STRENGTHS AND LIMITATIONS OF THIS STUDY

This study employed innovative smartphone measurement of markers of psychological distress and sleep duration among Australian adolescents.

This study enabled an investigation of determinants of sleep duration allowing for clear temporal associations between exposures and outcomes.

Study participants were predominantly female and older age adolescents (16–18 years), and therefore, did not represent the population.

High attrition with just over 25% of those recruited responding to at least one follow-up sleep ecological momentary assessment.

Introduction

Adolescence is a crucial phase of the life course involving significant social, emotional and biological development, which have lasting impacts on adult health.1 Sleep deprivation, generally defined as less than 8–9 hours of sleep per night for adolescents,2 3 has been associated with a number of adverse health outcomes among adolescents including increased risk of obesity, injury, depression, and suicidal ideation and behaviour.4–6 Despite the negative effects of sleep deprivation, studies report a lack of adequate nocturnal sleep among many adolescents while also noting an increase in sleep problems in these age groups.7–9 A recent study conducted in Australia reported about a quarter of adolescents aged 12–15 years, and more than half aged 16–17 years, get less than 8 hours of sleep per day.10

Prior evidence highlights the presence of mental health problems among adolescents with shorter sleep duration including increased risk of psychological distress,11 anxiety and depression.6 12 While the association between mental health and sleep deprivation among adolescents has been previously investigated, most of the evidence is based on cross-sectional surveys,12–24 which limits the temporal interpretation between mental health exposures and sleep deprivation. Recent studies have used mobile phones to follow up with subjects over time and conduct ecological momentary assessments (EMAs) in (near) real time to better assess temporal trends in key exposures associated with adverse health outcomes, including well-being and psychological distress.15 Accordingly, this study investigates the association between indicators of psychological distress and sleep duration among a cohort of adolescents in Western Sydney (Australia).

MethodsParticipants

This study was conducted as part of the Western Sydney Adolescent Health Study, a prospective cohort study investigating determinants of health and well-being among the general population aged 13–19 years from the Sydney (Australia) population catchment. A detailed description of the methods has been published elsewhere.15 Briefly, young people aged 13–19 years were recruited via social media platforms (Facebook and Instagram) from the general population. A youth Advisory Group was established to seek feedback on the design and implementation of project materials. The social media campaign ran from 8 November 2019 to 8 January 2020 to recruit participants and they were prospectively followed for 16 weeks. Data collection of the study spanned 27 weeks from the first participant’s recruitment to the last follow-up (from 8 November 2019 to 14 May 2020) and recruited a total of 1295 participants of which 80% were females aged 16–18 years (table 1), reflecting higher engagement in social media among females compared with males more generally.16 17

Table 1

Sociodemographic characteristics of study participants (13–19 years) at baseline (N=1295)

The study used the Ethica Data smartphone app (https://ethicadata.com/) to collect data from questionnaires, EMAs and smartphone sensors. Geolocation information (via Wi-Fi, Bluetooth and GPS) was collected automatically through the Ethica app among participants who provided consent. Following enrolment, each participant received a baseline questionnaire and a 16-week schedule of follow-up EMAs, with questions sent directly to their smartphone. There were a total of six EMAs relating to mental health factors and sleep duration (described below), sent to participants as part of a 16-week schedule of EMAs over the follow-up period. EMAs were sent to participants at random times and on varying days of the week (Monday to Saturday and one EMA per day), between 8:00 and 10:00 hours or between 15:00 and 20:00 hours, to avoid school hours and sleep periods. Among those recruited at baseline, 343 (26.5%) participants responded to at least one follow-up sleep duration EMA (table 1 and online supplemental figure 1). Participants who responded to at least one follow-up sleep duration EMA were not significantly different from the entire sample recruited at baseline (online supplemental table 1).

Outcome variable

The primary outcome variable for this study was self-reported sleep duration in the past 24 hours which was collected at baseline, via weekly EMAs for 16 weeks. Sleep duration was dichotomised into ‘less than 8 hours’ and ‘at least 8 hours’, to represent those participants not meeting the recommended minimum amount of daily sleep.3

Psychological distress

The main exposure of psychological distress was collected at monthly intervals for each participant and was measured using the Kessler Psychological Distress 6-item scale (K6).18 Responses for each K6 item were marked on an ordinal scale including ‘none of the time’, ‘a little of the time’, ‘some of the time’, ‘most of the time’ and ‘all of the time’ and were scored from 1 to 5, respectively. A K6 score ≥19 was used as indicative of a ‘probable serious mental disorder’ as described previously.18

Social media and homework

Social media use and homework were also included as potential factors associated with sleep duration.19 20 Self-reported social media use was collected via a weekly EMA that asked: In the past 24 hours, did you spend any time on the internet, social media (like Instagram, Youtube or Facebook). For participants who answered ‘yes’ to this question, a follow-up question was asked: ‘If Yes, how long did you spend on the internet, social media’. Self-reported number of hours completing homework was also collected via a weekly EMA that asked: ‘In the past 24 hours, how many hours have you spent studying or doing homework outside of class time?’

