From: Systematic review of the validity and reliability of consumer-wearable activity trackers
 | Sample characteristics | Tracker wearing protocol | Measurements | Validity results | |||||
---|---|---|---|---|---|---|---|---|---|
Author (year) | n | % female | Activity | Lab/field | Validity criterion (measure assessed) | Type | Placement | Measures | |
Adam Noah et. al (2013) [29] | 16 | 38 | 6 min each of treadmill walking (3.5 mph), walking with incline (3.5 mph at 5 %), jogging (5.5 mph), and stair stepping (30.5 centimeter step at 96 beats/min) | Lab | Two Actical accelerometers (steps), indirect calorimetry using K4b2 Cosmed (EE) | Ultra (Fitbit) | Waist (one on each side) | Steps/min, kilocalories/min | Fitbit Ultra vs. Actical ICC: average 0.94, range 0.80–0.99 (steps); Fitbit Ultra vs. Cosmed ICC: average 0.77, range 0.58-0.87 (kilocalories) |
23 | 43 | Classic (Fitbit) | Waist (one on each side) | Steps/min, kilocalories/min | Fitbit vs. Actical ICC: average 0.93, range 0.82–0.98 (steps); Fitbit vs. Cosmed ICC: average 0.74, range 0.18-0.72 (kilocalories) | ||||
Bai et. al (2015) [45] | 52 | 46 | 20Â min sedentary, 25Â min treadmill at self-selected speed, 25Â min resistance exercise | Lab | Indirect calorimetry using Oxycon Mobile (EE) | Flex (Fitbit) | Left wrist | Kilocalories/80- min trial | Overestimated overall EE by 20.4 kilocalories; Pearson CC 0.78; overall mean absolute error 16.8Â % |
UP24 (Jawbone) | Right wrist | Underestimated overall EE by 23.1 kilocalories; Pearson CC 0.77; overall mean absolute error 18.2Â % | |||||||
Case et. al (2015) [30] | 14 | 71 | Treadmill walking at 3.0 mph for 500 and 1500 steps, each done twice | Lab | Tally counter (steps) | One (Fitbit) | Waist | Steps/trial | 500 step trial (n = 27 observations) mean 498.6 (SD 3.7); 1500 step trial (n = 26 observations) mean 1497.0 (SD 10.7) |
Zip (Fitbit) | Waist | Steps/trial | 500 step trial (n = 27 observations) mean 498.6 (SD 10.8); 1500 step trial (n = 27 observations) mean 1498.4 (SD 10.4) | ||||||
Flex (Fitbit) | Wrist | Steps/trial | 500 step trial (n = 28 observations) mean 465.4 (SD 92.1); 1500 step trial (n = 28 observations) mean 1378.0 (SD 142.7) | ||||||
UP24 (Jawbone) | Wrist | Steps/trial | 500 step trial (n = 28 observations) mean 477.5 (SD 102.1); 1500 step trial (n = 28 observations) mean 1477.0 (SD 174.4) | ||||||
Dannecker et. al (2013) [31] | 19 (16 with Fitbit data) | 47 (from n = 19) | Resting, supine, sitting, standing, free living activity, and 6 random activities out of 8 (walking (2.5 mph, 3.5 mph, or 2.5 mph with 2.5 % grade), stepping, sweeping, cycling (75 watts), standing, sitting | Lab | 4 h stay in whole room calorimeter (EE) | Classic (Fitbit) | Belt at anterior superior iliac spine | Total EE during the 3.5-h period while in the room calorimeter (omitted first 30 minutes) | Root-mean-square error of tracker 28.7 % or 143 kilocalories; root-mean-square error of tracker after labeling activities 12.9 % or 64 kilocalories |
de Zambotti et. al (2015a) [47] | 28 | 100 | One nights sleep (n = 10), 2 nights sleep (n = 18) | Lab | Polysomnography (sleep) | UP (Jawbone) | Non dominant wrist | TST, sleep onset latency, WASO | Overestimated TST by 26.6 ± 35.3 min (p < 0.001) and sleep onset latency by 5.2 ± 9.6 min (p = 0.005); underestimated WASO by 31.2 ± 32.3 min (p < 0.001) |
