In search of lost sleep: Secular trends in the sleep time of school-aged children and adolescents

In search of lost sleep: Secular trends in the sleep time of school-aged children and adolescents

Sleep Medicine Reviews 16 (2012) 203e211 Contents lists available at ScienceDirect Sleep Medicine Reviews journal homepage: www.elsevier.com/locate/...

743KB Sizes 17 Downloads 66 Views

Sleep Medicine Reviews 16 (2012) 203e211

Contents lists available at ScienceDirect

Sleep Medicine Reviews journal homepage: www.elsevier.com/locate/smrv

CLINICAL REVIEW

In search of lost sleep: Secular trends in the sleep time of school-aged children and adolescents Lisa Matricciani a, c, *, Timothy Olds a, b, c, John Petkov a a

Health and Use of Time (HUT) Group, University of South Australia, GPO Box 2471, Adelaide SA 5000, Australia Sansom Institute for Health Research, University of South Australia, GPO Box 2471, Adelaide SA 5000, Australia c School of Health Sciences, University of South Australia, GPO Box 2471, Adelaide SA 5000, Australia b

a r t i c l e i n f o

s u m m a r y

Article history: Received 7 January 2011 Received in revised form 18 March 2011 Accepted 18 March 2011 Available online 25 May 2011

Background: Sleep deficits are associated with a wide range of detrimental physical and mental health outcomes. There is concern that children are not getting enough sleep, and that sleep duration has been declining. However, evidence is sparse. Methods: A systematic review of world literature was conducted to locate studies reporting the sleep duration of children aged 5e18 years. Monte Carlo simulation was used to generate pseudodata from summary data, which were combined with raw data and analysed by linear regression of sleep duration on year of measurement at the age  sex  day type  country level. Results: Data were available on 690,747 children from 20 countries, dating from 1905 to 2008. From these data, 641 regressions were derived. The sample-weighted median rate of change was 0.75 min nightly per year, indicating a decrease of more than 1 h per night over the study period. Rates of change were negative across age, sex and day type categories, but varied according to region, with Europe, the USA, Canada and Asia showing decreases and Australia, the UK and Scandinavia showing increases. Conclusion: Over the last 103 years, there have been consistent rapid declines in the sleep duration of children and adolescents. Ó 2011 Elsevier Ltd. All rights reserved.

Keywords: Sleep duration Children Adolescents Trends

Introduction Adequate sleep is important for the growth, maturation and health of children and insufficient sleep has been associated with an array of physical and psychosocial health deficits.1 These include an impaired ability to concentrate2,3 and retain information,4,5 mood disorders including anxiety, depression and hyperactivity4,6 as well as impaired motor skills7 and poorer overall health and immune function.8 Inadequate sleep has also been associated with impaired academic performance,2 an increased risk of injuries and accidents,9 suicide ideation10 and drug and alcohol use.11 It has been suggested that short sleep duration increases the risk of obesity by increasing sympathetic activity, elevating cortisol and ghrelin levels, decreasing leptin levels and/or impairing glucose tolerance.12,13 Both cross-sectional and prospective cohort studies on children support this association.14e16 In light of this connection, together with the growing prevalence of obesity worldwide,17 * Corresponding author. Health and Use of Time (HUT) Group, University of South Australia, GPO Box 2471, Adelaide SA 5000, Australia. Tel.: þ61 8 830 21741; fax: þ61 8 830 26558. E-mail address: [email protected] (L. Matricciani). 1087-0792/$ e see front matter Ó 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.smrv.2011.03.005

secular declines in sleep duration have been linked to the secular rise in obesity.16,17 The notion that children are sleeping less than they used to is widespread both in the scientific literature and the popular media, and is “moving out of the bedroom, home and neighbourhood, and into the courts, boardrooms and even parliament”.18 This secular decline, variously ascribed to electrification,14,19 increased use of technology,20,21 and modern lifestyle,15,22,23 is believed to have resulted in many children not getting enough sleep.24e26 Declines in children’s sleep duration over recent decades have been reported in the scholarly literature. Dollman and colleagues27 noted a 30 min decline between the years 1985 and 2004 for 10e15 year-old Australian children on school nights while Iglowstein et al.28 reported declines for 1e16 year-old Swiss children between 1974 and 1993. Similar trends have also been reported for children from Japan29 and Iceland.30 In spite of such findings, there is also evidence to suggest that that the sleep duration of children has not declined over the years. Hofferth and Sandberg31 reported an increase in the sleep duration of 3e12 year-old American children between the years 1981 and 1997, while Huysmans et al.32 reported no change for 15-18 yearold children in the Netherlands between 1980 and 2000.

