Keywords
actigraphy - sports - athletes - sleep
Introduction
In addition to the increasing reports on the prevalence of poor sleep in the general
adult population,[1] recent research has also revealed that short sleep duration is prevalent in elite
athletes.[2]
[3]
[4] Current research in elite athletes has shown that sleep disruption has been found
to have a negative effect on athletic performance, including speed,[5] endurance,[6] attention,[7] and aerobic capacity.[8] Sleep is vital to elite athletes' physical and cognitive recovery, including their
daily physiological growth and repair, conservation of energy, and reaction time.[9]
[10]
Adolescent athletes experience high levels of sleep disturbance due to increased academic
and training demands, often requiring early wake-up times and late-night training
sessions to fit in with school schedules.[11]
[12]
[13] Further exacerbating this issue, adolescents experience a delayed circadian rhythm
due to many varying factors including biological, psychological, and sociocultural
influences.[14] Sleep onset and offset is usually much later in adolescents, primarily due to a
change in the biological timing mechanism that causes a delay in sleep onset.[15] The reduced sleep duration and the later onset time (due to phase delay) in adolescents
has been referred to as the “Perfect Storm” model that impacts on sleep behavior.[14] Despite this, very little is known about the sleeping patterns of highly trained
adolescent athletes in sports where training schedules may limit the sleep opportunity
(e.g., swimming).
Increased electronic device usage before bedtime may impact the delayed sleep onset
adolescents experience.[16]
[17]
[18] A recent study by Galland et al.[19] showed that more than 84% of this group used at least one electronic media device
on 3 or more nights per week. Previous research has also shown that exposure to device
usage before bed suppresses melatonin in adolescents.[20] However, while it is known that shortened sleep duration can be associated with
increased electronic device usage, it is still unclear what effects device usage has
on adolescents' sleep indices.[21] Furthermore, to our knowledge, night time device use in adolescent athlete populations
has not been explored yet. It could be postulated that this group has lower levels
of device use in an attempt to manage their challenging training schedules.
With the delayed sleep onset experienced by adolescents, the offset must also be later
to cater for the delayed sleep/wake cycle, allowing adequate sleep quantity. However,
due to growing demands, such as earlier school starts, extracurricular activities,
and part-time work, the later offset that adolescents require is not possible.[22]
[23] Previous research has also shown that sleep deprivation has been related to injury
occurrence in adolescent athletes.[22]
[24]
[25] A study by Milewski et al.[24] in 112 adolescent athletes found that those who slept less than 8 hours per night
were 1.7 times more likely to have had an injury on average, when compared with those
who slept more than 8 hours per night, which highlights the need for this group to
obtain the recommended 8 to 10 hours.[24]
[25]
[26]
Adolescent athletes that participate in certain sports may be further implicated in
the challenges involved in obtaining adequate sleep, such as those where athletes
are often required to train both before and after school. These sports typically involve
large training volumes, such as swimming. Two previous studies by Steenekamp et al.,[13] and Gudmundsdottir[12] measured the sleep duration of adolescent swimmers using actigraphy, and found that
total sleep duration prior to early morning training (6:44 and 5:21 hour:min, respectively)
was well below the recommended 8 to 10 hours of sleep for adolescents.[12]
[13] Furthermore, Steenekamp et al.,[13] suggested that athletes did not compensate for early morning training sessions by
going to bed earlier the night before, and rather, their bedtime remained constant
regardless of training times the following day. Both Steenekamp et al.[13] and Gudmundsdottir[12] studied well-trained adolescent swimmers, but little is known about the performance
level above these cohorts – those that represent their country or are amongst top
placings in their age-group. Indeed, the increased training demands of the next tier
of competition may further exacerbate any sleep issues.
Another area that has not previously been researched in adolescent athletes is their
ability to self-predict sleep duration. Previous research has shown that elite adult
athletes overestimate their total sleep time by 19.8 minutes compared with actigraphy
monitor results.[27] Furthermore, Short et al.[28] reported that adolescent high school students significantly overestimate their perceived
sleep duration (8:17 hour:min) compared with their actigraphy-scored sleep (6:52 hour:min).
However, to the authors' knowledge, predicted versus measured sleep has not been investigated
in youth athletes as of yet.
While there has been an increase in sleep research occurring in the elite adult athlete
population over the past decade,[29] there is a paucity of research on high-performance adolescent athletes (15–18-years-old).
