A total of 100 participants were recruited. Approximately even distributions were recruited across sex and age groups (6-9, 10-12, 13-15, and 16-18 years). Participant characteristics are shown in Table 1.

All research visits were conducted in the Applied Physiology Laboratory and the surrounding area at the University of Tennessee Knoxville. The protocol entailed 2 visits on separate days, each lasting approximately 2-2.5 hours. The main feature of the protocol was the completion of 16 activities (Table S3 in Multimedia Appendix 1) that were divided across the 2 visits. A key goal of this study’s design was to simulate realistic transitions between activities. Therefore, each activity was performed twice, with 1 instance lasting approximately 60-90 s and the other lasting approximately 4-5 min. Within those ranges, participants self-selected when they transitioned to the next activity unless guidance was needed from the research team (eg, if the upper limit of 90 s or 5 min was reached or if the walking circuit reached a specific stopping point to facilitate speed calculations). They also self-selected the order of the 8 activities they were prescribed each day, except that the same activity could not be performed twice in a row, and the activities were grouped by location (ie, in an upstairs laboratory or a separate recreation area) to minimize total transition time. The order of short and long bouts for each activity was randomized.

Throughout each research visit, participants wore a portable indirect calorimeter and a variety of activity monitors. Participant behavior and posture were coded by direct observation, using a focal sampling approach that allowed continuous logging of behaviors, transitions, and posture. Video recording was also performed for most participants. The video recordings served as a backup in case the live coded observation records needed to be reviewed and revised for any reason (eg, adjusting timestamps if an error was noted during the live coding process).

For the first visit, participants were asked to refrain from eating or drinking (except water) for 2 hours before the visit and to wear lightweight athletic clothing and closed-toed shoes. After completing the enrollment and consent processes, they removed their shoes and socks and had anthropometric measurements taken. Standing and sitting height (cm) were measured using a wall-mounted stadiometer. Body mass (kg) and body fat (%) were measured using a Tanita BC-418 bioelectrical impedance analyzer (Tanita Co. of America, Inc).

Following the anthropometric measurements, participants were fitted with this study’s equipment and completed a 30-minute assessment of resting metabolic rate (RMR). The remainder of the visit involved completing the 8 activities assigned for that day. Participants received a US $30 gift card at the end of the visit.

For the second visit, participants were given the same previsit instructions. Upon returning to the laboratory, they were fitted with the same study equipment as visit 1 before performing the remaining 8 activities, such that all participants completed the full set of 16 activities under normal circumstances (ie, assuming there is no withdrawal, refusal, or space unavailability due to scheduling). Participants received a US $45 gift card at the end of the visit.

All participants wore 5 ActiGraph GT9X devices (GT9X; ActiGraph LLC), which were the primary activity monitors of interest for this study. One was placed on the right hip, 1 on each wrist, and 1 on each ankle. The hip-worn device was worn on an elastic belt and positioned over the right iliac crest along the anterior axillary line and secured using a manufacturer-specific belt clip. The wrist-worn devices were worn in manufacturer-specific watch bands that the devices clipped into. The watch bands were positioned proximal to the styloid processes with the devices oriented where the LCD could be read naturally by the participant, as a standard watch face would be. The ankle-worn devices were clipped into manufacturer-specific clips worn on slip-through Velcro bands. The clips were positioned so that the devices were on the lateral aspect of the ankle proximal to the lateral malleolus. Additional detail is given in Section S1.1 in Multimedia Appendix 1.

Secondary activity monitors, both research- and consumer-grade, were also worn in various locations on the body as detailed in Sections S1.2 and S1.3 in Multimedia Appendix 1.

Participants were observed using the Noldus Observer XT and Pocket Observer software platforms (Noldus Information Technology), which served as a criterion measure of behavior and posture. Participants also wore a Cosmed K4b² (COSMED s.r.l.), which was the criterion measure of EE. General details of these tools can be found in Sections S1.4 and S1.5 in Multimedia Appendix 1. Specific details applicable to part 1 of this study are discussed below.