204

L. Matricciani et al. / Sleep Medicine Reviews 16 (2012) 203e211

Pääkönen33 found that children in Finland were getting more sleep on weekends and less sleep on weekdays in 2000 compared with 1987. Randler,34 on the other hand, noted that the sleep duration of German children had increased from 1907 up until the 1970s at which time a decline commenced. The belief that children are getting less sleep than they should, and that this deficit is due to “modernism”, has persisted for at least a century. Given the conflicting evidence around secular trends in children’s sleep and the apparent scarcity of scientific evidence,35 we have undertaken a comprehensive review of historical studies of children’s sleep durations in order to chart changes in sleep duration over the last century.

Methods Data location A systematic literature review was conducted to locate relevant data. A pilot search determined the scope and relevance of candidate databases. The Scopus, EbscoHost, Ovid and Web of Science platforms were searched. Table 1 shows the search terms used. Studies returned from the database search were initially included and read in full if the abstract stated that the self- or proxy-reported sleep duration of children was measured. Papers were excluded if the abstract explicitly stated that the children were unhealthy, were younger than five years of age or older than 18 years of age (not inclusive) or that the sleep duration was measured by polysomnography or actigraphy. In cases where the exclusion criteria were not clearly stated, the paper was kept and read in full. The reference lists of relevant articles were pearled for further studies, and experts in the field were contacted. Studies were analysed to identify multiple publications of the same dataset, and the authors contacted if it was unclear whether duplication had occurred. Where two articles referred to the same dataset, the article with the most detailed summary detail was included for analysis. In addition to searches for summary data (data expressed in the form of means and standard deviations), strategies to retrieve raw datasets were also employed. Specifically, all located studies were screened for reference to large national studies for which raw data could potentially be retrieved and three electronic time use data archives36e38 (websites that provide raw datasets from studies that have been conducted to assess time use variables) were systematically searched. In addition, any national statistic office or time use study referred to within these databases was emailed with the request for both raw and summary data of interest. To retrieve data from the electronic time use data archives searched, appropriate navigation methods were followed and data were downloaded using the instructions offered by the website. If the data archive did not appear to contain any data of interest, they were emailed for confirmation.37,38 All data that were downloaded and retrieved for this study36 were interpreted and decoded with

appropriate codebook. Where coding was unclear, the organisation was contacted for clarification. All decoding of data was then checked by an independent reviewer. If any discrepancies were identified between the two reviewers and could not be resolved, the data were excluded from analysis. Data were also excluded from analysis in cases where data were missing (e.g., the child’s sex). Data extraction All summary data collected were extracted from hard copy form and recorded on an electronic spreadsheet. All data entered were cross-checked with original hard copies and then randomly checked for errors. Data collected from each study included year of measurement, sex, age, country of residence, sleep duration, day type, sample size and methodology. In cases where such data were missing or unclear, efforts were made to retrieve data by contacting the authors. The following principles guided data extraction: 1) Year of measurement was expressed as the midpoint of the measurement period if stated, or, if not stated, was estimated from an equation based on those studies for which year of measurement was reported. The equation was:

Year of Measurement ¼ 35:5 þ 0:981 Year of Publication ðn ¼ 97 studies; r ¼ 0:996; p ¼ 0:001; SEE ¼ 2:24Þ 2) Sex was “boy”, “girl” or “both sexes”. In cases where “both” was reported, the sample consisted of an approximately equal number of males and females. 3) Age was taken as the reported mean age. If the mean age was not reported, the midpoint of the age range provided was recorded. In cases where the study did not report the child’s age, the study was not included for analysis. Weighted age values were not calculated because Monte Carlo simulation was used. However, in cases where the average sleep duration of children of different ages was provided, the weighted mean age was recorded (if possible). For example, if the average sleep duration of children aged 10, 11 and 12 years old were recorded and the study specified that 60, 40 and 50 children from each age group was examined (respectively), the age recorded was 10.9 years. 4) If the country in which data were collected was not stated, the first author’s affiliation was used as a guide, in addition to any clues in the text. 5) Sleep duration was operationalised as much as possible as the self- or proxy-reported time in bed (TIB), taken to mean the elapsed time between turning out the lights to go to sleep and waking in the morning. Total sleep time (TST) (i.e., time in bed minus latencies) was only taken in cases where TIB was not reported. In many cases, particularly with older studies, explicit definitions of sleep were not included. In these cases, assumptions were made based on what was reported. While studies reporting both total sleep time (TST) (i.e., time in bed

