Bidirectional Relationship of Sleep with Emotional and Behavioral Difficulties : A Five-year Follow-up of Finnish Adolescents

Journal of Youth and Adolescence https://doi.org/10.1007/s10964-020-01203-3 EMPIRICAL RESEARCH Bidirectional Relationship of Sleep with Emotional and Behavioral Difficulties: A Five-year Follow-up of Finnish Adolescents Laura Kortesoja1,2 ●Mari-Pauliina Vainikainen1,2 ●Risto Hotulainen1●Arja Rimpelä3,4 ●Henrik Dobewall3● Pirjo Lindfors3●Sakari Karvonen5●Ilona Merikanto 6,7,8 Received: 18 November 2019 / Accepted: 24 January 2020 © The Author(s) 2020 Abstract The long-term effects of sleep on adolescent psychosocial well-being are mostly unknown, although insufficient sleep has been associated with emotional and behavioral difficulties in cross-sectional studies. With a five-year follow-up of Finnish adolescents (Time 1: n=8834; Mean age =13 years, 51.1% female, Time 2: n=5315, Mean age =15 years, 51.6% female, Time 3: n=3712; Mean age =17 years; 50.2% female), the purpose of this longitudinal study was to investigate the relations between self-reported sleep duration, sleep problems, and emotional and behavioral difficulties during adolescence. Emotional and behavioral difficulties were assessed using The Strengths and Difficulties Questionnaire (SDQ) measuring emotional symptoms, conduct problems, hyperactivity, peer problems and total difficulties. Sleep duration was calculated by counting the hours between self-reported bedtime and wake-up time. Sleep problems were assessed with a single question about the general sleep problems. According to the cross-lagged models for sleep and emotional and behavioral difficulties, the findings of this study indicate a developmental process during adolescence where, firstly, short sleep duration is a stronger predictor for current and prospective emotional and behavioral difficulties than vice versa. Secondly, increased emotional and behavioral difficulties expose adolescents to current and later sleep problems more strongly than reverse. Thus, the results show that short sleep duration predisposed to emotional and behavioral difficulties across adolescence, which then led to more prospective sleep problems. These findings suggest a developmental process where sleep and emotional and behavioral difficulties are intertwined in shaping adolescents’health. Keywords Adolescence ●Longitudinal study ●Emotional and behavioral difficulties ●Sleep duration ●Sleep problems Introduction Adolescence is a developmental period in life during which many biological, psychological, and social factors interact (Nelson et al. 2005). Changes in sleep (Carskadon 2011) and in mental health (Keyes 2006) during this period are reflected in adolescent’s well-being as increased propensity to depressive symptoms, problems in psychosocial functioning, and conduct problems. These aforementioned problems are classified as emotional and behavioral difficulties in childhood and adolescence (World Health Organization 2015), which can be divided to internalizing problems (e.g. depression and anxiety) and externalizing problems (e.g. hyperactivity and peer/conduct problems) (Goodman et al. 2010). Both internalizing problems and externalizing problems are relatively frequent among adolescents, with prevalence estimates varying from 10% to 30% (Merikangas et al. 2010; Patel et al. 2007). Early emotional and behavioral difficulties may lower the quality *Ilona Merikanto ilona.merikanto@helsinki.fi 1Centre for Educational Assessment, University of Helsinki, Helsinki, Finland 2Faculty of Education and Culture, Tampere University, Tampere, Finland 3Faculty of Social Sciences, Unit of Health Sciences, Tampere University, Tampere, Finland 4Department of Adolescent Psychiatry, Pitkäniemi Hospital, Tampere University Hospital, Nokia, Tampere, Finland 5Social Policy Research Unit, Finnish Institute for Health and Welfare, Helsinki, Finland 6SleepWell Research Program, Faculty of Medicine, University of Helsinki, Helsinki, Finland 7Mental Health Unit, Finnish Institute for Health and Welfare, Helsinki, Finland 8Orton Orthopaedics Hospital, Helsinki, Finland 1234567890();,: 1234567890();,: of adolescent life and even have long-term effects, e.g. externalizing behavior problems may increase the risk for school dropout, family problems, and economic difficulties (Colman et al. 2009). The association between sleep and mental health in adolescence is well-established. For example, short sleep duration during adolescence is associated with depressed mood and behavioral problems (Wolfson and Carskadon 