Given the phase delay compared with that of adults, combined with training loads and
extracurricular activities that adolescent athletes can experience, it is vital that
we understand their sleeping patterns to allow for adequate recovery for health and
performance. Therefore, the present study aims to assess the sleeping patterns in
highly trained adolescent aged swimmers across a 2-week training phase, specifically
comparing sleep on nights preceding an early morning training, daytime training, and
rest days. Our secondary aim was to compare the adolescent athletes' perceived and
measured sleep duration. Furthermore, the study's final aim was to determine if there
is a correlation between device usage prior to sleep onset and the effect on different
sleep measures. In accordance with previous literature, we hypothesize that this study
group's sleep will be impaired on the nights preceding early morning training when
compared with day training or rest days. Furthermore, they will overestimate their
perceived total sleep time compared with their measured results. Finally, we hypothesize
that adolescents' electronic device usage will have a negative relationship with their
sleep indices.
Materials and Methods
Participants
A total of 15 adolescent swimmers (15–18-years-old) volunteered to participate in
the current study (mean ± standard deviation [SD]; age 16.4 ± 1.0 years, 12 females/3
males). Participants were recruited through the local and national swimming organizations
in New Zealand. To be included in the current study, all participants had to attend
high school and either have represented their country or placed top three at a national
event in the year prior to recruitment. Based on the recommendations by Lakens,[30] after most of or the entire target population is measured, there is no need to perform
a sample size calculation. Therefore, due to sampling almost the entire population
available (∼30 national level adolescent swimmers representing their country), the
calculation was deemed unnecessary in this instance. Study approval was obtained through
the University of Waikato's Human Research Ethics Committee. All participants provided
informed written consent before taking part in this study and, for participants under
16 years, consent was also provided by a parent/caregiver.
Design
In this crossectional study, participants completed four validated sleep questionnaires
via an electronic survey link (Survey Monkey, CA, USA). Upon completion of the questionnaires,
participants wore a wrist-actigraphy device for a 2-week (14-day) period to monitor
sleep. This period was completed during a normal, in-season training phase, where
participants had approximately 11 training sessions per week on average, of which
7 were pool sessions and around 4 were gym sessions. All participants took part in
the study sometime in the up to 10 weeks leading up to their next major competition,
but not in the final 2 weeks prior to competition, when they were tapering. Therefore,
no participant competed during the monitoring period. The research was also completed
during a regular school term, in the participants' own home environment. Each morning
throughout the 2-weeks, participants were asked to fill out a subjective sleep diary,
and at the conclusion of each day they were asked to fill out a training diary.
Classifications of Nights
Each night was coded based on the training schedule of the following day. Therefore,
nights of sleep were placed into three categories dependent on the schedule for the
preceding day; early morning training (EARLY) which consisted of any training commencing
prior to 7am, daytime training (DAY) which consisted of training commencing any time
after 7am, and no training (REST) which consisted of a full day rest with no training
sessions. These classifications have been reported previously.[12]
[13]
Measures
Sleep Questionnaires
The sleep questionnaire contained personal characteristic questions, as well as four
validated and common sleep questionnaires; the Pittsburgh sleep quality index (PSQI),
the sleep hygiene index (SHI), the Epworth sleepiness scale (ESS), and the athlete
sleep behavior questionnaire (ASBQ).
The PSQI assesses sleep quality and disturbances over a 1-month period to give a global
score relating to overall sleep quality.[31] Global scores range from 0 to 21, with higher scores indicating worse sleep quality.
It is suggested that a global score of > 5 equates to severe sleep difficulties in
at least two areas or moderate sleep difficulties in more than three areas (out of
seven component areas) of the questionnaire.
The SHI is a 13-questions-long, self-administered questionnaire that assesses sleep
behavior and habits thought to compromise sleep hygiene.[32] Participants are asked to indicate how frequently they engage in specific behaviors
on a five-point rating scale ranging from 0 (never) to 4 (always). Item scores are
then summed providing a global score (0 to 52) for sleep hygiene, with a higher score
representing poorer sleep hygiene status. The SHI has been shown to be both valid
and reliable in a healthy population.[32]
The ESS provides a measurement for general daytime sleepiness rated on a scale of
0 to 3 for eight everyday activities.[33] Scoring ranges between 0 and 24, with numbers greater than 16 indicating high levels
of daytime sleepiness, those between 10 and 15 indicating abnormal daytime sleepiness,
and those from 0 to 10 indicating normal daytime sleepiness. This test is commonly
used to differentiate between individuals with and without sleep disorders and has
also been shown to correlate with objective measures of sleepiness.[34]
The ASBQ is an 18-item questionnaire about sleep habits and behaviors thought to be
of common concern for elite athletes.[35] It asks participants how frequently specific behaviors occur; never, rarely, sometimes,
frequently, or always (e.g., I go to bed with sore muscles; I use stimulants when I train/compete; I think, plan,
and worry about my sporting performance when I am in bed). Each response is weighted from 1 = never to 5 = always. A global score is produced
at the end of the questionnaire by summing the answers from each of the 18 items.