de Zambotti et. al (2015b) [48] | 65 | 43 | One nights sleep | Lab | Polysomnography (sleep) | UP (Jawbone) | Non dominant wrist | TST, sleep efficiency, sleep onset latency, WASO | Overestimated TST by 10.0 min (p < 0.001), sleep efficiency by 1.9 % (p < 0.001), and sleep onset latency by 1.3 min (p = 0.33); underestimated WASO by 10.6 min (p < .001) |
Diaz et. al (2015) [15] | 23 | 57 | 6 min each of treadmill walking (1.9 mph, 3.0 mph, 4.0 mph) and jogging (5.2 mph) | Lab | Counting from a video recording (steps), indirect calorimetry using Ultima CPX (EE) | One (Fitbit) | 2 on right hip, 1 on left hip | Steps/min, kilocalories/min | Pearson CC 0.97–0.99 and mean difference −3.1 to −0.3 (steps); Pearson CC 0.86-0.87 (kilocalories) and mean difference −0.8 to 0.4 kilocalories |
Flex (Fitbit) | 1 on each wrist | Steps/min, kilocalories/min | Pearson CC 0.77-0.85 and mean difference −26.3 to −2.9 (steps); Pearson CC 0.88 and mean difference −0.2 to 2.6 (kilocalories) | ||||||
One (Fitbit) | Right hip | Steps/day, MVPA min/day, kilocalories/day, sleep min/day | Pearson CC 0.99 (steps), 0.91 (MVPA), 0.76 (kilocalories), 0.92 (sleep); ICC 0.95 (steps), 0.46 (MVPA), 0.55 (kilocalories), 0.90 (sleep); mean absolute difference 779 (steps), 58.6 (MVPA), 349 (kilocalories), 23.0 (sleep); range of differences = −890 to 1849 (steps), 1.0 to 137.2 (MVPA), −1724 to −83 (kilocalories), 45 to 76 (sleep) | ||||||
Zip (Fitbit) | Right hip | Steps/day, MVPA min/day, kilocalories/day | Pearson CC 0.99 (steps), 0.88 (MVPA), 0.81 (kilocalories); ICC 0.98 (steps), 0.36 (MVPA), 0.57 (kilocalories); mean absolute difference 447 (steps), 89.8 (MVPA), 484 (kilocalories); range of differences −970 to 1596 (steps), 10.0 to 157.2 (MVPA), −1145 to 218 (kilocalories) | ||||||
Ferguson et. al (2015) [33] | 21 | 52 | 48 h (including sleep, excluding showering) of free-living conditions, no activity restrictions/guidelines | Field | BodyMedia SenseWear model MF (steps, physical activity, EE, sleep); ActiGraph GT3X+ (steps, physical activity) | UP (Jawbone) | Left wrist | Steps/day, MVPA min/day, kilocalories/day, sleep min/day | Pearson CC 0.97 (steps), 0.81 (MVPA), 0.74 (kilocalories), 0.89 (sleep); ICC 0.97 (steps), 0.70 (MVPA), 0.27 (kilocalories), 0.85 (sleep); mean absolute difference 806 (steps), 18.0 (MVPA), 866 (kilocalories), 22.0 (sleep); range of differences −1978 to 2252 (steps), −4.7 to 96.5 (MVPA), −1937 to −94 (kilocalories), − 31 to 132 (sleep) |
Gusmer et. al (2014) [34] | 32 | 78 | 30-min phases of treadmill walking at slow and brisk speeds (±10 % of selfselected comfortable walking speed) | Lab | ActiGraph G1TM (steps), CPX Ultima metabolic cart (EE) | Ultra (Fitbit) | Right hip | Steps/min, kilocalories/trial | Pearson CC: slow walk: 0.97 (steps: mean 105.3 ActiGraph vs. 105.9 Ultra), 0.69 (kilocalories: mean 100.9 cart vs. 88.0 Ultra); brisk walking: 0.996 (steps: mean 114.2 ActiGraph vs. 113.9 Ultra), 0.94 (kilocalories: mean 121.9 cart vs. 100.9 Ultra) |
Lauritzen et. al (2013) [23] | 6 | 0 | 20-meter walk at participant's normal pace | Lab | Counting from a video recording (steps) | Ultra (Fitbit) | 1 on belt/pants pocket on dominant leg, 1 on wrist of dominant hand | Steps/20-min trial | Hip error 2.9Â % (SD 2.3Â %); wrist error 31.3Â % (SD 30.7Â %) |