Table 1 Search strategy used for each database. Database

Date

Limitation

Search Terms

Scopus

19/12/09

Abstract

Ovid

20/12/09

Web of Science

21/12/09

Multi-field search Topic

EbscoHost

21/12/09

Abstract

((ABS(“sleep time” OR “sleep duration” OR “sleep quantity” OR “time spent asleep”) OR ABS(“time spent sleeping” OR “time sleeping” OR “time asleep” OR “sleep length”))) AND ((ABS(child* OR teen* OR youth* OR boy* OR girl*) OR ABS(adolescen* OR student*))). ((sleep time or sleep duration or sleep quantity or time spent asleep or time spent sleeping or time sleeping or time asleep or sleep length) and (child$ or teen$ or youth$ or boy$ or girl$ or adolescen$ or student$) Topic¼(“sleep time” or “sleep duration” or “sleep quantity” or “time spent asleep” or “time spent sleeping” or “time asleep” or “sleep length”) AND Topic¼(child* or teen* or youth* or boy* or girl* or adolescen* or studen*) AB (sleep time or sleep duration or sleep quantity or time spent asleep or time spent sleeping or time sleeping or time asleep or sleep length) and AB (child* or teen* or youth* or boy* or girl* or adolescen* or student*)

L. Matricciani et al. / Sleep Medicine Reviews 16 (2012) 203e211

minus latencies) and time in bed were included, when regressions were calculated they involved only estimates using the same operationalisation. 6) Day type was categorised as “schooldays” (usually Monday to Friday nights), “non-schooldays” (Saturday and Sunday nights) or “Weekly average”. Where day type was not specifically stated, it was assumed that the data referred to the weekly average. In cases where sleep was measured on a specific day Monday, Tuesday, Wednesday, Thursday, Friday were considered “schooldays” and Saturday and Sunday were considered “nonschooldays”. Holidays were classified as “non-schooldays”. 7) The methodology involved in sleep data collection was also recorded, specifically whether data were obtained via diary, questionnaire or interview methods or whether data were selfor proxy-reported. Data combination Data were available both as raw datasets and as summary data (data expressed as means and standard deviations). To combine datasets, Monte Carlo simulation was used to generate pseudodata. This technique attempts to “recreate” the unavailable raw data by using a random normal generator to produce data points based on reported means and standard deviations (SDs). It assumes that the original distributions were approximately normal, which was true of the available raw datasets. Pseudodata were combined with raw data to form a combined database for analysis. This approach has been used previously in similar research cumulating studies of children’s fitness39 as well as sleep.40 Where SDs were not presented in the studies identified and unavailable from authors, they were estimated based on data from studies where they were reported. These SDs were significantly related to day type, age and year of measurement, but not to sex or country. The following equations were developed to estimate missing SDs:

School days ¼ 1289:8 þ 2:20Age þ 0:660Year ðn ¼ 318; r ¼ 0:40; p < 0:0001Þ Non  school days ¼ 1728:1 þ 4:47Age þ 0:881Year ðn ¼ 268; r ¼ 0:51; p < 0:0001Þ Weekly average ¼ 717:7 þ 0:96Age þ 0:384Year ðn ¼ 239; r ¼ 0:34; p < 0:0001Þ Trend analysis Where sufficient data existed, simple linear regression was used to analyse trends, at the age  sex  country  day type level (for example, 10 year-old Swedish boys on schooldays). Rates of change were expressed as minutes per day per year. Linear regressions were only determined for studies that defined sleep in the same manner (i.e., TIB or TST). To compare rates of change across age groups, sexes, countries and day types, sample-weighted ANOVAs were used with the regression slopes as the dependent variable and age groups (5e8, 9e12, 13e15 and 16e18 years), sex (boys, girls, both sexes), day types (schooldays, non-schooldays, weekly average), and regions (Asia, Australia, Canada, Europe, Scandinavia, UK, USA) as the grouping factors. In addition, regressions were grouped into four time periods [Very early (1900e1939); Early (1940e1969); Late (1970e1989); Very Late (1990e2008)] according to the midpoint of the years covered by the regression, and rates of change were compared across time periods. Alpha was set at 0.05 for all analyses.