1998) and physical and emotional health problems (Owens et al. 2014). In addition, late adolescence is characterized by a circadian shift to eveningness (Roenneberg et al. 2007), which associates with increased risk for externalizing problem behavior (Merikanto et al. 2017). However, the complex associations between sleep and psychosocial development from childhood to late adolescence are still largely unknown. Understanding these associations is important as many mental disorders that are prevalent in adulthood (e.g. depression) begin to develop already during adolescence (Costello et al. 2011; Patel et al. 2007). Thus, longitudinal research is required to understand the complex interplay between sleep and mental health and their longterm effects on well-being. The present longitudinal study aims to answer these questions on how adolescent emotional and behavioral difficulties interact with sleep duration and sleep problems across adolescence. Sleep Functioning in Adolescence Both cross-sectional (Olds, Maher, Blunden, & Matricciani 2010) and longitudinal studies (Gradisar et al. 2011)have shown that sleep duration decreases during adolescence. Individual changes in sleep during childhood/adolescence may be hormonal and puberty-related (Carskadon et al. 1998) or environmentally induced, for example increased autonomy on bedtimes, heavier homework loads, working, sports, social and other activities, and use of electronic devices in adolescence (Harbard et al. 2016). Increasing autonomy and early school start time predispose adolescents to shortened sleep duration and sleep problems, such as sleep fragmentation, especially on school days (Lehto et al. 2016). As a result, this can increase daytime tiredness, depressive symptoms, and burnout, leading to poor daytime functioning (Pesonen et al. 2019). Teenage years are also characterized by a circadian shift to eveningness (Roenneberg et al. 2007), with a biological tendency towards later bedtime and the need to compensate this with a later wake-up time. Recently, it has been shown that chronotype, which refers to the genetic tendency towards morningness or eveningness, modifies the objective sleep timing already from childhood to late adolescence (Merikanto et al. 2018). It has also been postulated that sleep length has shortened over the last 100 years due to the cultural and environmental transitions (Keyes et al. 2015). Previous Studies on Association between Sleep and Emotional and Behavioral Difficulties during Adolescence The association of short sleep with problems in psychosocial functioning during adolescence is well established in cross-sectional studies concerning internalizing and externalizing behavior (Becker et al. 2015). Short sleep (Owens et al. 2014) and sleep problems (Tu et al. 2015) are both associated with internalizing and externalizing difficulties in adolescence. In a previous study, short sleep duration on school nights was associated with feelings of sadness and worthlessness, low motivation, anxiety, and thoughts of self-harm in youth (Yeo et al. 2019). Further, circadian preference for eveningness can lead to impaired daytime functioning (Wolfson and Carskadon 1998) and increased internalizing problems (Quach et al. 2018). Previous research has shown that the relation between eveningness and increased internalizing difficulties may be explained by shorter sleep duration and weakened sleep quality (Merikanto et al. 2017). In addition, later bedtimes during the school week are longitudinally associated with shorter sleep duration, worse educational outcomes and emotional distress (Asarnow et al. 2014). Previous studies suggest that later bedtimes and short sleepdurationcouldplay a role in the etiology of depression (Gangwisch et al. 2010) and lower self-esteem (Fredriksen et al. 2004). Short sleep duration may also increase the risk of adolescent anxiety disorders (Roberts and Duong 2017). Sleep problems were a significant predictor of later internalizing difficulties in a follow-up study of children from the age of 4 years until the age of 13 years, but the association was not significant in reverse (Quach et al. 2018). Compared to internalizing difficulties there are fewer studies about the role of sleep in adolescents’externalizing difficulties (Becker et al. 2015). Previous studies around the topic suggest that short sleep duration is associated with behavioral problems (Liu and Zhou 2002)anddifficulties in interpersonal relations, such as peer problems (Roberts et al. 2009). Furthermore, unhealthy sleep practices and sleep