One (Fitbit) | Waist | Kilocalories/trial | Mean absolute error 10.4 %; Pearson CC 0.81; root-mean-square error 40.1; did not fall in 90 % equivalence interval; systematic bias with slope −0.22 comparing One (x) to Oxycon (y); Pearson CC to ActiGraph 0.80 | ||||||
Lee et. al (2014) [35] | 60 | 50 | 13 activities that were all 5Â min in length except for treadmill (3Â min each) totalling 69Â minutes | Lab | Oxycon Mobile (EE); ActiGraph GT3X+ worn on hip, applied Sasaki et al. 2011 [39] algorithm (EE) | Zip (Fitbit) | Waist | Kilocalories/trial | Mean absolute error 10.1Â %; Pearson CC 0.81; root-mean-square error 40.8; fell within 90Â % equivalence interval from measured EE; systematic bias with slope - 0.29 comparing Zip (x) to Oxycon (y); Pearson CC to ActiGraph 0.77 |
UP (Jawbone) | Left wrist | Kilocalories/trial | Mean absolute error 12.2Â %; Pearson CC 0.74; root-mean-square error 45.8; did not fall in 90Â % equivalence interval; no systematic direction of bias with slope - 0.03 comparing UP (x) to Oxycon (y); Pearson CC to ActiGraph 0.65 | ||||||
Mammen et. al (2012) [36] | 10 | 50 | One min on the treadmill at each of 8 speeds (4 walking and 4 running) | Lab | Manually count (steps) | Ultra (Fitbit) | Waist, inside the pants pocket, shirt collar (men) or bra (women) | Steps/trial | Waist-worn Ultra under counted at 2 km/hour (31 steps/min; p < 0.05) but had similar counts at > =3 km/hour. Pocket- worn Ultra under counted during running (10, 19, 34, 38 steps/min at 8, 9, 10, and 11 km/hour, respectively; p < 0.05), but recorded similar counts when walking (2, 3, 4.5, and 6 km/hour). Similar counts across walk/run trials for collar-(males) or bra-(females) worn Ultras. |
Meltzer et. al (2015) [37] | 63 | 51 | One night's sleep | Lab | Polysomnography (sleep) | Ultra (Fitbit) | Non dominant wrist | TST, sleep efficiency, WASO | Normal mode overestimated TST by 41Â minutes and sleep efficiency by 8Â %, underestimated WASO by 32Â minutes; 87Â % sensitivity, 52Â % specificity, 84Â % accuracy. Sensitive mode underestimated TST by 105Â minutes and sleep efficiency by 21Â % and overestimated WASO by 106Â minutes; 70Â % sensitivity, 79Â % specificity, 71Â % accuracy. |
Montgomery- Downs et. al (2012) [38] | 24 | 40 | One night's sleep | Lab | Polysomnography (sleep) | Classic (Fitbit) | Non dominant wrist | TST, sleep efficiency | Polysomnography recorded 465.0Â min (SD 48.4) with 79.5Â % sleep efficiency and 370.9Â min (SD 70.3) TST; Fitbit measured 94.0Â % sleep efficiency and 438.0Â min TST; Fitbit overestimated sleep efficiency compared to polysomnography by 14.5Â % (SD 10.7Â %) and overestimated TST by mean 67.1Â min (SD 51.3). |
Sasaki et. al (2015) [39] | 20 | 50 | Visit 1: 6 min each of treadmill walking (3.0 at 5 % and 4.0 at 5 %) and jogging (5.5 mph), three trials; visit 2: 6 min each of household activities (choice from 3 activity routines) | Lab | Oxycon Mobile (EE) | Classic (Fitbit) | Belt around waist in line with the anterior axillary line | Total EE (rest plus activity) | Pearson CC 0.86; systematic underestimation of EE by the Fitbit with a mean bias of −4.5 ± 1.0 kcals/6 min; for 6 of 15 activities the Fitbit significantly underestimated EE (stairs, cycling, laundry, raking, treadmill 3.0 mph with 5 % grade, treadmill 4.0 mph with 5 % grade) and 1 of 15 activities the Fitbit significantly overestimated EE (carrying groceries) |