Global scores can range from 18 to 90, and it is suggested that a global score of
≤ 36 would equate to good sleep behaviors, with a score of ≥ 42 equating to poor sleep
behaviors.[35] The ASBQ has been reported to have acceptable reliability (intraclass correlation = 0.87)
when retested and has moderate to large correlations with validated questionnaires
such as the sleep hygiene index (SHI), the Epworth sleepiness scale (ESS) and the
PSQI. Furthermore, the ASBQ had acceptable levels of internal consistency in the sampled
population (Cronbach α = 0.73).
Sleep Monitoring
Participants were required to wear a wrist-actigraphy device (Fatigue Science, Readiband,
Vancouver, Canada) over a 2-week period to monitor and objectively quantify sleep
patterns. The sleep indices obtained from the actigraph were: total sleep time (TST,
h:min), total time in bed (TTB, h:min), sleep latency (SL, min), wake episodes per
night (WE, number), wake after sleep onset (WASO, min), sleep efficiency (SE, %),
sleep onset time (SOT, time of day), and wake time (WT, time of day). The raw activity
scores were translated to sleep-wake scores based on computerized scoring algorithms.[36] Participants were instructed to wear the actigraph device on the wrist they felt
most comfortable with,[37] continuously for the 2-week monitoring period, with the exception of time spent
in pool training sessions and showering. Sleep indices were quantified via the Fatigue
Science (Readiband, Vancouver, Canada) software at a sampling rate of 16 Hz. The Readiband
devices used in the current study have shown high levels on intra-device reliability,[38] and have been validated against the gold standard PSG with accuracies of approximately
90% for TST.[39]
Sleep and Training Diaries
For the duration of the sleep monitoring period, participants were also required to
fill out a daily sleep and training diary. The sleep diary consisted of five subjective
questions to estimate how long participants used an electronic light-emitting device
in the 2 hours prior to bedtime, how long it took them to fall asleep, how many times
they woke during the night, how long they slept for, and to rate their quality of
sleep on a scale of 1 to 5 (1 = very poor, 5 = very good). Similarly, the daily training
diary asked participants to outline how many training sessions they had per day, as
well as the time, type, and duration of each session.
Statistical Analysis
Simple group and descriptive statistics are reported as means ± SDs unless stated
otherwise. A linear mixed model (LMM) was conducted to compare types of night (EARLY,
DAY, and REST) for changes in all objective sleep measures. EARLY was used as the
reference level. The LMM included a fixed effect of night and random intercept for
participant.
A repeated measures standardised mean difference (SMD) was calculated by dividing
the mean difference (in raw units) of interest by the average within-subject SD (i.e.,
random effects residual error term). Statistical significance for fixed effects was
determined using F tests with the degrees of freedom for F statistics computed using
the Satterthwaite approximation method. Bonferroni corrections were made for post-hoc
analysis between nights to reduce the likelihood of type-1 error. Magnitudes of the
standardized effects between EARLY, DAY, and REST for all sleep measures were also
analyzed using Cohen's d and interpreted using thresholds of 0.2, 0.5, and 0.8 (small, moderate, and large, respectively).[40] An effect size of < 0.2 was considered to be trivial, and the effect was deemed unclear if its 95% confidence interval overlapped the thresholds for small positive and negative effects. Intraclass correlation coefficients (ICC's) were also
used to compare conditions for sleep metrics and the Pearson product-moment correlation
was run to assess the relationship between device usage and all sleep indices. The
magnitude of correlation between device usage and the sleep indices was assessed using
the following thresholds: 0.00–.19, very weak; 0.20–.39, weak; 0.40–.59, moderate; 0.60–.79, strong; and 0.80–1.0, very strong.[41] Analyses were performed using Statistical Package for Social Science (IBM Corp.
Armonk, NY, USA) version 22.0, and Jamovi, version 2.3.18.0, with statistical significance
set at p ≤ 0.05.