205

Results Data collected Search strategies employed in this study identified a total of 218 studies (Fig. 1) and 44 raw datasets. Sleep duration data were available for 822,105 children from 35 different countries at the age  sex  country  day type level. These data were collected between 1892 and 2008. Data included for trend analysis From the data collected, it was possible to derive 641 regressions at the age  sex  country  day type level, encompassing 690,747 children from 20 countries (Australia, Belgium, Brazil, Canada, China, Denmark, Egypt, Finland, France, Germany, Italy, Japan, the Netherlands, New Zealand, Norway, Poland, Spain, Switzerland, UK and USA). Twelve of these are European countries and eight Pacific Rim countries. Sixteen are classified by the World Bank41 as high income and four as low to middle income (Brazil, China, Egypt and Poland). Summary data were available for all countries while raw data were available for Australia, Canada, Denmark, Germany, Italy, the Netherlands, Norway, Spain, the UK and the USA only. Data were collected between 1905 and 2008. Regression analyses Of the 641 regressions, 371 (58%) were negative (decreases in sleep duration) or zero. Of these, 191 (52%) were statistically significant. The remaining 270 (42%) regression slopes were positive (increases in sleep duration), of which 110 (41%) were statistically significant. Of the 641 regressions, 594 assessed TIB and 47 assessed TST. Table 2 summarises the rates of change in sleep duration at the age  sex  country  day type level. Although sample-weighted means and medians for rates of change are presented, we will focus on the weighted median values which best represent real changes because they do not assume a normal distribution of rates of change. Across all 641 regressions, the sample-weighted median rate of change was 0.75 (interquartile range 1.19 to þ0.16) minutes per day per year. Fig. 2 shows funnel plots of rates of change from the 641 regressions against the span of years for each regression and sample size. For small spans and sample sizes, the rates of change are very labile, but for large spans and sample sizes they gravitate towards the sample-weighted median value (0.75 min/year). Secular trends according to age Decreases in sleep time occurred across all age groups (Fig. 3a). There was a general trend for secular decline to be greater among older children, with median changes ranging from 0.41 min per day per year for early primary school children to 0.91 min per day per year for older adolescents. These rates were significantly different for all age categories (p < 0.0001), with the exception of 13e15 and 16e18 year-old age groups. Secular trends according to sex As shown in Fig. 3b, rates of change were similar for boys (0.77 min per year), girls (0.67) and both sexes (0.72). Significant differences were found between all groups, but the difference was small (Effect size (ES) ¼ w0.1). Secular trends according to day type Rates of change were negative for all day types (Fig. 3c), with the rate of decline on schooldays (0.74 min per year) exceeding that of

206

L. Matricciani et al. / Sleep Medicine Reviews 16 (2012) 203e211

Fig. 1. PRISMA flowchart for the search.

non-schooldays (0.30). The weekly average rate of change was 0.77 min per year. Significant differences were seen across all day type groups (p < 0.0001). Secular trends according to geographical region As shown in Fig. 4, rates of change varied according to geographical region. Positive rates of change were found for Scandinavia, the UK and Australia, while negative rates of change were

found for Asia, Canada, the USA and Europe. Significant differences were found for all regional groups (p < 0.0001). Secular trends according to time period Significant differences were identified across the different time periods indicating that the overall rate of decline is not an artefact of year. As shown in Fig. 5, the rates of change were negative for all time periods, except for the “very early” period, which was positive.

L. Matricciani et al. / Sleep Medicine Reviews 16 (2012) 203e211

207

Table 2 Rates of change (minutes per day per year) in sleep duration according to sex, age, day type and geographical location.

All Analysis by age Age 5e8 Age 9e12 Age 13e15 Age 16e18 Analysis by sex Boys Girls Both sexes Analysis by day type Schooldays Non-schooldays Weekly average Analysis by region Asia Australia Canada Europea Scandinavia UK USA Analysis by time periodb Very Early (1900e1939) Early (1940e1969) Late (1970e1989) Very Late (1990e2008)

k

n

Mean

SD

Median

IQR

641

690,747

0.46

2.84

0.75

1.19 to þ0.16

90 165 172 214

49,090 149,709 331,993 159,955

0.13 0.26 0.56 0.54

2.83 3.24 3.01 2.10

0.41 0.57 0.86 0.91

1.38 0.67 1.20 1.23

184 184 273

121,003 123,068 446,676

0.40 0.17 0.57

2.34 2.25 3.12

0.77 0.67 0.72

0.91 to 0.18 0.94 to þ0.25 1.35 to þ0.09

263 234 144

217,371 121,093 352,283

0.79 0.97 0.17

2.49 4.05 2.50

0.74 0.30 0.77

1.42 to þ0.44 2.15 to þ0.73 0.89 to 0.63

63 42 26 245 46 97 120

261,770 30,072 6670 142,897 33,834 66,470 142,428

0.50 þ1.27 0.73 0.92 0.00 þ0.57 0.53

2.56 2.10 1.39 2.74 1.52 1.48 3.54

0.77 þ0.40 1.10 1.17 þ0.65 þ0.57 1.14

0.59 þ0.07 2.16 2.13 1.35 0.29 1.46

to to to to to to to

0.89 þ1.55 þ0.10 0.07 þ0.67 þ1.35 þ0.11

16 75 143 407

9460 276,423 84,779 320,085

þ2.57 0.79 0.48 0.27

1.55 0.39 1.34 4.04

þ2.39 0.78 0.35 0.40

þ1.84 0.87 1.35 2.00

to to to to

þ2.61 0.67 þ0.28 þ1.39

to to to to

þ0.22 þ0.17 0.09 þ0.21

Significant differences were found across age groups (with the exception of 13e15 and 16e18 year-old age categories), sexes, regions and between different day types (P < 0.05). k ¼ number of regressions assessed; n ¼ sample size; SD ¼ standard deviation; IQR ¼ interquartile range. a Other than Scandinavia and the United Kingdom. b Calculated according to the midpoint of the years covered by the regression.

