Identifying trends in children’s time use to address problematic lifestyles has been recognized as a global priority to minimize the burden of noncommunicable diseases in adulthood [1]. As compelling evidence regarding the quality of children and adolescents’ time use and its influence on health emerge, the World Health Organization [2] as well as several countries, including Singapore [3], have developed integrated movement guidelines to reinforce the importance of leading balanced and active lifestyles.

Understanding the lifestyle behaviors of children in relation to the integrated guidelines developed can potentiate the development of targeted behavioral interventions and programs aimed at improving and promoting healthy lifestyles that children can sustain and bring into adulthood. To do so, “adequate, affordable, and convenient” data collection measures to assess 24-hour movement behaviors are required [4].

Self-reported measures (eg, questionnaires, diaries) are commonly used for the collection of information on and assessment of movement behaviors [5,6]. They are inexpensive and easy to administer, however, are subject to the reliance on the respondent’s memory, resulting in recall bias and social desirability bias [5,7]. Children have been found to struggle with reporting durations and intensities of activities [8] and often lack the motivation to complete questionnaires [9], limiting the validity of these measures [4,10,11]. Furthermore, most questionnaires are specialized and focus on single behaviors, often requiring a combination of questionnaires to assess all movement behaviors. Having multiple questionnaires may be tedious for researchers to administer and increases participant burden. There is thus a demand for assessments that can capture the full spectrum of movement behaviors (ie, moderate-to vigorous physical activity [MVPA], light physical activity [LPA], sedentary behavior or inactivity, and sleep).

Objective methods, such as the use of accelerometers, offer a reliable and valid means of objectively capturing 24-hour movement behavior data [12,13]. However, they can be expensive and logistically challenging to administer [14], in that collecting, processing, and analyzing accelerometer data can be complex and require expertise [15]. In addition, accelerometers do not capture contextual information of these movements (eg, type and location of the activities undertaken) [11,14] and are unable to objectively assess screen-viewing, a pertinent behavior with well-established negative associations with the physical and psychosocial outcomes of children [16]. These limitations impede the understanding of children’s behaviors and consequentially limit the development of targeted behavioral interventions.

The limitations of existing methods that assess movement behaviors warrant a need for a valid, low-burden, and cost-efficient data collection method to collect 24-hour movement information from children. Advancements in technology may circumvent some of the challenges of existing data collection methods, yet few digital assessments exist. Multimedia activity recall for children and adolescents (MARCA) [9], Synchronised Nutrition and Activity Program (SNAP) [17], and MyDailyMoves [4] are digital assessments developed to provide ease of collecting self-reported 24-hour movement behavior information from children. However, not all self-reported behaviors reported on these assessments have been validated against objective measurements; therefore, the utility of these assessments for collecting 24-hour movement behavior information in comparison to objective measures remains unclear. These applications were also developed for Western populations (ie, Australia, the United Kingdom, and the Netherlands, respectively), and the behavioral patterns of children in Western versus Asian populations may differ (eg, the prevalence of children attending tuition or shadow education) [18], limiting the relevance of existing applications to the Asian setting.

To bridge this gap, My E-Diary for Activities and Lifestyle (MEDAL), an interactive web-based diary, was developed to collect time-use information from children of at least 10 years of age in Singapore [18]. Usability testing suggests that MEDAL is a feasible application for capturing the movement behaviors of children aged 10 to 12 years [18]. We aimed to validate the use of MEDAL for assessing children’s movement behaviors by comparing self-reported and accelerometer-measured time spent in MVPA, LPA, inactivity, and sleep among children aged 10 to 11 years in Singapore. We hypothesized that children at this age would be able to self- report their movement behaviors on MEDAL accurately although some differences between self-reported and objective measures would be expected.