Results
The characteristics and measurements of the 2-week monitoring period are presented
in [Table 1]. Participants averaged 7.1 pool training sessions and 3.5 gym sessions per week.
Table 1
Mean ± standard deviation values for the measured sleep questionnaires and training
load of the two-week monitoring period.
|
Mean ± SD
|
Sleep questionnaires
|
|
Pittsburgh sleep quality index
(range 0–21)
> 5 poor (n = 6)
≤ 5 adequate (n = 9)
|
5.2 ± 1.4 (poor)
|
Epworth sleepiness scale
(range 0–24)
>16 high (n = 1)
10–16 abnormal (n = 7)
0–10 normal (n = 7)
|
8.5 ± 4.0 (normal)
|
Sleep hygiene index
(range 0–52, with higher score indicating poorer sleep hygiene status)
|
18.1 ± 6.1
|
Athlete sleep behavior questionnaire
(range 18–90)
≤36 good (n = 5)
≥42 poor (n = 3)
|
36.1 ± 5.3
|
Training load, per week
|
|
Total training time
(h:min)
|
16:35 ± 3:30
|
Pool training sessions
(No.)
|
7.1 ± 1.5
|
Gym training sessions
(No.)
|
3.5 ± 1.5
|
Abbreviations: SD, standard deviation.
Type of Day
The dataset was distributed across the three types of nights as EARLY (n = 102 observations), DAY (n = 49 observations), and REST (n = 26 observations). The values for the comparison of variables between EARLY, DAY,
and REST can be observed in [Tables 2] and [3].
Table 2
Mean ± standard deviation values for the measured objective sleep variables and device
usage on all days, the night preceding an early morning training session, a day training
session, and a rest day.
|
ALL
|
EARLY
|
DAY
|
REST
|
Correlation (r)
|
Sleep index
|
|
|
|
|
|
Total time in bed
(h:min)
|
8:00 ± 0:30
|
6:55 ± 0:50#,^
|
9:06 ± 1:01
|
9:43 ± 1:41
|
0.14
|
Total sleep time
(h:min)
|
6:40 ± 0:46
|
5:53 ± 1:06#,^
|
7:40 ± 1:12
|
7:59 ± 1:19
|
0.17
|
Sleep efficiency
(%)
|
84.3 ± 7.1
|
85.2 ± 7.0
|
84.2 ± 9.8
|
82.9 ± 6.9
|
0.23
|
Sleep latency
(min)
|
27.3 ± 13.2
|
26.5 ± 13.4
|
28.5 ± 18.4
|
24.5 ± 12.9
|
-0.12
|
Wake episodes per night
(No.)
|
3.8 ± 1.7
|
3.0 ± 1.8^
|
4.3 ± 2.5^
|
5.8 ± 2.0#
|
-0.32
|
Wake after sleep onset
(min)
|
36.7 ± 30.5
|
25.3 ± 22.6^
|
43.7 ± 42.3
|
61.9 ± 43.2#
|
-0.18
|
Sleep onset time
(time of day)
|
21:54 ± 0:34
|
21:38 ± 0:49^
|
21:49 ± 0:49^
|
22:36 ± 1:01
|
0.06
|
Wake time
(time of day)
|
5:40 ± 0:30
|
4:20 ± 0:27#,^
|
6:40 ± 0:41^
|
8:01 ± 1:16#
|
0.33
|
Subjective sleep quality
(1–5 scale)
|
3.63 ± 0.28
|
3.65 ± 0.28
|
3.70 ± 0.46
|
3.68 ± 0.66
|
0.19
|
Device usage (min)
|
63.7 ± 27.9
|
59.7 ± 27.7
|
65.1 ± 30.1
|
62.8 ± 40.4
|
n/a
|
Training times (time of day)
|
n/a
|
5:23 ± 0:46
|
13:27 ± 3:25
|
n/a
|
n/a
|
Notes: Comparison of electronic device usage to sleep indices on ALL nights using the Pearson
moment correlation (r). #Significantly different to DAY (p < 0.05). ^Significantly different to REST (p < 0.05).
Table 3
Mean ± standard deviation data for differences between nights for objective sleep
indices, including raw differences between conditions and effect sizes (d) with 95% confidence limits (± 95% CL).