problems (e.g. insomnia) are also related to conduct problems and attention-deficit/hyperactivity disorder (Becker et al. 2015). Sleep problems are also associated with loneliness, especially in early and middle adolescents. Adolescent sleep problems may lead to dysfunction of working memory, which can further increase risky behavior in late adolescence (Thomas et al. 2015). The associations between sleep and psychosocial wellbeing are rather bidirectional (Pieters et al. 2015); psychosocial factors can influence sleep and sleep may influence psychosocial functioning (Becker et al. 2015). Psychosocial dysfunction can lead to insufficient sleep. For instance, Journal of Youth and Adolescence emotional and behavioral difficulties, especially hyperactivity, are associated with later delayed sleep phase in adolescence (Hysing et al. 2018) It is still largely unknown whether sleep is differentially associated with different dimensions of psychosocial difficulties and vice versa (Quach et al. 2018). Thus far, only few studies have investigated the bidirectional relationships between sleep and emotional and behavioral difficulties in a longitudinal design throughout adolescence, even though the need for longitudinal research has been established (Gregory and Sadeh 2012). What is more, most studies have examined shorter time lags between sleep and emotional and behavioral difficulties. However, there are previous studies that examine reciprocal relations between sleep and psychosocial functioning with multiple time points in adolescence. For instance, daily demands that cause stress are associated with receiving less sleep at night (Fuligni and Hardway 2006). Furthermore, short sleep duration was not associated with psychological well-being 6 or 12 months later, while there was evidence for the reverse relationship (Brand et al. 2014). Externalizing problem behavior is associated with later sleep problems during the elementary school transition period and vice versa (Quach et al. 2018). As noted above, adolescence is a critical developmental period with an increase of both sleep and mental health difficulties (Becker et al. 2015). Insufficient sleep puts youth at risk for a range of mental health difficulties. A longitudinal approach in later adolescence and early adulthood is required in order to elucidate the complex interplay between sleep and emotional and behavioral difficulties and their long-term effects on health prospects across critical transition periods. Gender Differences in Adolescent Emotional and Behavioral Difficulties and Sleep Previous research has shown differences in girls and boys concerning the quality and appearance of mental health problems (Wang et al. 2015). A recent research review revealed that internalizing problems may be on the rise in adolescent girls (Bor et al. 2014). Adolescent girls also report more emotional problems (Fink et al. 2015) and less peer and conduct problems than boys (Van Roy et al. 2006). In contrast, Mishina et al. (2018) found boys to have less peer problems compared to girls. Additionally, previous studies have demonstrated that gender might play a role in the development of sleep length; girls sleep less (Patte et al. 2018) and report more sleep problems than boys (Kechter and Leventhal 2018) during adolescence. This can be due to the developmental differences, since females sleep more than males in late adolescence and early adulthood (Maslowsky and Ozer 2014). However, more research is needed to understand the different developmental paths in girls and boys concerning the associations between sleep and psychosocial well-being. Current Study Using cross-lagged models, this study seeks to examine for the first time how adolescents’sleep duration and sleep problems predict later emotional and behavioral difficulties and vice versa from early to late adolescence within a single study. Thus far, no previous study has explored the longitudinal bidirectional links between emotional and behavioral difficulties and sleep with a follow-up from early to late adolescence. In addition, the present study aims at investigating the continuity of sleep duration, sleep problems and emotional and behavioral difficulties across adolescent years. Considering sleep functioning in adolescence, it is expected that sleep duration decreases throughout adolescence. Based on prior research, sleep problems and emotional and behavioral difficulties are hypothesized to increase from early to late adolescence leading to more emotional and behavioral problems in those with shorter sleep duration