Scientific Papers

U.S. Adolescent Rest-Activity patterns: insights from functional principal component analysis (NHANES 2011–2014) | International Journal of Behavioral Nutrition and Physical Activity

In NHANES 2011–2014, we identified the top four rest-activity profiles of adolescents with the use of fPCA. Both correlations and differences were observed between weekday and weekend profiles. Moreover, we observed notable variation of these rest-activity profiles across subgroups of age, sex, race/ethnicity, and household income, suggesting broad influences from demographic characteristics and family socioeconomic status.

High amplitude profile

PCA1 was primarily a measure of day-time activity level, and explained the largest portion of variance (42.8%) in rest-activity profiles, similar to previous studies in adults [5, 8]. The high amplitude profile was strongly and inversely associated with age. This is consistent with a large body of literature showing a decline in physical activity with age in adolescents [33, 34]. Physical inactivity is an important risk factors for numerous diseases, such as obesity and diabetes [35]. Moreover, health behaviors established in adolescence have been shown to track into adulthood [36], and may affect health outcomes later in life [24, 25]. Preventing the decline of physical activity in adolescents, for example through multicomponent intervention involving schools, families and communities [37], is an important public health priority. Other than older age, a lower amplitude was also observed in Asian adolescents compared with non-Hispanic whites, which agrees with previous studies showing a lower activity level in Asian adolescents and adults [38, 39].

The weekday and weekend profiles showed a relatively high correlation (0.59) and were similarly associated with demographic and family socioeconomic status. However, although boys and girls were similar with regard to the high amplitude profile on weekdays, girls showed a greater value for this profile on weekends. Given that the high amplitude profile appeared to be primarily driven by day-time activity levels, this finding appeared somewhat unexpected as previous studies generally showed a higher physical activity level among boys than girls [40]. However, most of the previous studies that investigated sex differences in physical activities in adolescents did not focus on weekdays and weekends specifically. Moreover, many previous studies focused on comparing the amount of moderate-to-vigorous physical activity [41], while the fPCA derived profiles did not specifically focus on a specific time period or apply a prefixed cutoff to define physical activity intensities. Thus, the high amplitude profile reflects the overall physical activity pattern that include both high- and low-intensity activities. Indeed, a more recent study using NHANES 2011–2014 data to compare total activity levels between boys and girls reported findings similar to ours [42].

Early activity window profile

PCA2 captured the timing of the active phase and explained 21.1–24.4% of total variance. The timing aspect was also captured in previous fPCA studies in adults, explaining 11.4–23.0% of variance [5,6,7,8,9], supporting the timing aspect as an important feature of the 24-hour rest-activity patterns. The timing of the activity window is directly related to sleep timing, and both are determined by many internal and external factors. Previous studies have consistently established an age-related delay in chronotype (i.e., time preference of the sleep window) during adolescence [43], which is consistent with the later activity window found in the oldest age group in our study. The fact that both weekday and weekend PCA2 were associated with older age suggested this association may be primarily driven by biological rather than environmental factors such as school schedule. Although it is unclear what factors drove this shift in chronotype, some studies suggested it could be partly due to changes in circadian patterns of hormones during adolescence [44,45,46].

We also observed an earlier active window in Mexican American compared with white participants. Few studies have examined racial and ethnical differences in either activity timing or chronotypes, but the result for Mexican Americans was directionally similar to findings in the previous fPCA study in adults [5]. It is also noteworthy that the relationship was not observed on weekends, suggesting that factors contributing to an earlier activity window in Mexican Americans are likely weekday specific. It would be interesting for future studies to investigate underlying contributors, internal or external, to such racial and ethnic differences. Adolescents with the lowest household income had a later active phase, and a similar association was reported in the previous study in the NHANES adult [5]. However, our study is the first that examined the relationship between family socioeconomic status and activity timing in adolescent, and more evidence is needed to clarify this relationship.

Early activity peak profile

This profile, which explained slightly less than 10% of total variance, was distinct from the early activity window profile (PCA2) in that instead of featuring a temporal shift of the active phase, it is characterized by the timing of peak activity within the active phase. The only characteristic that was associated with this profile was age, with the oldest age group (18-19.9) exhibiting the highest eigenvalue. Notably, the association between different age groups and the profile characterizing the timing of peak activity differed for weekday (PCA3) and weekends (PCA4), suggesting that this profile may be partially affected by external factors, such as school schedule and after-school sports activities. Specifically, the earlier timing of peak activity among the 18 + age group may be driven by a lower participation in after-school sports training after graduating from high school. Recently, a number of studies suggested that the timing of exercise may have an impact on its health effects [47, 48], however the evidence is still limited and research among adolescents is lacking. Future studies should examine how timing of exercise may affect physical and mental health in adolescents, which will provide valuable information for designing school-based and extracurricular exercise programs among children and adolescents.