Simpson et. al (2015) [46] | 42 | 74 | 8 trials of walking 15 meters (self selected speed and 0.3-0.9 m/s at 0.1 increments) | Lab | Counting from a video recording (steps) | One (Fitbit) | Right waist, right ankle | Steps/trial | % error: 0.3 m/s: ankle 14.5, waist 98.4; 0.4 m/s: ankle 5.9, waist 82.0; 0.5 m/s: ankle 4.1, waist 40.4; 0.6 m/s: ankle 3.2, waist 21.6; 0.7 m/s: ankle 2.5, waist 10.5; 0.8 m/s: ankle 2.8, waist 7.0; 0.9 m/s: ankle 2.8, waist 5.6; Bland Altman mean difference −0.4 to 5.7 steps for ankle and 1.4 to 48.0 for waist |
Stackpool et. al (2014) [40] | 20 | 50 | 20Â min each of: treadmill walking, treadmill running, elliptical cross-training, agility-related exercises | Lab | Manually counting (steps); Oxycon Mobile (EE) | Ultra (Fitbit) | Hip | Steps and kilocalories for each 20-min bout | Pearson CC: treadmill walking (0.99 steps, 0.24 kilocalories), treadmill running (0.44 steps, 0.63 kilocalories), elliptical (0.99 steps, 0.47 kilocalories), agility (0.47 steps, 0.67 kilocalories) |
UP (Jawbone) | Wrist | Steps and kilocalories for each 20-min bout | Pearson CC: treadmill walking (0.98 steps, 0.87 kilocalories), treadmill running (0.98 steps, 0.69 kilocalories), elliptical (0.99 steps, 0.40 kilocalories), agility (0.34 steps, 0.65 kilocalories) | ||||||
Stahl and Insana (2014) [41] | 10 | 30 | During waking hours for 10 consecutive days | Field | Self-reported estimation of expended kilocalories/week from CHAMPS questionnaire (EE). Note: kilocalories/week divided by 7 to obtain kilocalories/day; then basal metabolic rate was added to the kilocalories/day. | Classic (Fitbit) | Waist | Kilocalories/day | Pearson CC 0.61; Fitbit underestimated by a mean of 195.0 kilocalories/day; 70Â % of participant's data were within 1 SD and 100Â % were within 2 SD |
Storm et. al (2015) [42] | 16 | 38 | 11-min walking protocol (included indoor and outdoor walking and steps) repeated at self-selected natural, slow, and fast speeds | Lab | OPAL sensors placed on each ankle (steps) | One (Fitbit) | Left waist | Steps/11-min trial | 1.1 % self-selected walk, 1.0 % fast walk; limits of agreement 15 ± 35 steps; under estimated for slow walk (−25 mean steps), self-selected walk (−12 mean steps), fast walk (−9 mean steps) |
UP (Jawbone) | Right wrist | Steps/11-min trial | Mean absolute error 10.1 % slow walk, 2.5 % self-selected walk, 2.1 % fast walk; limits of agreement 16 ± 135; under estimated for slow walk (−35 mean steps), self-selected walk (−4 mean steps), fast walk (−9 mean steps) | ||||||
Takacs et. al (2014) [43] | 30 | 50 | 5 min each of treadmill walking (0.90, 1.12, 1.33, 1.54, 1.78 meters/second) | Lab | Motion capture system and manually counting (steps); treadmill output (distance) | One (Fitbit) | 1 right hip, 1 left hip, 1 in front pocket of the dominant leg | Steps/trial, distance/trial | Steps: no significant difference (p > 0.05) between observed and One step counts at any of the 3 locations, ICC 0.97-1.00, relative error <1.3 %. Distance: significant differences between observed and One distance, ICC 0.0-0.05, relative error 5.0-39.6 %. |
Tully et. al (2014) [44] | 42 | 60 | 7Â days of free-living wear excluding water activities and sleep | Field | ActiGraph GT3X and Yamax CW700 pedometer (steps, physical activity) | Zip (Fitbit) | Right waist | Steps/day, MVPA min/day | Spearman CC: 0.91 (ActiGraph steps), 0.86 (ActiGraph MVPA), 0.91 (Yamax steps) |