Sleep index
|
EARLY – DAY (effect size)
|
EARLY – REST (effect size)
|
DAY – REST (effect size)
|
Total time in bed
(h:min)
|
-2:11 ± 0:11#
d = -2.61 ± 0.80
Large
|
-2:48 ± 0:51#
d = -2.73 ± 1.17
Large
|
-0:37 ± 0:40
d = 0.09 ± 0.73
Unclear
|
Total sleep time
(h:min)
|
-1:47 ± 0:06#
d = -1.68 ± 0.54
Large
|
-2:06 ± 0:13#
d = -2.02 ± 0.68
Large
|
-0:19 ± 0:07
d = 0.29 ± 0.41
Unclear
|
Sleep efficiency
(%)
|
1.0 ± 2.8
d = -0.09 ± 0.46
Unclear
|
2.3 ± 0.1
d = -0.16 ± 0.53
Unclear
|
1.3 ± 2.9
d = -0.05 ± 0.56
Unclear
|
Sleep latency
(min)
|
-2.0 ± 5.0
d = 0.12 ± 0.69
Unclear
|
2.0 ± 0.5
d = -0.22 ± 0.44
Unclear
|
4.0 ± 5.5
d = -0.25 ± 0.60
Unclear
|
Wake episodes per night
(No.)
|
-1.3 ± 0.7
d = 1.06 ± 0.61
Large
|
-2.8 ± 0.2#
d = 1.37 ± 0.59
Large
|
1.5 ± 0.5#
d = 0.20 ± 0.55
Unclear
|
Wake after sleep onset
(min)
|
-18.4 ± 19.7
d = 1.11 ± 0.74
Large
|
-36.6 ± 20.6#
d = 1.22 ± 0.56
Large
|
18.2 ± 0.9
d = 0.06 ± 0.56
Unclear
|
Sleep onset time
(h:min)
|
0:11 ± 0:00
d = 0.13 ± 0.58
Unclear
|
0:58 ± 0:12#
d = 1.15 ± 0.73
Large
|
0:47 ± 0:12#
d = 1.02 ± 0.83
Large
|
Wake time
(h:min)
|
2:20 ± 0:14#
d = 4.43 ± 0.63
Large
|
3:41 ± 0:49#
d = 7.78 ± 1.57
Large
|
1:21 ± 0:35#
d = 2.21 ± 1.02
Large
|
Notes:
#Significant difference between nights (p < 0.05).
The LMM identified a main effect of day (F[2,28] = 22.2, p < 0.001), for TST, with substantial reductions observed between EARLY and DAY (-1:47 ± 0:06 hour:min,
d = -1.68 ± 0.54, p < 0.01) and between EARLY and REST (-2:06 ± 0:13 hour:min, d = -2.02 ± 0.68, p < 0.01).
The LMM identified a main effect of day (F(2,28) = 26.8, p < 0.001), for TTB, indicating that it was different from baseline values, with post
hoc analysis revealing reductions between EARLY and DAY (-2:11 ± 0:11 hour:min, d = -2.61 ± 0.80, p < 0.01) and between EARLY and REST (-2:48 ± 0:51 hour:min, d = -2.73 ± 1.17, p < 0.01).
The LMM identified a main effect of day (F(2,28) = 7.42, p = 0.003) with a significant difference for SOT between EARLY and REST (0:58 ± 0:12 hour:min,
d = 1.15 ± 0.73, p < 0.05) and DAY and REST (0:47 ± 0:12 hour:min, d = 1.02 ± 0.83, p < 0.05).
The LMM also identified a main effect of day (F(2,28) = 11.7, p < 0.001) for WE, between EARLY and REST (-2.8 ± 0.2, d = 1.37 ± 0.59, p < 0.05) and DAY and REST (1.5 ± 0.5, d = 0.20 ± 0.55).
The LMM also identified a main effect of day (F(2,28) = 5.99, p = 0.007) for WASO between EARLY and REST (-36.6 ± 20.6, d = 1.22 ± 0.56, [Table 3 & 4]). There were no significant differences observed in SE, SL, or subjective sleep
quality for comparison between nights ([Table 2]).
Subjective versus Objective Measures
There were significant differences observed for all three sleep indices (SL, WE, and
TST) when comparing participants' objective and subjective sleep diary data from the
2-week monitoring period ([Table 4]). Participants significantly underestimated their subjective SL compared with actigraphic
SL (17.5 ± 5.6 minutes and 27.3 ± 13.2 minutes, respectively, d = 0.97 ± 0.55, p < 0.05) and WE (0.8 ± 0.6 and 3.7 ± 1.8, respectively, d = 1.75 ± 0.63, p < 0.05). As hypothesized, participants significantly overestimated their subjective
TST compared with objective results (7:42 ± 0:42 hour:min and 6:40 ± 0:42 hour:min,
respectively, d = -1.00 ± 0.68, p < 0.05, [Table 5]).