and more sleep problems as compared to adolescents with at least 8 h sleep duration per night and less sleep problems. Moreover, this study examines gender differences in sleep and emotional and behavioral difficulties. Based on previous findings concerning to gender differences, girls were expected to have shorter sleep duration, more sleep problems and emotional symptoms but less hyperactivity and conduct problems as compared to boys. Methods Participants and Procedure This study is conducted by Metropolitan Longitudinal Finland (MetLoFin) Study Group focusing on the development of adolescent learning and health. The total number of the age cohort was around 13,500 in 14 municipalities of the Helsinki metropolitan area, observed three times during the years 2011–2016. School data were collected by online classroom surveys, supervised by teachers. All adolescents in the target grade in the participating lower secondary schools of the Helsinki metropolitan area were invited to participate. The three time points were not spaced evenly apart since the study design was not planned primarily for studying sleep but the survey points were based on the transitions points in the Finnish school system. 7th grade was the beginning of the lower secondary school, 9th grade the last grade of the lower secondary school and the last Journal of Youth and Adolescence survey point was the second year of the upper secondary school. The original sample was collected in 2011 (T1), when the adolescents were in the seventh grade (12–13 years of age), and a total of 9723 adolescents (50.5% girls) out of the age cohort of around 13,500 students participated in the study. Two follow-ups were done: in 2014 at the end of the ninth grade (T2, 15–16 years of age) and in 2016 in the second year of upper secondary education (T3, 17–18 years of age). At T2, altogether 9359 adolescents (49.2% girls) and at T3, a total of 6547 adolescents (52.4% girls) participated in the MetLoFin study. The study flow is illustrated in Fig. 1. The analytic sample used here consists at T1 of 8743 participants (51.1% girls), at T2 of 5271 (51.6% girls) and at T3 of 3703 (50.2% girls) with information on sleep duration, sleep problems and psychosocial well-being based on The Strengths and Difficulties Questionnaire. The median ages for the analytic samples used here were 13.0 (SD =0.31) years at T1, 15.0 (SD =0.49) years at T2, and 17.0 (SD =0.37) years at T3. A total of 2479 participants (54.0% girls) provided data at every time point (T1, T2, and T3), i.e. around 18.4% of the age cohort. Adolescents who participated in the study at both T1 andT2(n=5271) or at both T1 and T3 (n=3703) were compared with adolescents who participated in the study only at T1 (n=2241). Participants who took part in the study only at T1 were more likely to have emotional and behavioral difficulties, to have shorter sleep duration, to be boys, and to have mothers with secondary education or lower (p< 0.001 for all independent samples t-tests). The Ethics Committee of the Finnish Institute for Health and Welfare approved the study design for each data collection. Also, the agency for education in each municipality approved the study every study year. Participation was voluntary for the adolescents. As the research was conducted as part of the school routine, information was collected only on students on the target grade who were present during the school day. Approximately 10% of students are known to be absent each day, some did not want to participate or parents had denied participation. All schools did not want to participate in the surveys or did not distribute the questionnaire to all classes. Parents were also given the right to refuse participation on behalf of their child. In two of 14 municipalities, local authorities required parental consent when the adolescents were 12–13 years of age. Measures Sleep duration Sleep duration was assessed using two questions on sleep and wake-up times. Question 1 asked participants “What time do you usually go to sleep during the school week?”The response scale consisted of seven options: 9 pm or earlier, around 9:30 pm, around 10 pm, around 10:30 pm, around 11 pm, around 11:30 pm, 12 pm or later. Question 2 asked participants “What time you usually wake up during the school week?”The response scale comprised seven options: 7 am or earlier, around 7:30 am, around 8 am, around 8:30 am, around 9 am, around 9:30 am, 10 am or later. Sleep duration was calculated by counting the hours between bedtime and wake-up time. The anchors of bedtime and wake-up time were used to determine estimated sleep duration. Sleep problems Sleep problems were assessed with a single question: “During the past six months have you had (and if so, how often) difficulty falling asleep or waking up at night?”