Discussion Main findings This study identified a secular decline of 0.75 min per year in children’s sleep duration over the last 100 years. Significant differences were identified across most age groups, between sexes and regions, and on the different day types. The greatest rate of decline in sleep occurred for older children, boys and on schooldays. Regional analyses indicated secular declines in Asia, Canada

the USA and Europe, while increases were identified in Australia and in the other parts of Europe (the UK and Scandinavia). Europe (excluding the UK and Scandinavia) was found to have experienced the greatest rate of decline. Strengths and limitations This is the largest and most comprehensive study to examine secular trends in children’s sleep duration to date. As previously identified,35 while notions of a secular decline in children’s sleep

Fig. 2. Funnel plots of changes in sleep duration (Y-axis, min/year) against the span of years for each regression, and the total sample size for each regression (X-axes). The dashed line is the sample-weighted median rate of change (0.75 min/year).

208

L. Matricciani et al. / Sleep Medicine Reviews 16 (2012) 203e211

Fig. 3. Box plots showing sample-weighted rates of change for age (Fig. 3a), sex (Fig. 3b) and day type (Fig. 3c) sub-groups. The dashed line is the sample-weighted median rate of change (0.75 min/year). k ¼ number of regressions assessed; SD ¼ standard deviation; IQR ¼ interquartile range.

duration are widespread, the evidence to support such claims is scarce and conflicting. Unlike previous studies,35 which have examined relatively small and selective samples of children, this study combined raw and summary data from a large number of sources, without language restrictions, to determine trends for each age  sex  country  day type slice. Therefore, this study provides the best estimate of secular trends in children’s sleep duration to date. This is of particular importance given that secular trends in

children’s sleep duration have been used to strengthen arguments to change public policies, guidelines and practices. Furthermore, unlike previous studies, which have generally examined the sleep of children from two points in time, this study analysed the sleep of children at multiple time points, minimising artefactual volatility. This study also examines longer year spans, reducing the risk of statistical volatility. Additionally, this study identified sleep trends for 690,747 children from 20 different countries, some of which have never been discussed previously. Such a large sample size adds power to this study. Lastly, major variables known to affect sleep (i.e., age, sex, methodology, country and day type) have all been taken into account, an analysis that has not been conducted previously. There are several limitations of this study that also need to be addressed. In particular, this study combines data from a large number of sources to determine secular trends in children’s sleep duration. In any attempt to collate a large number of datasets which use different methodologies, and across a wide temporal and geographical span, a number of assumptions and approximations must be made. While these assumptions and approximations increase statistical “noise” and are limitations of this study, they do not, by themselves, bias trend estimates. Although this study assessed secular trends according to the definition of sleep (TIB or TST), it did not assess trends according to the method of reporting (i.e., self- or proxy-reported). However, available studies comparing self- and proxy-reported sleep times suggest that the two methods appear to yield similar results.42,43 Consequently, there is no reason to believe that selfand proxy-reported sleep times are different. Furthermore, analysis revealed that only 32 of the 641regressions consisted of a mixture of self- and proxy-reported sleep times. Consequently, even if later studies show that these methods yield different estimates of sleep duration, it is unlikely that the results of this study are artefacts of mixing self- and proxy-report. While the majority of regressions consisted of homogeneous methods of reporting, virtually all 5e8 year-old children’s sleep was proxyreported, while virtually all 13e15 and 16e18 year-old children’s sleep was self-reported. There were an approximately equal number of self- and proxy-reported sleep times for the 9e12 year-old age group. While objective measures of sleep, such as polysomnography and actigraphy are preferred methods of measuring sleep, these techniques have only recently been developed and have been used on relatively small samples. In contrast, self- and proxy-reported sleep durations have been reported for children since the 1890s on populations as large as 4000.44 While self- and proxy-reported sleep times are usually overestimated in nonclinical children when compared to objective measures, studies3,45e48 have shown a good correlation for both questionnaires (r ¼ 0.60e0.78) and diaries (r ¼ 0.97) when compared with objective measures in the adolescent and child population. Given that these measures are consistently correlated with criterion measures, they are suitable for trend analysis. Another limitation of this study that needs to be addressed is the categorisation of day type. While it would have been ideal to have classified “schooldays” as Sunday-Thursday nights and “non-schooldays” as Friday- Saturday nights, only 14 of the studies included for analysis explicitly stated that day type was classified in this manner. Given that the majority of studies defined “schooldays” as Monday-Friday nights and “non-schooldays” as Saturday-Sunday nights, it was decided that “schooldays” would refer to Monday-Friday while “non-schooldays” would refer to Saturday-Sunday nights. Although this decision was undesirable and adds statistical “noise”, it is unlikely to bias trend estimates.