Table 4
Linear mixed model (LMM) fixed and random effect parameters for the model examining
the mean difference in objective sleep indices between nights, with p-values, standardized mean differences (SMD), and intraclass correlation coefficients
(ICC).
|
TTB (h:min)
|
TST (h:min)
|
SE (%)
|
SL (mins)
|
WE (No.)
|
WASO (mins)
|
Fixed effects
|
Estimates
|
p value
|
SMD
|
Estimates
|
p value
|
SMD
|
Estimates
|
p value
|
SMD
|
Estimates
|
p value
|
SMD
|
Estimates
|
p value
|
SMD
|
Estimates
|
p value
|
SMD
|
Intercept (EARLY)
|
6:55
|
< 0.001
|
|
5:53
|
< 0.001
|
|
85.2
|
< 0.001
|
|
26.5
|
< 0.001
|
|
3.0
|
< 0.001
|
|
25.3
|
< 0.001
|
|
DAY a
|
2:11
|
< 0.001
|
1.98
|
1:47
|
< 0.001
|
1.92
|
-1.0
|
0.60
|
-0.19
|
2.0
|
0.65
|
0.17
|
1.3
|
0.04
|
0.80
|
18.4
|
0.09
|
0.63
|
REST a
|
2:48
|
< 0.001
|
2.54
|
2:06
|
< 0.001
|
2.25
|
-2.3
|
0.24
|
-0.44
|
-2.0
|
0.66
|
-0.16
|
2.3
|
< 0.001
|
1.77
|
36.6
|
0.00
|
1.26
|
Random effects
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Between subjects SD
|
0:35
|
|
|
0:44
|
|
|
6.1
|
|
|
9.2
|
|
|
1.4
|
|
|
23.4
|
|
|
Within subjects SD
|
1.06
|
|
|
0:56
|
|
|
5.3
|
|
|
12.0
|
|
|
1.6
|
|
|
29.0
|
|
|
ICC
|
0.21
|
|
|
0.38
|
|
|
0.57
|
|
|
0.37
|
|
|
0.44
|
|
|
0.40
|
|
|
Abbreviations: TTB, total time in bed; TST, total sleep time; SE, sleep efficiency; SL, sleep latency;
WE, wake episodes; WASO, wake after sleep onset. Notes:
a reference – EARLY (pre 7am).
Table 5
Mean ± SD data for differences between subjective and objective sleep for objective
sleep indices, including effect sizes (d) for comparison between conditions.
Sleep index
|
Subjective sleep
|
Objective sleep
|
Objective / subjective sleep
difference
(raw)
|
Objective sleep - subjective sleep
(effect size ± 90% CL)
|
Sleep latency
(min)
|
17.5 ± 5.6
|
27.3 ± 13.2
|
9.8 ± 7.6
#
|
0.97 ± 0.55
Moderate
|
Wake episodes per night
(No.)
|
0.8 ± 0.6
|
3.8 ± 1.7
|
3.0 ± 1.1
#
|
1.75 ± 0.63
Large
|
Total sleep time
(h:min)
|
7:42 ± 0:54
|
6:40 ± 0:46
|
-0:58 ± -0:08
#
|
-1.00 ± 0.68
Moderate
|
Abbreviations: SD, standard deviation. Notes:
#Significant difference between subjective and objective measures of sleep (p < 0.05).
Device Usage
Electronic light-emitting devices were used on average for 63.7 ± 27.9 minutes in
the 2 hours prior to bedtime ([Tables 1] and [2]). There was no significant difference between EARLY, DAY, and REST conditions for
device use, with average use being 59.7, 65.1, and 62.8 minutes in the 2 hours preceding
bedtime, respectively. The Pearson product-moment correlation was run to assess the
relationship between device usage and sleep indices (TTB, TST, SE, SL, WE, WASO, SOT,
and WT). There were no statistically significant correlations observed between device
usage and sleep indices (p > 0.05), and all relationships were considered weak – very weak.
Discussion
The primary aim of this study was to investigate the sleeping patterns of highly trained
adolescent swimmers over a 2-week period. In this study, participants averaged 6:40 hour:min
of total sleep duration over the monitoring period, which falls below the recommended
sleep duration of 8 to 10 hours sleep for adolescents.[26] Furthermore, differences in sleep indices were observed between the three types
of nights (EARLY, DAY, and REST), with TST significantly reduced on nights preceding
EARLY training compared with both DAY training and REST by 1:47 and 2:06 hour:min,
respectively.
