Counts describe this library, not the strength of evidence.
How to read this evidence
Validation is specific to a metric, model, reference and setting. A useful heart-rate measurement does not validate calorie estimates or sleep stages. Even processing the same accelerometer data differently can change estimated activity–disease relationships.
Over 24 hours in a single participant, the Apple Watch 3 differed from an ambulatory ECG by -1.80 bpm on average and the Fitbit Charge 2 by -3.47 bpm, with mean absolute error near 6% for both. With one person (the first author) and two devices, this is a preliminary check that cannot be generalized.
A systematic review of 35 studies found that data from phone and wearable sensors were associated with symptoms of schizophrenia, bipolar disorder and depression. Most studies covered bipolar disorder, depression and stress, with few on schizophrenia, and the authors say clinical usefulness for supporting therapy is not yet fully assessed.
In 10 healthy runners tested across speeds and slopes, ground reaction force metrics were not strongly correlated with tibial load; for impact peak versus peak tibial load, r = -0.29±0.37, and 76 of 80 individual correlations were not strong. With only 10 people, the sample is small, but it warns against treating impact force from a wearable as a measure of bone stress.
In a System Usability Scale (SUS) survey of 388 participants rating seven wearables, scores ranged from a mean of 61.4 (Apple Watch) to 67.6 (Huawei Honor B2), with no significant difference among brands. Scores were tied to how long people had used the device and their job background, but the devices were tested in 2018-era models and the sample's selection is not described.
In a systematic review of 46 studies in people with affective disorders, 85 behavioural features from phones and wearables were examined, and some, such as time at home, sleep duration and vigorous activity, correlated consistently with depressive mood. Others, such as texts sent and phone screen use, pointed in different directions across studies, and methods were not standardised.
In 201 college students followed for 1 month, wearable sensor features such as skin conductance and temperature classified high versus low self-reported stress with 78.3% (148/189) accuracy and mental health groups with 87% (41/47) accuracy. These are in-sample classification results from one university's students, and the mental health group is only 47 people.
Average daily acceleration was lower with cardio-metabolic disease in UK Biobank: 31 ± 12 mg in men with type 2 diabetes plus cardiovascular disease versus 42 ± 15 mg in healthy men. Disease was self-reported about 5 years before wrist monitoring, so the direction of cause is unclear.
In 31 older adults doing two two-minute walk tests, the Fitbit Flex and ChargeHR agreed well with directly observed steps (ICC 0.86, 95% CI 0.76-0.93), but both underestimated step counts. The sample was small and the home comparison used another device as the yardstick, so it supports consistency more than exact accuracy.
In 89 older adults (average age 73.7 years), the International Physical Activity Questionnaire (IPAQ) showed weak reliability and poor validity for sitting time and moderate-vigorous activity against a thigh-worn accelerometer, and classified behaviour correctly only 2% of the time. The sample was small, so the finding is a caution about questionnaires rather than a precise error estimate.
In 44 healthy people wearing 6 devices and 2 phone apps, heart rate, steps, distance and sleep duration had a mean absolute percentage error (MAPE) of about 0.10, but energy use reached up to 0.44. Results differed widely by brand and activity, and these are older models, so they do not carry over to today's watches.
After 6 weeks with a Fitbit Flex, 60 Australian adolescents aged 13-14 years mostly found it easy to use for tracking daily activity, with more difficulty for sleep tracking. This was a focus-group study at 3 schools in one city, so it describes opinions rather than whether the device changes activity.
Across 26 studies (14,739 participants, 3523 with obesity), children and adolescents with obesity consistently spent less than the recommended 60 min/day in moderate-to-vigorous activity (MVPA), and somewhat less than non-obese peers. Studies differed too much to pool, so the review is narrative, and sedentary time showed no marked difference between groups.
Among 157 patients with rheumatoid arthritis (83) or axial spondyloarthritis (74) tracked for 3 months, only 27.4% (43/157) reached the WHO target of 150 minutes per week of moderate to vigorous activity, and the two diseases had similar levels. Participants all owned smartphones and joined a research study, so they may be more active and tech-ready than typical patients.
In an online survey of 237 Australian adults (200 current and 37 former tracker users), 51-81% said their physical activity improved, compared with 14-40% for diet and 11-24% for sleep, and 70% had experienced functionality problems. Respondents were mostly young (mean age 33.1 years) and self-selected, and perceived change is not measured change.
Among 711 people in France given a Fitbit Zip, 73.9% were still tracking after 100 days and 16.0% after 320 days, with an average use of 129 days. Participants were handed a tracker as part of a study, so people who buy their own may behave differently.
Pooling 15 randomised trials (1402 people), computer, mobile and wearable tools reduced sitting by a mean of -41.28 minutes per day (95% CI -60.99 to -21.58) in healthy adults. Almost all trials had high or unclear risk of bias, heterogeneity was high, and the benefit had almost vanished beyond 6 months.
The international journal of behavioral nutrition and physical activity
In 15 healthy young adults, three Fitbit models correlated well with true step counts on a treadmill (ICC > 0.75) but poorly when walking with a walker (ICC < 0.60). The sample was small and young, so it says little about older people or the aid users the findings matter for.
Pooled accelerometer data from 9509 adults in England, Portugal, Norway and Sweden showed an average of 530 min/day of sedentary time and 36 min/day of moderate to vigorous activity, with 72% not meeting activity recommendations. The data come from five cross-sectional studies, so they describe a snapshot and cannot show who changes over time.