The Observer XT desktop platform (version 12.5) was used to create a 2-class focal sampling coding scheme, which was then exported for use in the Noldus Pocket Observer application (version 3.2). The latter application was run on Samsung Galaxy Tab 4 tablets (Samsung) and used for real-time coding of participant behavior and posture. Several activity monitoring studies have taken a similar approach to direct observation using Noldus products [22,28,32-34].

The coding scheme included a 19-item list of activity behaviors (ie, for the 16 activities listed in Table S3 [Multimedia Appendix 1], plus labels for the start time of the observation period, transitions between activities, and unknown activities) and a 6-item list of postures (lying, reclining, sitting, standing, stepping, or unknown). The label for unknown activities was used whenever an unlisted activity occurred, such as a water break or bathroom break, and was accompanied with a posture label of unknown.

Whenever a new activity was initiated, users tapped the corresponding label in Pocket Observer, which issued a timestamp for the new activity and then prompted the user to indicate the participant’s posture. This created a continuous and timestamped record of activities and postures. After finishing the protocol, live-coded observations were imported back into the Noldus Observer XT software program for conversion to Excel format. The resulting files were then processed in R (R Foundation).

Section S1.5 in Multimedia Appendix 1 describes the procedures that were used when syncing data from the various sources.

While the activity monitor data were collected for flexible use (eg, to be processed and tested in different ways across future studies), the criterion data from direct observation and indirect calorimetry were designed for consistent usage across future studies. Therefore, we describe the processing and aggregation of criterion data in the sections that follow, while monitor data processing is left for future studies to describe according to the unique needs of the study.

Each participant was observed by a single trained observer in real time during each study visit. The observation files were inspected for comments and anomalies that indicated errors (eg, the user entered a comment indicating a button was pressed at the wrong time or that the activity duration was outside the expected range). Corrections were made by a senior reviewer cross-referencing the video recordings and adjusting labels or timing accordingly. For participants without video recordings, observations were used as they were created from the live coded direct observation (n=3). Each row of the observation files indicated a start time and duration for the corresponding activity and posture, and by definition, there were no gaps between activities. Thus, every second of the data collection period could be mapped to a specific activity and posture label that was active at that time, allowing a second-by-second record to be constructed from the information in the observation files.

EE was calculated based on oxygen consumption (VO₂) and RMR, both measured by the K4b². The values were expressed in youth metabolic equivalents (MET_y), where

As described in the Sections S2.1 to S2.3 in Multimedia Appendix 1, calculations for RMR and continuous VO₂ were the same for parts 1 and 2 of the FLPAY study. Below, we provide additional detail about calculating steady state EE as this was unique to part 1 of this study.

Steady state periods were defined for the longer activity bouts. All bouts longer than 3 minutes, 40 seconds were included in the analysis. The steady state period was defined as the last 60 seconds of each activity, after discarding the final 10 seconds. For each of these periods, the corresponding K4b² data were extracted in breath-by-breath format. An average VO₂ was then calculated using manufacturer-specified procedures (Section S2.1 in Multimedia Appendix 1). This value was divided by RMR to obtain MET_y for the steady state period. Posture was determined by cross-referencing the direct observation data from the steady state period.

Activity intensity was derived from EE (MET_y) and posture, using the following accepted definitions [35]: for sedentary behavior, ≤1.50 MET_y and seated, lying, or reclining posture; for light physical activity, 1.51-2.99 MET_y regardless of posture (or ≤1.50 MET_y if posture was upright); and for moderate-to-vigorous physical activity, ≥3.00 MET_y regardless of posture. For each activity and age group (ie, those aged ≤12 versus >12 years), data were cleaned by removing values <0.2 MET_y, or >2 SD above or below the mean.

A total of 84 participants were recruited to this study, of whom one was excluded (final N=83). Participant characteristics are shown in Table 2.

The overall goal of this part of this study was to collect free-living data with diverse activities and transitions between them. Therefore, research visits were conducted in varied settings, such as in the community, participants’ homes, local parks, and approved facilities of community partners. Data were collected across 2 separate days, with the goal to have one in the participant’s home and the other somewhere else.