When comparing to objective measures, adolescent swimmers in this study objectively
overestimated their TST by approximately 1-hour per night, while also substantially
underestimating SL and WE when compared with their objective sleep measures determined
by actigraphy. The final aim of this study was to determine if there was a correlation
between device usage prior to sleep onset and the effect on different sleep measures.
No correlation was observed for any of the sleep indices in adolescent swimmers.
As reported by Steenekamp et al.,[13] sleep duration was significantly reduced on nights preceding early morning training
(6:44 hour:min) compared with nights with no early morning training session (8:45 hour:min)
in 32 adolescent swimming and rowing athletes.[13] A further study by Gudmundsdottir[12] reported sleep duration was significantly shorter on nights preceding early morning
training (5:21 hour:min) compared with later morning training (6:37 hour:min) and
no morning training (6:53 hour:min) in 108 Icelandic adolescent swimmers.[12] The results of the current study are consistent with previous findings,[12]
[13] with TST significantly reduced on the night prior to EARLY training sessions (5:53 hour:min)
compared with DAY training sessions (7:40 hour:min) and REST (7:59 hour:min). However,
unlike in previous studies, it appears that the participating athletes adjusted their
bedtimes to offset the earlier wake times due to the early morning training, as SOT
was significantly earlier on nights preceding early morning and day training compared
with rest days.
Adolescent athletes in the current study had significantly less WASO on EARLY training
days (25.3 minute) compared with REST (61.9 minute), which is a similar finding to
the Gudmundsdottir[12] study, where WASO was higher when there was no morning training on the next day.
Given TTB and TST are significantly lower on nights preceding EARLY training days,
it is not surprising that WASO and WE would also be significantly reduced. Given the
importance of sleep on athletes' physical and cognitive recovery,[8] there is clearly a need for further research in this area. Furthermore, the increased
risk of injury in this population when less than 8 hours of sleep per night is attained[24] emphasizes the importance of adolescent athletes regularly obtaining the recommended
hours of sleep. Therefore, practitioners and coaches working with this population
should ensure that scheduling of training sessions allows for adequate sleep, or that
athletes compensate for early training times by going to bed earlier.
Mean scores from the PSQI placed adolescent swimmers in the “poor” sleep category.
However, with an average score of 5.2 and only 6 athletes scoring >5, classifying
this population group as poor sleepers may be somewhat harsh, as it has been suggested
that appropriate cut-off scores for poor sleep in adolescents may be >8.[42]
Additionally, results from the ESS suggest that adolescent athletes have normal amounts
of daytime sleepiness (mean global score of 8.5). The SHI results suggest that adolescent
athletes in the current study displayed better sleep behaviors than a similar aged
cohort,[43] with a mean global score of 18.1. These results contrast what Setyowati et al.,[43] found in 101 Indonesian adolescent school students, who scored 32 for the SHI on
average. A potential reason for the difference in results observed between the two
studies could be the differing school times, with a 7am start and 1pm finish in Indonesia,
compared with approximately 8:40am and 3:20pm in New Zealand. Furthermore, as participants
in the current study were athletes, it is possible that they had some basic understanding
of the importance of sleep (e.g., from athlete education sessions) in comparison to
the participants of Setyowati et al.,[43] which had a nonathlete cohort.
Results from the ASBQ also showed that adolescent athletes in the current study may
have better sleep behaviors (mean global score = 36.1) than adult individual and team
sport elite athletes, as reported previously (44.3 and 42.6, respectively).[27] Adolescent athletes may report having better sleep behaviors than elite ones on
the ASBQ due to the relevance of certain questions. For example, questions relating
to alcohol, travel, foreign sleeping environments, sleeping pills, and use of stimulants
scored extremely low in adolescent athletes, which may offer a reason why this group
scored lower than elite athletes on the questionnaire. We suggest that a modified
version of the ASBQ, specifically for adolescent athletes might be required.
To our knowledge, the current study is the first to examine the difference between
subjective and objective sleep measures in adolescent athletes. In elite adult athletes,
the use of subjective self-perceived sleep measures has been shown to be in better
agreement with objective actigraphy-derived results, with differences in TST being
of approximately 20 minutes.[28] The current study indicates that adolescent swimmers significantly overestimate
TST (by ∼54 minutes) and underestimate SL and WE, suggesting that there may be a greater
deficit between subjective and objective sleep measures in this group.