L. Matricciani et al. / Sleep Medicine Reviews 16 (2012) 203e211

209

Fig. 4. Box plots showing sample-weighted rates of change for different regions. The dashed line is the sample-weighted median rate of change (0.75 min/year). k ¼ number of regressions assessed; SD ¼ standard deviation; IQR ¼ interquartile range.

Previous literature

Cause of secular trends

Although the results of this study support popular notions of a secular decline in sleep,35 the rate of change is less than the popularly reported (1.526e2.049 min of sleep per year). Furthermore, this study identified significant differences across age groups, between sexes and regions, and on schooldays and non-schooldays, considerations rarely acknowledged by publications containing claims.

While this study does not determine possible causes for trends identified, it has been speculated that secular declines in children’s sleep duration have occurred as a result of progressive delays in children’s bedtimes, but unchanged wake times. Dollman and colleagues,27 for instance, found that although the bedtimes of 10e15 year-old Australian children on schooldays in 2004 were later than they were in 1985, children’s wake times had not

Fig. 5. Box plots showing sample-weighted rates of change for different year periods. The dashed line is the sample-weighted median rate of change (0.75 min/year). k ¼ number of regressions assessed; SD ¼ standard deviation; IQR ¼ interquartile range.

210

L. Matricciani et al. / Sleep Medicine Reviews 16 (2012) 203e211

changed. Similarly, Iglowstein and colleagues,28 in examining three birth cohorts (1974, 1979, 1986) of 1e16 year-old Swiss children, found that although the bedtimes had gotten later, wake times had not changed. Consistent with these findings, later bedtimes have also been reported for children living in Finland,33 Germany34 and Iceland.30 Delays in children’s bedtimes may be attributed to activities that keep children awake and contexts that allow them to do so. Activities such as technology use, schoolwork and part-time employment have all been associated with delayed bedtimes and reduced sleep duration.50e52 Similarly, electrification, caffeine use and parental attitudes30 have been identified as contexts that allow children to have later bedtimes.19,26,53 Implications Secular delays in children’s bedtime and declines in children’s sleep are often put forward as evidence that many children today do not get enough sleep52,54e57 and are chronically sleep deprived.20,25 However, we should also consider the hypothesis that children used to sleep more than they needed, or that sleep is discretionary and a specific amount is not required.58 Indeed, despite concerns that too many children do not get enough sleep, whether children need a specific amount of sleep is controversial and has not yet been established, despite the existence of sleep guidelines. Although there is a lack of consensus58,59 regarding what constitutes “adequate” sleep and whether children are in need of more sleep, the results of this study are concerning in light of evidence suggesting that short sleep duration is associated with multiple social, physical and mental health deficits. However, it is difficult to determine whether negative health consequences have arisen as a result of a reduction of more than 1 h of sleep over the last 103 years. Before such conclusions can be drawn, further research is required to quantify the deficit associated with sleep decrements.

Practice Points 1) There has been a secular decline of 0.75 min per year in children’s sleep duration over the last century. 2) Significant differences exist across age groups, sexes, regions and between different day types. 3) It is possible that children today are more sleepdeprived than their parents or grandparents were as children.

Research agenda 1) What are the health implications of decreasing sleep amongst children? 2) How much sleep do children and adolescents actually need? 3) Why has sleep duration declined, and why are there differences across age groups, sexes, regions and different day types?

Financial support This study received no financial support.