. The scale contained four response options: seldom or never, approximately once a month, approximately once a week, almost every day. Fig. 1 Study flow in the three study years Journal of Youth and Adolescence Emotional and behavioral difficulties The Strengths and Difficulties Questionnaire (SDQ) was used for assessing emotional and behavioral difficulties (Goodman et al. 2010). The SDQ is a psychometric measurement tool for a child’s emotional and behavioral difficulties and has been validated for Finnish school-aged children and adolescents (Koskelainen and Kaljonen 2000). The questionnaire is widely used in the School Health Promotion Study nationwide in Finland. The SDQ includes five subscales that each comprise of five items (responses ranging from 0 to 2): emotional symptoms, conduct problems, hyperactivity, peer problems, and prosocial behavior. The SDQ can be used for measuring both internalizing and externalizing behaviors (Goodman et al. 2010). Total difficulties on psychosocial behavior The total difficulties on psychosocial behavior includes the first four subscales of The Strengths and Difficulties Questionnaire, excluding the prosocial subscale measuring positive and helpful behavior. Total difficulties score was calculated by summing the scores of the four subscales (range 0–40). The higher the sum score, more total difficulties there were. Cronbach’s alpha coefficients varied over time (α=0.80–0.82) for total difficulties. Emotional symptoms Emotional symptoms were assessed with items asking if participants had, for example, psychosomatic symptoms (i.e., headaches, belly pain), if they were worried, felt themselves unhappy or were scared of new situations. Emotional symptoms subscale was formed by summing the scores of these items (range 0–10). Cronbach’s alpha coefficients varied over time (α=0.72–0.75) for emotional symptoms. Conduct problems Conduct problems subscale measured tendency to lie, stealing things, agreeableness and loosing temper. Conduct problems scores were formed by summing item scores (range 0–10). Cronbach’s alpha coefficients varied over time (α=0.56–0.63) for conduct problems. Hyperactivity Hyperactivity subscale measured restlessness, concentrating, considering before taking action and completing tasks. Hyperactivity scores were formed by summing item scores (range 0–10). Cronbach’s alpha coefficients varied over time (α=0.73–0.75) for hyperactivity. Peer problems Peer problems subscale measured tendency to do things alone, being bullied, preferring adult companion and popularity of the peer group. Peer problems scores were formed by summing item scores (range 0–10). Cronbach’s alpha coefficients varied over time (α=0.54–0.56) for peer problems. For all subscales, higher score indicated more difficulties. Socioeconomic status (SES) Socioeconomic status (SES) was measured by mother’s educational level. The mother’s educational level was recorded as either secondary education or lower versus tertiary education. Gender Participants reported their gender as either male or female. Data Analyses The descriptive statistical analyses were performed using IBM SPSS (SPSS Inc., version 24.0). First, repeated measures ANOVAs were conducted to compare the differences in SDQ and sleep variables from T1 to T2 and from T2 to T3 and to study gender differences in sleep and emotional and behavioral difficulties. Second, Pearson correlations were calculated for examining the relations between sleep duration, sleep problems, and emotional and behavioral difficulties. To determine whether sleep duration and sleep problems predict adolescent emotional and behavioral difficulties or vice versa, cross-lagged structural equation models (SEM) were ran with Mplus (Mplus 8.3). In this study, data were used for participants who provided information either at T1 only (n=8743; 51.1% girls), at both T1 and T2 (n=5271; 51.6% girls), or at T1 and T3 (n=3703; 54.0% girls), thus for around 27.4–64.8% of the age cohort. As adolescent socioeconomic status (SES) may play a role in physical health and sleep in adolescence (Marco et al. 2012), analyses were controlled for mothers’educational level. Furthermore, because gender differences were expected to be found in adolescents’sleep (Maslowsky and Ozer 2014), analyses were adjusted for gender. The structure of the SEM model