Prolonged activity/reduced rest window profile

The length of day-time activity was captured by the PCA4 which explained 7.1–7.7% variance. A similar profile was reported in previous fPCA studies in adults, explaining 9.1-14.8% variance [5,6,7,8,9]. Since the 24-h rest-activity cycle can be divided into an active and a resting phase, a higher value of this profile therefore indicates both a prolonged window for activity and a reduced window for rest or sleep. Thus, the more prolonged activity window found in the older adolescents essentially reflected a shorter sleep duration, which is consistent with the results found in the Youth Risk Behavior Surveys showing U.S. students in grade 12 had a higher prevalence of short sleep duration than students in grade 9 (77.6% vs. 65.6%) [49]. Another nationally representative study of the U.S. adolescents also found those aged 18–19 were 68–75% less likely to report ≥ 7 h of sleep per night, compared with those aged 12–13 [15]. In addition, girls on average showed a more prolonged activity window than boys did in our analysis, which is consistent with previous studies which showed adolescent girls in the U.S. had a higher prevalence of short sleep duration than boys (75.6% vs. 69.9%) and were about 30% less likely to report ≥ 7 h of sleep per night [15, 49]. A few mechanisms have been proposed to explain shorter sleep duration in girls relative to boys, including hormonal changes during puberty, higher stress and mental problems [50,51,52]. However, such sex difference was only observed for weekdays, not weekends, suggesting social factors may play an important role.

Non-Hispanic Black participants on average showed a higher eigenvalue for this profile than Non-Hispanic White participants, which is consistent with findings in NHANES adults [5]. This association may be primarily driven by a shorter sleep duration in the Black participants. Multiple studies reported a shorter sleep duration in Black adolescents than White adolescents [15, 49, 53]. Such racial disparities in sleep may be explained by various factors, including environmental disturbances such as light and noise due to poor housing and neighborhood conditions [54, 55], higher stress [56], and a higher prevalence of chronic conditions such as obesity and cardiovascular diseases in African Americans [53, 57]. Similar to sex differences, racial difference in this profile appeared stronger on weekdays, while the Black-White difference almost vanished on weekends. This suggested that environmental constrains specific to weekdays (e.g., school schedules, commute, school work) may have contributed the disparities in sleep duration between racial groups.

Strengths and limitations

A major strength of our study is the use of fPCA to derive activity profiles that are important for explaining population-level differences in 24-hour rest-activity patterns. Compared to other commonly used methods for characterizing rest and activity behaviors, including measurements of total physical activity levels and moderate-to-vigorous physical activities, duration of sedentary behaviors and/or sleep, and metrics derived from cosine-based models or non-parametric algorithms (e.g., interdaily stability, intradaily variability), a notable strength of fPCA is its flexible, data-driven, and shape-naïve approach. Specifically, fPCA does not rely on prefixed intensity cutoffs or assumptions about 24-hour shapes of rest-activity rhythms, which are often insensitive to differences in activity levels and patterns that vary according to population characteristics (e.g., sociodemographic compositions, health status). As a result, the fPCA is able to capture the most prominent activity profiles or features specific to a study sample. This is a particular strength for our study, which utilized a representative sample of American adolescents, because profiles derived from our analysis are reflective of rest-activity patterns on a national scale and thus have important public health implications. Moreover, the data-driven and flexible approach of fPCA has allowed us to uncover profiles that represent not only previously established crucial domains of rest and activity patterns (e.g., amplitude), but also novel features that have not been documented by earlier studies: For example, we found that the early peak profile was a unique feature distinct from the early activity window profile, which revealed a potentially important nuance in the timing of rest-activity rhythms that has not been reported before.

Beside the application of fPCA, our study has several additional strengths. The data from NHANES allowed us to assess the rest-activity profiles in a national representative sample of U.S. adolescents. The large sample size made it possible to compare across sociodemographic subgroups. The use of accelerometer data provided objective and accurate rest-activity measurements. We also stratified the analysis by weekdays and weekends to examine any differences between schooldays and weekends, which provided important evidence about the role of social and environmental factors in shaping rest-activity patterns in this population.

Despite the strengths, our study also has limitations. First, the rest-activity profiles derived by fPCA are sample specific which tend to weaken external validity. However, we believe the use of the national representative sample in NHANES allows our findings to be generalized to the wide population of American adolescents. However, more studies are warranted to investigate special populations, particularly those with underlying health conditions that affect the rest-activity rhythm, such as sleep and mental disorders. Second, the data were collected from 2011 to 2014, therefore, the activity patterns and the associations observed in our study may not necessarily apply to adolescents in different birth cohorts and/or periods. Third, the excluded participants were generally older and more likely from the highest household income group, compared with the included participants. Therefore, the external validity may be reduced and the results should be interpreted with caution. Fourth, the profiles generated from fPCA could sometimes be difficult to interpret. However, the rest-activity profiles described in our study provided distinct and interpretable features which captured characteristics that were also observed in previous fPCA studies. Fifth, the accelerometer data in NHANES only captured the levels of activity and do not provide information about the specific types of activity or in what context the participants were engaging in these activities. Additionally, since the accelerometer was worn on the wrist, it may not accurately measure activities that primarily involve the lower body. Future studies are needed to examine the specific types of activities in which adolescents participate, and explore their association with demographic characteristics and health outcomes. Lastly, we only examined five demographic and socioeconomic status characteristics. Other factors, such as lifestyle and environmental factors, may also impact the rest-activity profiles and need to be examined by future studies.

Source link