In this study, we evaluated the validity of MEDAL by comparing self-reported time spent in each movement behavior reported (ie, MVPA, LPA, inactivity, and sleep) on MEDAL to accelerometer measurements among children in Singapore aged 10 to 11 years. We found that the children reported lower LPA, higher MVPA, higher inactivity, and higher night sleep durations than those measured by the accelerometer. Moderate-to-strong correlations were observed between accelerometer-measured and self-reported MVPA, inactivity, and night sleep, while LPA demonstrated a poor correlation. There was generally poor agreement between accelerometer-measured and self-reported values for all behaviors. Differences between weekdays and weekend days in correlation and agreement between self-reports and accelerometer measurements were also observed.

We demonstrated that some movement behaviors were better reported on MEDAL compared to others. In particular, LPA was most poorly reported, as it had the lowest correlation with accelerometer measurements, and a similar strength of correlation was reported previously (r=0.09) [32]. The literature suggests that it is difficult to define LPA and, consequentially, difficult to capture LPA using subjective assessments such as self-reports, especially among children [33]. This is attributed to the sporadic and spontaneous nature of activity among children of this age [14]; periods of light activity may be intermittent in nature, which is attributable to children’s short attention span and quick disinterest in continuous activity. These short bursts of light activity contribute to differences in objectively measured and self-reported physical activity and inactivity durations, as short periods of movement or inactivity might not be salient during their recall and thus, unintentionally misreported. The strength of associations between self-reported and accelerometer-measured MVPA and inactivity was similar or better than that reported in other studies, respectively. Recent systematic reviews [10,11] found that self-reported MVPA durations among children aged around 10 to 11 years had moderate correlation with objective measurements (r=0.21-0.48) [32,34-37], whereas correlations were poor for self-reported inactivity (r=0.06-0.14) [35,38]. Remembering frequencies, durations, and intensities of all activities undertaken in a day is difficult for children [8], contributing to the recall bias that underlies observable differences in time spent in each movement behavior between objectively measured and self-reported measures.

Lastly, for sleep, we demonstrated that night sleep was better reported on MEDAL compared to the other behaviors. Based on a systematic review of sleep questionnaires developed for children and adolescents [39], the criteria for assessing sleep duration in previous studies differ from the current one, and thus the ability to make comparisons to other studies was limited. Notwithstanding, it has been established that sleep latency, defined as the time a person takes to fall asleep after going to bed, may influence reporting of sleep times [40]. Periods between bed time and falling asleep might be reported as sleeping but detected from accelerometer data to be otherwise, resulting in differences between objectively measured and self-reported sleep onset time and, consequentially, total sleep time.

Previous studies have typically only investigated the correlation between objective measurements and self-reports, which is a limitation when assessing validity, as correlation does not provide information on the strength of agreement between the methods [41]. In our study, we examined agreement and found that while MVPA, inactivity, and night sleep demonstrated moderate-to-strong correlations between accelerometer measurements and self-reports on MEDAL, agreement for all behaviors was poor. To our knowledge, only MVPA has been examined previously in terms of agreement between objective measurements and self-reports [37,42]. The findings of our study were in line with these studies (ICC=0.06-0.25) in that agreement was poor. Although children are able to report some behaviors using MEDAL, actual time spent in these behaviors may differ from estimates by wrist-worn accelerometers.

Finally, we investigated whether the validity of MEDAL differed on weekdays and weekend days. This adds to the existing literature, as few studies have investigated differences in reporting accuracy on weekdays versus weekend days. Our study suggests that the children were able to report weekend days’ physical activity better than weekdays’; conversely, they were able to report weekdays’ inactivity and night sleep better than weekend days’. The greater correlation and agreement for inactivity and night sleep demonstrated on weekdays might be attributed to weekdays being more structured, characterized by consistent wake and bed times [43] as well as regular school-day routines in which the majority of activities are inactive (eg, lessons in school) [44], making them easier to recall. Explanations for differences in reporting accuracy for physical activity (MVPA and LPA) on weekdays compared to weekend days are less clear. Participants might have reported their physical education lessons or after-school sports trainings on weekdays as entirely physical activity (regardless of MVPA or LPA). In reality, these periods of “activity” might include organizing the lesson or game, watching demonstrations, or taking turns to rest and play [33]. The accelerometer only measured the amount of time when the individual was actually moving, resulting in the discrepancy between periods of physical activity reported on MEDAL and those measured by the accelerometers. Conversely, the participants might have participated in a specific sport and might have been active throughout the duration of the physical activity they reported, resulting in the greater correlation and agreement between accelerometer-measured and self-reported physical activity durations on weekend days. This hypothesis warrants further investigation.