was simplified by adding each component to the model as latent. This missing data method (missing at random approach) was used for estimating the model from full data without imputing. The cross-lagged path model for studying the relations between emotional and behavioral difficulties and sleep across the three measurement points is presented in Fig. 2. The model was fitted by applying maximum-likelihood estimation to use all available information instead of imputing data missing at random Journal of Youth and Adolescence (Allison 2012). The model assessed the individual variation separately at T1, T2, and T3. All variables used in the model were normally distributed. Goodness-of-fit was evaluated using the following descriptive indices: Comparative Fit Index (CFI), Tucker Lewis Index (TLI), and Root Mean Square Error of Approximation (RMSEA). Due to the large sample size, x2-test results are not reported here. First, full model was tested for sleep (sleep duration and sleep problems) and total difficulties on psychosocial behavior, including stability for sleep (T1–T3) and total difficulties (T1–T3), the paths from sleep at T1 to total difficulties at T2, the paths from sleep at T2 to total difficulties at T3, the paths from total difficulties at T1 to sleep at T2, and the paths from total difficulties at T2 to sleep at T3, among all participants. Finally, the full models for sleep and emotional symptoms, conduct problems, hyperactivity and peer problems were examined. Results Stability of Sleep and Emotional and Behavioral Difficulties during Adolescence Repeated measures ANOVA revealed significant main effects for total difficulties, emotional symptoms, hyperactivity, conduct problems, and peer problems (Table 1). Adolescents’sleep duration decreased over time, with 36 min less sleep from T1 to T2 and 13 min less sleep from T2 to T3 (p< 0.001). In addition, sleep problems increased significantly from T1 to T2 and from T2 to T3 (p< 0.001). Total difficulties increased significantly over time (p< 0.001). Further, emotional symptoms, hyperactivity, and peer problems increased throughout adolescence (p< 0.001 for all repeated measures ANOVAs). Conduct problems, on the other hand, increased from T1 to T2 (p< 0.001), but decreased from T2 to T3 (p< 0.001). Adolescents who slept less than 8 h throughout adolescence (n=105, 59.0% girls) were also compared with adolescents who slept 8 h or more (n=1036; 51.3% girls) at every measurement point. For both groups, emotional and behavioral difficulties measured by sum score for total difficulties on psychosocial behavior increased with age. However, adolescents who slept less than 8 h/night throughout adolescence had significantly more emotional and behavioral difficulties at T1 (p< 0.001), T2 (p< 0.001), and T3 (p< 0.001) than adolescents who slept at least 8 h/ night at each measurement time (Fig. 3). Differences in Sleep and Emotional and Behavioral Difficulties by Gender As shown in Table 2, girls had shorter sleep duration than boys at every time point. Girls also had more sleep problems than boys at each time point (p< 0.001). Regarding emotional and behavioral difficulties, the score for total difficulties was higher among girls than boys at each time point (p< 0.001). In addition, girls had more emotional problems than boys throughout adolescence (p< 0.001) and also more hyperactivity (p< 0.001). However, boys had more conduct problems (p< 0.001) and peer problems (p< 0.001) than girls throughout adolescence. Significant interaction effects were found for sleep duration, emotional symptoms, hyperactivity, and conduct problems. Therefore, the sleep duration of girls decreased more rapidly than that of boys (p< 0.001), the emotional symptoms and hyperactivity of girls increased more rapidly than that of boys (p< 0.001), and the conduct problems of boys changed more rapidly than that of girls (p< 0.001). Cross-Sectional Correlations between Sleep and Emotional and Behavioral Difficulties across Adolescence Table 3presents Pearson correlations between sleep duration, general sleep problems, total difficulties, emotional symptoms, conduct problems, hyperactivity and peer problems at T1, T2, and T3. Bivariate correlations (T1–T3) showed that shorter sleep duration correlated with more total difficulties (r=−0.04 to −0.20, p<0.01 for all correlations), emotional symptoms (r=−0.04 to −0.14, p<0.01 for all correlations), conduct problems (r=−0.04 to −0.22, p<0.01 for all correlations), and hyperactivity (r=−0.06 to −0.22, p<0.01 for all correlations) throughout