Counts describe this library, not the strength of evidence.
Cómo interpretar la evidencia
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.
Across 25 accelerometer studies, children and adolescents with heart, lung, diabetes or cancer conditions generally did less than the recommended 60 minutes a day of moderate-to-vigorous activity, and mostly matched healthy peers except those with malignancies (SMD -2.2, 95% CI -4.08 to -0.26). That malignancy result rests on only 2 studies (n = 90), and the other comparisons were not significant.
Sixty adults wore seven wrist-worn devices while sitting, walking, running and cycling in a laboratory. Median heart rate error against an electrocardiogram (ECG) ran from 1.8% on the bike to 5.5% walking, but no device achieved a calorie error below 20 percent (device medians from 27.4% to 92.6%). These were consumer models on sale at the time, tested in a lab, so the results do not rank today's watches or show how they perform in daily life.
In 62 people on a treadmill, heart rate error (MAPE) was 1.14% to 6.70% for the Apple Watch, 2.38% to 16.99% for the Fitbit Charge HR and 7.87% to 24.38% for the Garmin Forerunner 225, while calorie error was far larger on all three. These are older wrist devices tested in a short lab session, so they do not describe current models or daily life.
In a pilot of 40 inactive adults aged 55-79, a 12-week program of a Jawbone Up24 tracker plus weekly phone counseling was feasible and rated at least 4 out of 5 for acceptability, with small effects on steps (d=0.26). With only 40 people and no test of statistical significance reported, the effect estimates are rough.
An armband prototype combining ECG and pulse (PPG) sensors estimated systolic blood pressure with a mean error of 1.63 ± 4.44 mmHg on unseen data and heart rate with a mean absolute error of 0.21 beats per min. Data came from only ten volunteers, so this is a proof of concept and not evidence that it is accurate enough for real-world use.