typical walking falls between 20% and 40%; lower generally means better balance [3].
• Walking asymmetry, the share of steps where one leg moves at a different speed from the other.

Apple validated these against an instrumented pressure mat in a study of more than 500 people, split into a design set and a validation set of 179 [3].

Walking speed and step length agree well with the lab. Double support time agrees only moderately. Accuracy depends on conditions Apple states plainly: the phone in a pocket or on a belt rather than in a bag, flat ground, and the user’s current height entered in the Health app [3].

There is also a sensitivity question. Apple reports the smallest change each metric can reliably detect in one person. For walking speed, the median was about 0.14 meters per second in the validation set [3]. That matters because a clinically meaningful decline, around 0.1 meters per second in a year [3], is at the edge of what a single comparison can detect. Trends over weeks and months, built from many walks, are more reliable than any two readings.

What is Walking Steadiness, and what is known about it?

Walking Steadiness, added in iOS 15, combines the mobility metrics into a rating of OK, Low or Very Low, which Apple relates to the risk of falling over the next 12 months. Users can turn on notifications if their steadiness drops [4].

Apple says the feature was built with data from the Apple Heart and Movement Study, a research study with Brigham and Women’s Hospital and the American Heart Association that has enrolled more than 100,000 participants across ages, who completed hundreds of thousands of surveys including on falls [5][6].

What we could not find is a peer-reviewed publication of how well Walking Steadiness predicts falls: how many people rated Low went on to fall, and how many falls happened to people rated OK. Apple’s published validation covers the gait measurements, not the fall prediction built on top of them [3]. The feature has been used as an outcome in independent clinical research, for example comparing recovery of steadiness after hip and knee replacement [7], which is a sign of clinical interest rather than validation of its predictive accuracy.

Why is this an iPhone-only signal?

Because Android has no equivalent platform feature.

On iPhone, HealthKit exposes walking speed, step length, double support time, walking asymmetry, six-minute walk distance and stair speeds from iOS 14, and Walking Steadiness and steadiness events from iOS 15 [8]. An app can request permission to read them.

On Android, Health Connect has a general speed record but no data types for step length, double support, asymmetry or steadiness [9]. Samsung and Google offer fall detection on their watches, but neither platform publishes a passive gait-quality rating for apps to read.

For a program serving older adults, that is a real equity issue. Gait-based fall risk is available for iPhone users and largely unavailable for everyone else, which is one reason phone and wearable ownership gaps need to be part of program design.

Does acting on fall risk help?

Finding people at risk only matters if something reduces the risk, and here the evidence is solid. A 2019 Cochrane review of community-dwelling older adults found that exercise programs reduced the rate of falls by about 23%, with balance and functional exercises showing the clearest effect [10].

That suggests the most valuable use of passive gait data is not the rating itself but the loop around it: identify people whose walking is slowing or becoming less steady, offer a balance and strength program, and use the same metrics to see whether walking improves.

How should care programs use gait data?

• Use trends, not single values. Look for sustained declines in walking speed or steadiness over weeks, not a single low day. The metrics are noisy within a person and depend on how the phone is carried [3].
• Confirm with a clinical assessment. A timed walking or standing test in a clinic or a home visit turns a phone signal into a care decision.
• Pair detection with an intervention. The value is in offering exercise, physiotherapy or a medication review to people whose gait is declining [10].
• Plan for missing data. People who walk less, or who carry their phone in a bag, produce fewer measurements. Silence can mean low activity, not good balance.
• Respect the line between wellness and diagnosis. Reporting a person’s walking speed trend is different from telling them they are at high risk of falling. Keep the wording factual and point to an assessment.
• Account for the Android gap. Programs that rely on Walking Steadiness will only see iPhone users. Phone step counts and activity, available on both platforms, are a weaker but broader signal; see step counting accuracy, phone versus wrist.

The short version

Falls affect more than one in four adults over 65 every year, and gait speed predicts survival as well as a medical history does [1][2]. iPhone measures walking speed and step length well against a gait lab, double support time less well, and combines them into a Walking Steadiness rating of OK, Low or Very Low [3][4]. Apple built that rating on a 100,000-person study but has not published how well it predicts falls [5][6]. Android has no equivalent [9]. Exercise reduces falls by about a quarter [10], so the best use of the data is to find people whose walking is declining, confirm with an assessment, and track whether a program helps.