Baseline variables

The study collected a range of sociodemographic and other health factors at baseline. For this study, baseline characteristics included sex (‘male’, ‘female’, ‘other’), age (in years), socioeconomic status (SES) and body mass index (BMI) (based on self-reported height and weight and categorised into ‘underweight’, ‘normal’, ‘obese’, overweight’). Individual SES was estimated using the area level Index of Relative Socio-economic Disadvantage based on the postcode of residence.21 Postcode of residence was not included in the baseline questionnaire and was imputed from mobile phone GPS signals and was based on the most frequent GPS position between 1:00 and 5:00 hours for all available GPS data.

Data analysis

Preliminary analyses investigated counts and percentages for sleep duration activity and descriptive plots of trajectories of sleep activity were examined over the baseline and 27-week follow-up period based on weekly self-report EMAs. Multilevel mixed effects logistic regression models were used to examine the univariate and multivariate associations between time variants (psychological distress, social media use, homework hours) and invariant covariates (gender, age, BMI, employment, SES) and sleep duration. Multivariate associations between psychological distress, social media use, homework and sleep activity were examined using two separate confounders adjusted models. One of the above models was adjusted for gender, age, BMI category and SES whereas SES was excluded from the second model due to missing values in nearly 50% of the participants. A subanalysis was also conducted to examine trajectories of social media usage, hours spent on homework and K6 score by dividing sleep trajectories among participants into ‘decreasing’, ‘increasing’ and ‘stable’ groups. Linear regression analysis was conducted for individuals with more than two sleep records (n=214) and were classified as (1) ‘increasing’ if regression coefficients were positive and p≤0.25; (2) ‘decreasing’ if the regression coefficients were negative and the p≤0.25 and ‘stable’ if p>0.25. All the analyses were conducted by using STATA V.1622 and R software, V.3.6.0.23

Results

Substantial variation in sleep duration was observed among individuals over the follow-up period. The median number of weekly sleep hours was 8.33 for males (IQR 7.12–9.25 hours) and 8.25 for females (IQR 7.18–9.20 hours) (figure 1). Sleep duration was slightly higher during mid-December and the end of January, a period coinciding with the school holidays.

Figure 1

Sleep hours by week for adolescents aged 13–19 years (N=1295). The horizontal line for each box plot in figure 1 is the median number of hours of sleep, the blue bars represent the IQR, and the vertical line is the minimum and maximum (outliers excluded).

Those with high psychological distress had higher odds of sleeping less than 8 hours (OR 1.51 (95% CI 1.17 to 1.96)), compared with those with low psychological distress (table 2, online supplemental figure 2). Higher levels of social media use (OR 1.34 (95% CI 1.01 to 1.79)) and longer homework hours (OR 1.37 (95% CI 1.07 to 1.75)) were also associated with a sleep duration of less than 8 hours (table 2, online supplemental figure 2 (panel 2)). Effect sizes of associations were similar between unadjusted and adjusted models (table 2, online supplemental figure 2 (panel 2)).

Table 2

The association between hours of sleep duration and psychological distress (K6), social media use, hours of homework and other sociodemographic factors among adolescents aged 13–19 years

An inverse association was also observed between sleep hour trajectories and psychological distress, with a decreasing trajectory of sleep hours associated with increasing psychological stress (figure 2). This was also evident in the regression models, which showed a decreasing sleep trajectory was associated with increased psychological distress (β=0.069, (95% CI −0.019 to 0.149)) and increased homework hours (β=0.078, (95% CI 0.036 to 0.120)) (table 3). An increasing sleep trajectory was weakly associated with decreased psychological distress (β=−0.015 (95% CI −0.121 to 0.092)) and decreased homework hours (β=−0.006, (95% CI −0.079 to 0.066) (table 3).

Figure 2

Distribution of sleep hours and monthly psychological distress (average K6 score) by type of sleep trajectory among adolescents aged 13–19 years (n=214). The horizontal line for each box plot in figure 2B is the median number of hours sleep, the blue bars represent the IQR and the vertical line is the minimum and maximum (outliers excluded). K6, Kessler Psychological Distress 6-item scale.