Despite known limitations of self-reports relating to recall bias, self-reporting offers advantages that objective assessments of movement behaviors, like use of accelerometers, do not [6]. Objective assessments are independent of recall bias, while self-reported measures, like MEDAL, are easier to administer, process, and analyze, particularly when extended to population-based studies or large cohorts [45]. They also offer the potential of collecting information on screen-viewing and contextual information, such as location and type of activity, without the need for complementary questionnaires or devices (eg, a location tracker) [45], streamlining the assessment and identification of problematic health behaviors and possible contexts to target health behavior interventions. Hence, the impetus for selecting an appropriate method to assess movement behaviors, whether self-reports or objective assessments or a combination of both, is dependent on the research question [45].

Our findings suggest that based on self-reports on MEDAL, the children are able to estimate their time spent in MVPA, inactivity, and night sleep although actual time spent in these behaviors may differ from estimates by wrist-worn accelerometers. On the other hand, self-reported LPA on MEDAL warrants cautious interpretation. Finally, MVPA and LPA might be better reported by children on weekend days, and inactivity and night sleep might be better reported on weekdays.

There are several strengths of the present study. High compliance with accelerometer wear allowed activity patterns of a full or close to a full 24-hour day to be objectively measured for comparison with self-reports on MEDAL. The reporting of all 4 movement behaviors was investigated, allowing us to review the validity of MEDAL as a tool for capturing time use across the full movement spectrum of 24 hours. Stratifying the analyses by weekdays and weekend days provided additional interesting findings to contribute to the current literature.

However, there are limitations to be acknowledged. First, we note that some participants might have removed their accelerometers during certain sports and activities (eg, contact sports, martial arts); therefore, accelerometer-measured MVPA and LPA might have been underestimated. As our results demonstrate that participants reported much lower LPA than did the accelerometer, the underreporting may be more pronounced than reflected. Sleep duration detection algorithms for raw acceleration data have only been developed to detect the onset of sleep [22], which may differ from bed time and would be intuitively self-reported. The difference in time between sleep onset (ie, when the participant truly falls asleep) and bed time (ie, the time the participant goes to bed to go to sleep) would consequentially influence estimations of sleep durations. As accelerometers and MEDAL assessed different constructs of sleep (ie, sleep onset to wake time versus bed time to wake time), accelerometers might be less adequate as a means of comparison against self-reported sleep durations. We also removed 1 MEDAL entry of a participant whom we deemed to have reported implausible values. Sensitivity analyses, however, revealed that including the entry did not affect results substantially. We also acknowledge that the sample size of this study was modest, largely due to the exclusion of participants that did not have any day of recording on MEDAL that corresponded with the days the accelerometers were worn. Notwithstanding, MEDAL was reported previously to be feasible and acceptable among children of this age group [18]. Some differences in characteristics exist between participants included in our study and those recruited but not included (ie, those excluded from movement behavior validation analyses and those involved in the other validation studies; Multimedia Appendix 2). These may potentially limit the generalizability of the findings of this study to the population.

This is among the first studies to assess the validity of self-reported behaviors across the full movement spectrum and to compare reporting accuracy on weekdays versus weekend days. This study suggests that MEDAL may be useful in assessing movement behaviors of children aged 10 to 11 years although estimates may differ from wrist-worn accelerometer measurements. Self-reported estimates for inactivity and sleep on weekdays might be more accurate than those on weekend days, whereas self-reported estimates for MVPA and LPA on weekend days might be more accurate than those on weekdays. Findings of this study will facilitate the interpretation of future data collected using MEDAL.