The first study visit was generally conducted in the participant’s home environment and began with the informed consent and assent process and participant screening. Once enrolled, participants had anthropometric measurements taken, including standing and sitting height (cm) using a portable stadiometer and body mass (kg) and body fat percentage measured using a Tanita BF-350 bioelectrical impedance analyzer. Anthropometric measurements were taken with shoes and socks removed. Following the completion of the anthropometric measurements, a 30-min RMR assessment was conducted using a Cosmed K4b² or K5 (used in breath-by-breath mode to align with the K4b²) portable indirect calorimeter before participants engaged in activities of their choice for up to 2 hours. In some cases, the RMR assessment was conducted on a separate visit (for a total of 3). Participants were compensated US $25 for the RMR assessment and US $25 for the free-living data collection period.

For the second study visit, an alternative environment was sought for each participant. When this was not possible, participants were asked to engage in different activities than what they chose during their initial visit, which helped to mimic different environments and promote engagement in a wide range of activity behaviors. Similar to the first visit, the participants engaged in behaviors of their choice for up to 2 hours. An additional US $25 of compensation was given for completing the second study visit.

Throughout this study’s visits, participants were given free rein to select activities, so long as they had parental permission, did not put the equipment at risk (eg, from water damage), and did not risk harming themselves or others.

All participants wore the same research-grade devices as described in the equipment portion of part 1 including the primary devices of interest (ActiGraph GT9X on the hip, both wrists, and both ankles; see also Section S1.1 in Multimedia Appendix 1) and the secondary research- and consumer-grade monitors (see Sections S1.2 and S1.3 in Multimedia Appendix 1, respectively). Participants wore a Cosmed K4b² or K5 portable indirect calorimeter, both of which collected breath-by-breath data as a criterion measure of EE (see Section S1.5 in Multimedia Appendix 1 for more detail). During the consent process, participants were able to opt-in to be video recorded throughout each research visit, and only 3 participants did not do so. Live coding was performed for those who did not opt into the video recordings. For the others, the video recordings were later used for direct observation to obtain a record of the behaviors and postures they engaged in, along with relevant contextual factors (eg, indoors versus outdoors and alone versus with others). These video-based observations were performed using the Noldus Observer XT software program for Windows desktop computers.

Most equipment matched what was used for part 1 of the FLPAY study, with the addition of the Cosmed K5 alongside the K4b². The main difference in part 2 was the coding scheme used for direct observation and the reliance on video recordings for direct observation rather than live-coded observations. For participants who did not opt into video recording, real-time observations were performed using similar procedures to those described in part 1 above, except that a new coding scheme was used (Table S2 in Multimedia Appendix 1). Otherwise, the video recordings were coded using the Noldus Observer XT Direct Observation software program (version 12.5) for Windows desktop machines.

The coding scheme for part 2 included a longer list of potential activities and postures, as discussed below. It also included prompts about the activity context (eg, indoor or outdoor and alone or with others). Otherwise, the general structure of the coding scheme matched with part 1, such that every second of the data collection period could be mapped to a single entry in the observation record.

Table S4 in Multimedia Appendix 1 lists the main activities that were included in the coding scheme. As before, extra labels were included for coding the start of the observation period, transition behaviors, and unknown activities (ie, activities not previously defined), and events where the participant was off-camera (ie, in a private space, bathroom, etc). After the primary activity behavior was coded, posture was coded as lying, sitting, standing, standing or stepping, squatting or kneeling, reclining, mixed posture, or unknown (used only when the activity behavior was “off camera”). Section S1.5 in Multimedia Appendix 1 describes the procedures that were used when syncing data from the various sources.

The data processing procedures for part 2 broadly mirrored what has already been discussed in part 1 resulting in a second-by-second dataset with information about EE, behavior, posture, and context. The direct observation procedures differed from part 1 due to the open-ended nature of this study and the reliance on video recordings. Specifically, at least 2 trained reviewers coded each video, and a third senior reviewer then compared their observations side-by-side. Additional detail is available in S1.4.3 in Multimedia Appendix 1.