References 1. Eisenmann J. Insight into the causes of the recent secular trend in pediatric obesity: common sense does not always prevail for complex, multi-factorial phenotypes. Preventive Medicine 2006;2(5):329e35. 2. Wolfson A, Carskadon M. Sleep schedules and daytime functioning in adolescents. Child Development 1998;69(4):875e87. 3. Wolfson AR, Carskadon MA, Acebo C, Seifer R, Fallone GP, Labyak SE, et al. Evidence for the validity of a sleep habits survey for adolescents. Sleep 2003;26(2):213e6. 4. Steenari M, Vuontela V, Paavonen J, Carlson S, Fjällberg M, Aronen E. Working memory and sleep in 6-to 13-year-old schoolchildren. Journal of the American Academy of Child and Adolescent Psychiatry 2003;42(1):85e92. 5. Walker M, Stickgold R. Sleep, memory, and plasticity. Annual Review of Psychology 2006;57(1):139e66. 6. Blunden S, Hoban TF, Chervin RD. Sleepiness in children. Sleep Medicine Clinics 2006;1(1):105e18. 7. Kuriyama K, Stickgold R, Walker MP. Sleep-dependent learning and motor-skill complexity. Learning and Memory 2004;11(6):705e13. 8. Sekine M, Chandola T, Martikainen P, Marmot M, Kagamimori S. Work and family characteristics as determinants of socioeconomic and sex inequalities in sleep: the Japanese civil servants study. Sleep 2006;1(29):2. 9. Koulouglioti C, Cole R, Kitzman H. Inadequate sleep and unintentional injuries in young children. Public Health Nursing 2008;25(2):106e14. 10. Liu X. Sleep and adolescent suicidal behavior. Sleep 2004;27(7):1351e8. 11. Wong M, Brower K, Fitzgerald H, Zucker R. Sleep problems in early childhood and early onset of alcohol and other drug use in adolescence. Alcoholism, Clinical and Experimental Research 2004;28(4):578e87. 12. Spiegel K, Leproult R, Van Cauter E. Impact of sleep debt on metabolic and endocrine function. Lancet 1999;354(9188):1435e9. 13. Spiegel K, Leproult R, L’hermite-Balériaux M, Copinschi G, Penev P, Van Cauter E. Leptin levels are dependent on sleep duration: relationships with sympathovagal balance, carbohydrate regulation, cortisol, and thyrotropin. The Journal of Clinical Endocrinology and Metabolism 2004;89(11):5762e71. 14. Gangwisch J, Malaspina D, Boden-Albala B, Heymsfield S. Inadequate sleep as a risk factor for obesity: analyses of the NHANES I. Sleep 2005;28(10):1289e96. 15. Cappuccio F, Taggart F, Kandala N, Currie A, Piele E, Stranges S, et al. Metaanalysis of short sleep duration and obesity in children and adults. Sleep 2008;31(5):619e26. 16. Patel S, Hu F. Short sleep duration and weight gain: a systematic review. Obesity 2008;16(3):643e53. 17. Van Cauter E, Spiegel K, Tasali E, Leproult R. Metabolic consequences of sleep and sleep loss. Sleep Medicine 2008;9(1 Suppl.):23e8. 18. Williams S. Sleep and society: sociological ventures into the (un)known. London: Routledge Taylor & Francis Group; 2005. 19. Cizza G, Skarulis M, Mignot E. A link between short sleep and obesity: building the evidence for causation. Sleep 2005;28(10):1217e20. 20. Taheri S. The link between short sleep duration and obesity: we should recommend more sleep to prevent obesity. Archives of Disease in Childhood 2006;91(11):881e4. 21. Pearson H. Sleep it off. Nature 2006;443(2):261e3. 22. Spiegel K, Knutson K, Leproult R, Tasali E, Van Cauter E. Sleep loss: a novel risk factor for insulin resistance and Type 2 diabetes. Journal of Applied Physiology 2005;99(5):2008e19. 23. Smedje H. Trends in the duration of school-day sleep among 10-to 15-year-old South Australians between 1985 and 2004. Acta Paediatrica 2008;96(7):954e5. 24. Curcio G, Ferrara M, De Gennaro L. Sleep loss, learning capacity and academic performance. Sleep Medicine Reviews 2006;10(5):323e37. 25. Loessl B, Valerius G, Kopasz M, Hornyak MD, Riemann D, Voderholzer U. Are adolescents chronically sleep-deprived? An investigation of sleep habits of adolescents in the Southwest of Germany. Child: Care, Health and Development 2008;34(5):549e56. 26. Calamaro C, Mason T, Ratcliffe S. Adolescents living the 24/7 lifestyle: effects of caffeine and technology on sleep duration and daytime functioning. Pediatrics 2009;123(6):1005e10. *27. Dollman J, Ridley K, Olds T, Lowe E. Trends in the duration of school-day sleep among 10- to 15-year-old South Australians between 1985 and 2004. Acta Paediatrica 2007;96(7):1011e4. *28. Iglowstein I, Jenni O, Molinari L, Largo R. Sleep duration from infancy to adolescence: reference values and generational trends. Pediatrics 2003;111(2):302e7. 29. Kohyama J. Sleep duration. Shounika 2005;46(1. Suppl.):88e9. 30. Thorleifsdottir B, Björnsson J, Benediktsdottir B, Gislason T, Kristbjarnarson H. Sleep and sleep habits from childhood to young adulthood over a 10-year period. Journal of Psychosomatic Research 2002;53(1):529e37. *31. Hofferth S, Sandberg J. Changes in American children’s time, 1981e1997. In: Owens T, Hofferth S, editors. Children at the millennium: where have we come from, where are we going?; 2001. p. 193e229. *32. Huysmans F, Zeijl E, van den Broek A. Adolescents’ leisure and well-being in the Netherlands: trends and correlates. Society and Leisure 2005;28(2):531e48.

Disclosures The authors declare no conflict of interest.

* The most important references are denoted by an asterisk.

