The science behind the measurements

The time from one heel strike to the next closely tracked an instrumented treadmill at walking speeds from 3.5 to 6.5 km/h. Observed step times ranged from about 0.42 to 0.70 s.

Explore the measurement: The Pearson correlation between phone and treadmill was 0.946. Step time is the event-level timing measure in this study.

Stride duration, step duration, stance duration, swing duration and cadence each agreed closely with a pressure mat. Together they describe the rhythm and phases of walking.

Explore the measurement: ICCs were:

• stride duration: 0.86
• step duration: 0.88
• stance duration: 0.83
• swing duration: 0.86
• cadence: 0.80

Participants repeated the walking protocol a week later. The phone reproduced stride duration, step duration, stance duration, swing duration and cadence across the two visits.

Explore the measurement: In healthy adults, test–retest ICCs were:

• stride duration: 0.88
• step duration: 0.78
• stance duration: 0.93
• swing duration: 0.86
• cadence: 0.90

In people with obesity, depression or chronic low back pain, test–retest ICCs for stride, stance, swing and cadence were 0.75–0.88 for the phone and 0.84–0.89 for the watch.

Step length estimated from motion sensors was compared step by step with an instrumented treadmill. Combining phone and watch data gave the smallest mean absolute error.

Explore the measurement: Per-step ICCs were 0.89 for the phone, 0.86 for the watch and 0.91 combined. When each person's steps were averaged, MAE fell to 1.72, 1.77 and 1.52 cm, with ICCs of 0.92, 0.86 and 0.94.

Phone and watch measurements of stride duration, stance duration, swing duration and cadence were compared with a pressure mat across two visits.

Explore the measurement: Phone/watch ICCs were:

• stride duration: 0.83–0.90/0.94
• stance duration: 0.78–0.84/0.80–0.91
• swing duration: 0.54–0.57/0.54–0.59
• cadence: 0.83–0.84/0.96–0.98

A heel strike starts a stride, and a toe off divides it into stance and swing. In paired recordings, the phone and optical system gave nearly identical mean stride durations.

Explore the measurement: The timing comparison also covered stance duration (0.688/0.696 s), swing duration (0.387/0.379 s) and cadence (56.3/56.3 strides/min), optical/phone, left side. Stride length and speed from the same study appear in card 2.

The same optical comparison tested spatial gait measures. Left-side mean stride length was 0.862 m optically and 0.868 m by phone; mean stride speed was 0.802 and 0.807 m/s. The phone estimated length using stride frequency and acceleration variance, then calculated speed from length and time.

Explore the measurement: Across sides, the reported mean bias was 0.7 cm for stride length and 0.004 m/s for speed. For the left side, 95% limits of agreement were −7.5 to +6.3 cm for length and −0.091 to +0.080 m/s for speed.

The same gait method was compared with optical motion capture in laboratory recordings. It measured stride time, cadence, stance, swing, stride length and stride speed.

Explore the measurement: Stride-time bias was 0.000 s, with limits of agreement from −0.100 to +0.061 s. Bias was −0.005 m for stride length and −0.005 m/s for speed. Relative limits of agreement ranged from 7.5% for cadence to 17.4% for swing.

Over three walks on a 250 m outdoor route, six stride measures were repeatable wherever the phone was carried: at the belt, in a trouser pocket, in the hand, in a shirt pocket or in a shoulder bag. The measures were stride time, cadence, stance time, swing time, stride length and stride speed.

Explore the measurement: ICC ranges across carrying positions were:

• stride time: 0.91–0.97
• cadence: 0.90–0.96
• stance: 0.91–0.97
• swing: 0.85–0.96
• stride length: 0.89–0.98
• stride speed: 0.90–0.98

During the same outdoor walks, a phone at the belt, in the hand, in a shirt pocket or in a shoulder bag was compared with one in a trouser pocket. Agreement was calculated separately for each stride measure.

Explore the measurement: Limits of agreement ranged from:

• stride length: 4.4–7.3%
• stride time: 6.5–9.7%
• cadence: 6.7–9.8%
• stride speed: 7.0–11.9%
• stance: 8.5–13.5%
• swing: 11.8–17.5%

Repeated outdoor walks let the researchers estimate the smallest detectable change for each stride measure. For stride length, that change was 1.0–3.4 cm across carrying positions.

Explore the measurement: MDC ranges were:

• stride time: 0.041–0.075 s
• cadence: 1.9–4.7 strides/min
• stance: 0.023–0.047 s
• swing: 0.015–0.047 s
• stride speed: 0.026–0.067 m/s

The phone added up individual step lengths within a six-minute window, with the window defined by step times. Across 60 comparisons covering roughly 180–340 m, estimated distance tracked treadmill distance.

Explore the measurement: Pearson r was 0.967. The report evaluates summed distance during treadmill walking.

Phone motion and demographic features estimated the metabolic intensity of treadmill walking, compared with a portable metabolic system. Mean bias was 0.01 MET.

Explore the measurement: Pearson r was 0.896. The dataset came from 38 healthy adults.

Mean walking time was 14.9 min/day at month one, 18.2 at month three and 9.3 at month seven, during COVID-19 confinement. The longitudinal record also captured GPS travel distance, stride length and stance duration during detected walks. Weekly walking, balance and tapping tests sat alongside daily survey and medication records.

Explore the measurement: Seventeen of 20 participants completed seven months, and 99.9% of expected data streamed correctly.

A score combining everyday stride length and stance duration was associated with Timed Up and Go time measured in the clinic. The gait record was averaged over the two weeks before each visit.

Explore the measurement: The study built the composite score from the two passive gait features and compared it with the clinic time.

Stride length combined with balance-test frequency tracked clinical gait and balance scores. Tap count and tap distance tracked a clinical tapping score. Daily survey scores tracked patient and clinician impressions.

Explore the measurement: Correlations were:

• gait and balance: r = 0.64 and 0.63
• tapping: r = 0.51
• survey: r = −0.61 and −0.56

Walking time changed sharply when daily life changed. Among 16 monitored people, mean walking minutes fell during confinement. The three who left the city increased their walking time by 99%.

Explore the measurement: Medication adherence logs remained stable, and clinical disability scores changed little over the period.