There is a possibility that the adolescent athlete population in the current study
could have inflated their sleep measures in their sleep diaries, for fear of parent/caregivers
reviewing their sleep metrics, which should be considered in future research. However,
these findings are similar to previous research in adolescents in the general population
who also significantly overestimate their perceived TST when compared with objective
sleep measures.[44]
[45] The current study's results suggest that objective sleep measures should be used
whenever possible in this cohort of athletes. However, it is important to note that
in most adolescent athlete settings, objective sleep measures will not be readily
available and could be cost prohibitive.
Research from the current study suggests that adolescent athletes' sleep is disrupted
by early morning training, leading to the accumulation of sleep debt throughout the
week, and resulting in them trying to 'catch up' on rest days. Although not all rest
days occurred during weekends, a majority of them did, which aligns with previous
research that highlights the significant change in adolescent's sleeping patterns
between weekdays and weekends.[46]
[47] These patterns lead to disruption in the circadian timekeeping system.[48] Similarly, Steenekamp et al.[13] found that adolescent athletes extended their sleep on weekends without training
when compared with weekdays with early morning training.
While it is evident that changes in scheduling of training for adolescent athletes
could result in improved sleeping patterns, it is not always practical, and coaches
are often unwilling to accommodate. Therefore, the implementation of sleep hygiene
education and strategies[49]
[50] is often the best option for adolescent athletes to understand the importance of
sleep, as well as to learn tools to improve it. Furthermore, educating coaches on
sleep hygiene may also be beneficial so they have the understanding on why training
schedules need to change for adolescent athletes. Based on the results from the current
study, a target for sleep education may be electronic device usage at night.
Electronic light-emitting devices were used by adolescent athletes in this study on
average for 1h per night in the 2-hours prior to bedtime. It is known that increased
device usage can be associated with shortened sleep duration (11–24 minutes).[21]
[51] Interestingly, no correlations were found for device usage and sleep indices (TTB,
TST, SE, SL, WE, WASO, SOT, and WT) in the current study. Perhaps differences would
have been found if some of the participants abstained from any device usage in the
lead-up to bedtime. However, this was not the case, as all athletes used their devices
for at least 30 minutes in the 2-hours prior to bed. Despite the results in the current
study, research that has focused specifically on adolescents has found that a 1 and
2-hour exposure to light from self-luminous devices (computers, tablets, and cell
phones) suppresses melatonin in adolescents by 23 and 28%, respectively.[20] Much like adolescents, adults who use electronic light-emitting devices before bed
had their evening levels of melatonin suppressed by 20%.[52]
As this was a field-based study, there were multiple variables (diet, caffeine intake,
environment, and school scheduling) that we were unable to control for, which could
have had an impact on the overall results of the study. A further limitation was that
female menstruation cycles were not controlled for in this study, nor was any data
collected on this variable. We would suggest that this be considered in future research
in female adolescent athletes.
Furthermore, we did not collect data on napping habits despite the actigraph used
in the current study being able to detect them. Therefore, we cannot rule out the
potential that short naps may have taken place (e.g., on the weekend) without being
detected. Future studies should address napping habits, as they may be used to supplement
night time sleep.
Lastly, although the current study had a relatively small sample size (n = 15), the ability to recruit highly trained adolescent swimmers in New Zealand is
somewhat limited. Indeed, the overall population to draw from that fits the inclusion
criteria is small (n = ∼30). Therefore, we believe that our sample is representative of this cohort. However,
given this study took place in only one country (New Zealand), the results may not
be representative of elite adolescent swimmers in different countries due to various
contextual, cultural, and environmental factors.
Conclusion
In summary, the results of the current study showed that, across a 2-week monitoring
period, early morning training resulted in a reduced sleep duration (∼5h 53mins) and
time in bed in adolescent swimmers. Additionally, adolescent athletes significantly
overestimate their sleep duration by 1 hour and underestimate their time to fall asleep
by 10 minutes, approximately. Interestingly, there were no links found between electronic
device use at night and any of the sleep measures.
The findings of the current study show the importance of coaches, parents, and school
administrators being aware of the sleeping habits of adolescent athletes and the subsequent
impact that early morning training sessions have on their sleep. Whenever possible,
adjusting schedules so they can better balance their schedules around training and
school requirements has the potential to benefit their sleep, reduce injury risk,
and improve academic performance. Further research is required to investigate interventions
that may enhance the sleep/wake behavior of adolescent athletes to better support
them.