L. Matricciani et al. / Sleep Medicine Reviews 16 (2012) 203e211 33. Pääkkönen H. What do school children in Finland do with their time? Loisir et Société 2005;28(2):425e9. *34. Randler C. Sleep length in German children and adolescents: comparing 1907 with 2006e2008. Somnologie: Schlafforschung und Schlafmedizin 2009;13(2):89e91. *35. Matricciani L, Olds T, Williams M. A review of evidence for the claim that children are sleeping less than in the past. Sleep, in press. 36. Center for Time Use Research. Center for time use research. viewed 10 May 2010. Available from: http://www.timeuse.org/mtus/download/; 2010. 37. UK data archives. UK data archive. viewed 12 September 2009. Available from: http://www.data-archive.ac.uk/home; 2008. 38. Harmonised European Time Use Surveys. HETUS. viewed 12 February 2010. Available from: https://www.h2.scb.se/tus/tus/Default.htm; 2009. *39. Tomkinson GR, Léger L, Olds T, Cazorla G. Secular trends in the fitness of children and adolescents 1980-2000-an analysis of 20 m shuttle run studies. Sports Med 2003;33(4):385e400. *40. Olds T, Blunden S, Petkov J, Forchino F. The relationship between sex, age, geography and time in bed in adolescents: a meta-analysis of data from 23 countries. Sleep Medicine Reviews 2010;14(6):371e8. 41. World Bank. Country and lending groups. viewed 10 June. Available from: http:// data.worldbank.org/about/country-classifications/country-and-lending-groups; 2010. 42. Csikszentmihalyi M, Graef R. Socialization into sleep: exploratory findings. Merrill-Palmer Quarterly 1975;21(1):3e18. 43. Beebe D, Lewin D, Zeller M, McCabe M, MacLeod K, Daniels S, et al. Sleep in overweight adolescents: shorter sleep, poorer sleep quality, sleepiness, and sleep-disordered breathing. Journal of Pediatric Psychology 2007;32(1):69e79. 44. Hertel N. Overpressure in high schools in Denmark. California: Macmillan; 1885. 45. Bauer K, Blunden SL. How accurate is subjective reporting of childhood sleep patterns: a review of the literature and implications for practice. Current Pediatric Reviews 2008;4(2):132e42. 46. Acebo C, Sadeh A, Seifer R, Tzischinsky O, Wolfson A, Hafer A, et al. Estimating sleep patterns with activity monitoring in children and adolescents: how many nights are necessary for reliable measures? Sleep 1999;22(1):95e103.

211

47. Gaina A, Sekine M, Chen X, Hamanishi S, Kagamimori S. Validity of child sleep diary questionnaire among junior high school children. Journal of Epidemiology 2004;14(1):1e4. 48. Lockley SW, Brainard GC, Czeisler CA. High sensitivity of the human circadian melatonin rhythm to resetting by short wavelength light. Journal of Clinical Endocrinology and Metabolism 2003;88(9):4502e5. 49. Shinomiya H, Takeuchi H, Martoni M, Natale V, Harada T. Comparative study on circadian typology of Japanese and Italian students aged 12-18 years. Sleep and Biological Rhythms 2004;2(1):93e5. 50. Van den Bulck J. Television viewing, computer game playing, and Internet use and self-reported time to bed and time out of bed in secondary-school children. Sleep 2004;27(1):101e4. 51. Van den Bulck J. Text messaging as a cause of sleep interruption in adolescents, evidence from a cross-sectional study. Journal of Sleep Research 2003;12(3):263. 52. Carskadon M. Patterns of sleepiness in adolescents. Paediatrician 1990;17(1):5e12. 53. de Sousa I, Arau cjo J, De Azevedo C. The effect of a sleep hygiene education program on the sleep-wake cycle of Brazilian adolescent students. Sleep and Biological Rhythms 2007;5(4):251e8. 54. Fredriksen K, Rhodes J, Reddy R, Way N. Sleepless in Chicago: tracking the effects of adolescent sleep loss during the middle school years. Child Development 2004;75(1):84e95. 55. Spilsbury J, Drotar D, Rosen C, Redline S. The Cleveland adolescent sleepiness questionnaire: a new measure to assess excessive daytime sleepiness in adolescents. Journal of Clinical Sleep Medicine 2007;3(6):603e12. 56. Walsh J, Dement W, Dinge D. Sleep medicine, public policy, and public health. In: Kryger M, Roth T, Dement W, editors. Principles and practice of sleep medicine. Philadelphia: Elsevier/Saunders; 2005. p. 648e56. 57. Yu Y, Lu B, Wang B, Wang H, Yang J, Li Z, et al. Short sleep duration and adiposity in Chinese adolescents. Sleep 2007;30(12):1688e97. *58. Horne J. Sleepfaring: a journey through the science of sleep. New York: Oxford University Press; 2006. *59. Harrison Y, Horne J. Should we be taking more sleep? Sleep 1995;18(10):901e7.