adolescence, both at the measurement age and contributing to later emotional and behavioral difficulties. Regarding peer problems, shorter sleep duration at either Fig. 2 Cross-lagged relationships between emotional and behavioral difficulties, sleep duration, and sleep problems. *p< 0.05 for correlation, **p< 0.01 for correlation, ***p< 0.001 for correlation Journal of Youth and Adolescence T2 or T3 correlated significantly with more peer problems at T3 (sleep duration at T2: r=−0.06, p<0.01 and at T3: r=−0.05, p<0.05). Vice versa, T2 peer problems correlatedwithshortersleepdurationatT3(r=−0.04, p<0.05). Self-reported sleep problems were also related to emotional and behavioral difficulties throughout adolescence, similarly to shorter sleep duration. The more self-reported sleep problems present at T1, T2, and T3, the more total difficulties (r=0.32 to 0.39, p< 0.01 for all correlations), emotional symptoms (r=0.36 to 0.41, p< 0.01 for all correlations), conduct problems (r=0.12 to 0.22, p< 0.01 for all correlations), hyperactivity (r=0.21 to 0.26, p< 0.01 for all correlations), and peer problems (r=0.14 to 0.17, p< 0.01 for all correlations) arose throughout adolescence, both at the measurement age and contributing to later emotional and behavioral difficulties. Cross-Lagged Models for Sleep and Emotional and Behavioral Difficulties The full model for total difficulties and sleep fitted the data well (RMSEA =0.02, SRMR =0.01, CFI =1.0, TLI = 0.99). Figure 2shows the cross-sectional and longitudinal pathways in SEM model for total difficulties and sleep variables. From a longitudinal perspective, short sleep duration (at T1 or T2) was associated with slightly more total difficulties later in adolescence (at T2: β=−0.05, Table 1 Mean (M), standard deviation (SD), and repeated measures ANOVA p-values (p) for sleep, total difficulties on psychosocial behavior, emotional symptoms, conduct problems, hyperactivity, and peer problems at T1 (n= 8730–8743), T2 (n=5271), and T3 (n=3703) T1 T2 T3 ANOVA Variable (range) M (SD) M (SD) M (SD) Fp Sleep duration, h:min 8:38 (0:46) 8:02 (0:48) 7:49 (0:46) 1228.31 <0.001 Sleep problems (0–4) 1.78 (1.02) 1.85 (1.02) 2.06 (1.02) 85.79 <0.001 Total difficulties on Psychosocial Behavior (0–40) 9.23 (5.32) 11.01 (5.88) 10.84 (5.65) 195.75 <0.001 Emotional symptoms (0–10) 2.44 (2.14) 3.11 (2.39) 3.32 (2.40) 204.35 <0.001 Conduct problems (0–10) 1.85 (1.58) 2.03 (1.76) 1.68 (1.60) 30.27 <0.001 Hyperactivity (0–10) 2.84 (2.12) 3.32 (2.28) 3.37 (2.28) 141.99 <0.001 Peer problems (0–10) 2.10 (1.66) 2.55 (1.79) 2.47 (1.69) 88.39 <0.001 Fig. 3 Mean scores for emotional and behavioral difficulties in adolescents who slept less than 8 h/night (n=105, 59.0% girls) throughout adolescence as compared with adolescents who slept at least 8 h/night (n=1036; 51.3% girls) throughout adolescence Journal of Youth and Adolescence p< 0.001 [95% CI, −0.07 to −0.02]; at T3: β=−0.05, p< 0.05 [95% CI, −0.09 to −0.01]). Further, sleep problems (at T1 or T2) were associated with slightly more total difficulties later in adolescence (at T2: β=0.05, p< 0.001 [95% CI, 0.03–0.08]; at T3: β=0.05, p< 0.01 [95% CI, 0.02–0.09]). Vice versa, total difficulties (at T1 or T2) were associated with slightly more sleep problems later in adolescence (at T2: β=0.10, p< 0.001 [95% CI, 0.08–0.13]; at T3: β=0.10, p< 0.001 [95% CI, 0.06–0.13]). However, total difficulties were not associated with sleep duration later in adolescence. Cross-Lagged Models for Sleep, Emotional Symptoms, Conduct Problems, Hyperactivity and Peer Problems The full models for emotional symptoms, conduct problems, hyperactivity, peer problems and sleep duration and sleep problems fit the data well. Table 4shows the fit indices for each model. First, the model for sleep and emotional symptoms was examined. The full model for the emotional symptoms and sleep variables fit the data well (RMSEA =0.02, SRMR =0.01, CFI =0.99, TLI =0.97). Short sleep duration at T1 was associated with slightly more emotional symptoms at T2 (β=−0.04, p< 0.05 [95% CI, −0.07 to −0.02]), but the association was not significant from T2 to T3 (β=−0.03, p> 0.05 [95% CI, −0.06–0.01]). Sleep problems either at T1 or T2 were associated with more emotional symptoms later in adolescence (at T2: β= 0.05, p< 0.001 [95% CI, 0.02–0.08]; at T3: β=0.04, p< 0.01 [95% CI, 0.00–0.08]). Vice versa, emotional symptoms either at T1 or T2 were associated with more sleep problems later in adolescence (at T2: β=0.15, p< 0.001 [95% CI, 0.12–0.18]; at T3: β=0.14, p< 0.001 [95% CI, 0.10–0.17]). However, emotional