Table 3

Change in hours of social media, hours of homework and psychological distress (K6) score by type of sleep trajectory

Discussion

This study investigated the association between indicators of time-varying mental health factors and trajectories of sleep duration among adolescents (aged 13–19 years) in Sydney (Australia). The study found that high levels of psychological distress were associated with a higher risk of sleeping less than 8 hours. Similarly, more hours spent on social media and on homework were also associated with a higher risk of sleeping less than 8 hours.

Previous studies have found psychological distress to be associated with shorter sleep duration among young adults.11 24 There is also evidence of a reinforcing relationship between sleep duration and depression, that is, lack of sleep increases the risk for major depression which in turn increases the risk of sleep deprivation.25 Our analysis also found evidence of this relationship with decreasing sleep trajectories associated with increased psychological distress (and vice versa). While there is a strong association between sleep difficulties and psychological disorders,26 27 recent evidence suggests that insomnia (inability to fall asleep and/or sleep through) might also be a causal factor for the occurrence of mental health problems.28 This negative impact of lack of sleep on mental well-being can be explained through brain functioning where proper sleep promotes mental and emotional resilience while lack of sleep has the opposite impact.29

While cross-sectional studies have reported the adverse impacts of heavier social media use on sleep hours and sleep quality among young adults and adolescents,20 30 31 the prospective design of this study was able to clearly delineate the temporal nature of this association. The adverse impact of social media usage on sleep hours can be interpreted in at least three ways.32 First, heavier social media use may displace sleep hours (eg, staying up late posting pictures and commenting on social media). Second, social media may stimulate emotional and/or psychological arousal which might impede sleep (eg, engaging in social media interaction). Third, mobile phones are light-emitting devices and using them before bedtime might prolong the time takes to fall asleep.33

This study also found evidence of more hours spent on homework increasing the probability of sleeping less than 8 hours, which is consistent with previous cross-sectional studies, showing more time spent on homework is associated with shorter sleep duration among adolescents.34–36 The need to complete homework is likely ubiquitous for this age cohort who were attending school or beginning tertiary education. Longer duration of time spent completing homework may reflect higher levels of engagement and mastery of content, however, homework may also be a source of stress for some students, especially as homework might relate to academic performance. While excessive amounts of homework might compel students to spend more time on it, more time spent on homework might also be indicative of difficulties in completing the homework rather than the excessive amount of homework per se.

There are several limitations when interpreting findings from the current study, also discussed in detail elsewhere.15 Briefly, study participants were predominantly female (80%) and older age adolescents (16–18 years). Despite this, the distribution of responses by key well-being outcomes was not substantially dissimilar to other representative prevalence studies of adolescents.37 38 Additionally, the high attrition rate is a potential source of selection bias and limits the generalisability of findings. Just over 25% of those recruited responded to at least one follow-up sleep EMAs. However, the baseline characteristics of the subgroup responding to follow up sleep EMAs did not significantly differ from those recruited at baseline.15 Furthermore, because the measurement of some exposures was based on self-report, recall bias must be considered as a potential source of limitation. However, the (near) real-time nature of the EMAs potentially reduced the likelihood of recall bias as questions related to the immediate period (usually the previous 24 hours), as opposed to longer index periods. Finally, part of the follow-up period also coincided with the initial period of the COVID-19 pandemic (March–April 2020), a period associated with decreased well-being and physical activity, and increased psychological distress and screen time in this cohort.15 However, the mixed models employed in the current study implicitly accounted for participant variation and period effects, and psychological distress and social media use were mutually adjusted for in the final statistical model.

In conclusion, the findings of this study highlight the adverse impacts of higher psychological distress and greater numbers of hours spent on social media and homework on trajectories of sleep duration. While prior cross-sectional studies have reported similar findings, our study was able to explore the temporal nature and reinforcing feedback loops in these associations, that is, high psychological distress was associated with less sleep duration and decreasing sleep duration trajectories was associated with increased psychological distress. The results of this study need to be replicated in larger longitudinal studies, especially with longer follow-up periods and in different geographic and cultural settings. Also, while it is a challenging task to modify health behaviours, health education programmes in schools focusing on the advantages of mental well-being and adequate nocturnal sleep may be beneficial.

Data availability statement

Data are available on reasonable request. Data are not publicly available according to the ethics approval. Data may be available on reasonable request.

Ethics statementsPatient consent for publication

Not applicable.

Ethics approval

This study involves human participants and institutional ethics approval for the study was obtained from the Western Sydney University Human Research Ethics Committee (HREC Approval Number: H13302). Participants gave informed consent to participate in the study before taking part.