Participant recruitment
The study was conducted in three cohorts. The first cohort [N = 25; mean (SD) age = 12.76 (.72)] provided data in February and March of 2015, the second cohort [N = 35; mean (SD) age = 11.15 (.43)] provided data in September and October of 2015, and the third cohort [N = 27; mean (SD) age = 12.74 (.52)] provided data in February and March of 2016. The study was approved by the Institutional Review Board of the University of California, Irvine and the Long Beach Unified School District Research Review Committee. Students were recruited from a public middle school in Southern California via flyers and oral presentations that were given during their physical education (PE) class. Students and parents/guardians provided written informed assent and consent prior to participation. In Cohorts One and Three, students were recruited without regard to level of participation in physical activity. In Cohort Two, students were required to be a member of a sports team to participate in the study. All study participants were required to be eligible to participate in regular physical education, thus ensuring that they were generally healthy. The data collected for the present study were obtained as part of two larger ongoing studies. The inclusion criteria related to activity levels were imposed by the study protocols.
Procedures
Each study participant underwent an assessment of height, weight and cardiorespiratory fitness (conducted at the school in a converted classroom) followed by 1 week of activity monitoring with the Fitbit Zip (San Francisco, CA, USA) attached to the belt of the ActiGraph activity monitor (model GT3X, ActiGraph, Pensacola, FL, USA). Prior to sending participants into the field with the activity monitors, the ActiGraph was fully charged and a new battery was installed in the Fitbit. Participants were instructed to wear the belt with both activity monitoring devices every day for 7 days, except when sleeping or bathing. Consistent with recommendations for obtaining a valid estimate of daily activity [12], 4 days of valid data was the minimum required to be considered complete. If a student returned incomplete data, as determined by the ActiGraph, the data collection was repeated. Fitbit accounts were established for each device by the research team. Students did not have access to their own accounts and were not given the password to view their data on-line.
Measures
ActiGraph
The ActiGraph activity monitor is a tri-axial accelerometer that is attached to a belt that wraps around the hip and is not waterproof. It is marketed exclusively as a research device, and the cost of the ActiGraph is approximately $225. The ActiGraph is widely used in physical activity research and has been validated against objective measures of motion and of energy expenditure [13, 14]. The ActiGraph can store data for up to 40 days, is rechargeable, can run on battery power for up to 30 days, and syncs to a computer through a cable. A newer version of the device, not used in this study, can sync using Bluetooth.
Fitbit Zip
The Fitbit Zip, a tri-axial accelerometer, is marketed as a consumer-oriented device. The device is held by a silicon clip that can be attached essentially anywhere on the body, and is water-resistant. The Fitbit Zip can store data up to 7 days, and syncs wirelessly and automatically up to a 20-foot range. The cost of a Fitbit Zip is approximately $60 U.S., and it requires a non-rechargeable replaceable battery every 6 months. The Fitbit Zips used in this study were purchased new directly from the company in the fall of 2014.
Demographics
Students self-reported their age and ethnicity. Two questions determined ethnicity according to the format used by the National Institutes of Health. Students were first asked to indicate if they were Latino/Hispanic (yes/no) and then asked to check a category indicating race (American Indian, Asian, African-American, Hawaiian/Pacific Islander, White, Multiracial, Other).
Data analysis
Data from the ActiGraph were aggregated using the Actilife software with the following parameters specified: (1) a valid day included at least 8 h of wear time; (2) for an hour to be included in wear time, it could not include a string of 30 min with zero activity; (3) data must be available for a minimum of 4 valid days. The Actilife software was used to yield both mean daily minutes of MVPA and mean daily steps across all valid days. The threshold for activity to be included as MVPA was computed using the formula recommended by Freedson [15] for children and using 4 METS (metabolic equivalents) as the minimum threshold for MVPA. For Cohorts One and Three, the average age of participants was 12 years, and the cutoff for MVPA yielded by the Freedson equation was 2058 counts per minute. Cohort Two participants were slightly younger (average age was 11 years old) so the cutoff for MVPA yielded by the equation was 2220 counts per minute. These parameters are easily specified in the software, and can be used to analyze all participant data simultaneously to yield a number that represents average daily minutes-per-day of MVPA. The selection of the relatively stringent criterion for non-wear time (i.e., a string of 30 min with zero activity) has been criticized in other studies for the potential to create a biased sample [16]. However, since we repeated the assessment until all participants met the inclusion criterion, no participants were excluded on the basis of failing to meet the criterion for valid wear time.
Utilizing the Fitabase software (Small Steps Lab, San Diego, CA, USA), data from the Fitbit Zip were exported as minute-by-minute data. The minute-by-minute data were used to verify that each participant included in the analyses had at least 4 valid days of data, with a valid day being defined as at least 8 h of valid data (a valid hour being one that did not contain more than 20 consecutive minutes of zero steps). Including only valid hours, the minute-by-minute data then were used to compute mean daily steps and MVPA (i.e., the total of what Fitbit calls “active” and “very active” minutes) for all valid days. Fitbit does not provide access to the raw counts-per-minute data. We explored using a string of 30 consecutive zero steps minutes as the criterion for a valid hour, to mirror the approach used with the ActiGraph, but found that this criterion was overly strict and resulted in very few valid hours on the Fitbit data. Using 20 consecutive zero steps minutes effectively eliminated obvious non-wear time and avoided the exclusion of periods during which the participant may have been minimally active but still wearing the device.
Comparisons of the average minutes-per-day of MVPA and average steps per day derived by the two instruments were conducted using Pearson’s correlations and Bland–Altman analyses. To examine whether there was a systematic difference between the devices, a one-sample t test was conducted to compare the mean difference between the estimates from the two devices (e.g., ActiGraph steps—Fitbit steps) and zero. A regression analysis in which the difference between the two device estimates was regressed on the mean of the two estimates was used to expose any proportional bias (i.e., change in the correlation between the two devices in relation to the magnitude of activity). In constructing the Bland–Altman plot, the y-axis represents the difference between the ActiGraph and the Fitbit estimates, and the x-axis represents the mean of the two estimates [17].