symptoms were not associated with sleep duration later during adolescence. Second, the model for sleep and conduct problems was examined. The full model for the conduct problems and sleep duration and sleep problems fit the data well (RMSEA =0.03, SRMR =0.02, CFI =0.99, TLI =0.94). Short sleep duration at T2 was associated with more conduct problems at T3 (β=−0.05, p< 0.001 [95% CI, −0.08 to −0.01]). Conversely, conduct problems were not associated with later sleep duration. Conduct problems at T1, however, were associated with more sleep problems at T2 (β=0.05, p< 0.001 [95% CI, 0.02–0.08]). Sleep problems were not associated with conduct problems later in adolescence. Third, the model for sleep and hyperactivity was examined. The full model for the hyperactivity and sleep fit the data well (RMSEA =0.02, SRMR =0.01, CFI =0.99, TLI =0.96). Short sleep duration at T1 was associated with more hyperactivity at T2 (β=−0.03, p< 0.05 [95% CI, −0.05–0.00]). However, hyperactivity was not associated with sleep duration later during adolescence. Further, sleep problems either at T1 or T2 were associated with more hyperactivity later in adolescence (at T2: β=0.06, p< 0.001 [95% CI, 0.04–0.09]; at T3: β=0.06, p< 0.01 [95% CI, 0.02–0.09]). Hyperactivity either at T1 or T2 was associated with more sleep problems later in adolescence (at Table 2 Mean (M), standard deviation (SD), and repeated measures ANOVA p-values (p) for sleep duration, sleep problems, total difficulties on psychosocial behavior, emotional symptoms, conduct problems, hyperactivity, and peer problems in boys and girls at T1 (n=8730–8743; 51.1% girls) or at T1 and T2 (n=5271; 51.6% girls) or at T1 and T3 (n=3703; 54.0% girls) T1 M (SD) T2 M (SD) T3 M (SD) ANOVA Variable (range) Boys Girls Boys Girls Boys Girls Fp Sleep duration, h:min 8:44 (0:47) 8:32 (0:45) 8:05 (0:50) 7:58 (0:46) 7:49 (0:45) 7:49 (0:46) 12.72 <0.001 Interaction: 12.22 <0.001 Sleep problems (0–4) 1.64 (0.96) 1.91 (1.06) 1.68 (0.95) 2.01 (1.05) 1.88 (0.97) 2.21 (1.04) 111.31 <0.001 Interaction: 2.98 0.051 Total difficulties on psychosocial behavior (0–40) 8.76 (5.32) 9.68 (5.29) 10.56 (6.17) 11.44 (5.56) 10.19 (5.83) 11.39 (5.43) 34.26 <0.001 Interaction: 1.66 0.191 Emotional symptoms (0–10) 1.79 (1.87) 3.06 (2.21) 2.33 (2.16) 3.85 (2.37) 2.55 (2.16) 3.98 (2.39) 414.90 <0.001 Interaction: 2.91 <0.001 Conduct problems (0–10) 1.97 (1.66) 1.74 (1.50) 2.30 (1.91) 1.78 (1.56) 1.82 (1.76) 1.56 (1.45) 42.78 <0.001 Interaction: 10.89 <0.001 Hyperactivity (0–10) 2.75 (2.07) 2.93 (2.17) 3.20 (2.18) 3.43 (2.36) 3.13 (2.16) 3.56 (2.35) 14.43 <0.001 Interaction: 6.00 <0.01 Peer problems (0–10) 2.25 (1.73) 1.96 (1.58) 2.73 (1.88) 2.37 (1.69) 2.68 (1.74) 2.29 (1.63) 44.16 <0.001 Interaction: 1.72 0.179 Journal of Youth and Adolescence Table 3 Pearson correlations between sleep duration, sleep problems, and emotional and behavioral difficulties at T1, T2, and T3 Variable 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 1. T1: Sleep duration 2. T2: Sleep duration 0.35b 3. T3: Sleep duration 0.25b0.37b 4. T1: Sleep problems −0.12b−0.06b−0.07b 5. T2: Sleep problems −0.09b−0.09b−0.02 0.34b 6. T3: Sleep problems −0.07b−0.08b−0.08b0.29b0.43b 7. T1: Total difficulties on Psychosocial Behavior −0.20b−0.08b−0.10b0.36b0.22b0.16b 8. T2: Total difficulties on Psychosocial Behavior −0.12b−0.17b−0.08b0.20b0.39b0.26b0.47b 9. T3: Total difficulties on Psychosocial Behavior −0.11b−0.14b−0.12b0.18b0.27b0.32b0.42b0.55b 10. T1: Emotional symptoms −0.12b−0.04a−0.08b0.37b0.24b0.19b0.75b0.35b0.33b 11. T2: Emotional symptoms −0.12b−0.14b−0.07b0.21b0.41b0.30b0.38b0.77b0.46b0.47b 12. T3: Emotional symptoms −0.11b−0.10b−0.08b0.18b0.27b0.36b0.32b0.42b0.76b0.41b0.57b 13. T1: Conduct problems −0.22b−0.08b−0.09b0.22b0.12b0.06b0.71b0.33b0.26b0.32b0.14b0.10b 14. T2: Conduct problems −0.09b−0.13b−0.04b0.10b0.20b0.09b0.29b0.71b0.33b0.10b0.33b0.09b0.38b 15. T3: Conduct problems −0.07b−0.10b−0.11b0.06b0.11b0.12b0.25b0.34b0.68b0.09b0.15b0.30b0.31b0.41b 16. T1: Hyperactivity −0.22b−0.09b−0.07b0.22b0.16b0.11b0.73b0.40b0.33b0.33b0.21b0.15b0.53b0.31b0.26b 17. T2: Hyperactivity −0.13b−0.17b−0.06b0.15b0.26b0.16b0.36b0.72b0.43b0.18b0.35b0.21b0.29b0.47b0.29b0.48b 18. T3: Hyperactivity −0.08b−0.12b−0.11b0.14b0.20b0.21b0.32b0.43b0.72b0.18b0.25b0.33b0.22b0.29b0.43b0.40b0.56b 19. T1: Peer problems 0.01 −0.01 −0.05a0.16b0.08b0.07b0.58b0.23b0.22b0.35b0.18b0.18b0.21b0.05a0.05b0.12b0.01 0.05a 20. T2: Peer problems 0.01 −0.03 −0.04a0.10b0.17b0.14b0.30b0.61b0.32b0.20b0.37b0.23b0.13b0.31b0.13b0.10b0.14b0.09b0.42b 21. T3: Peer problems −0.03 −0.06b−0.04 0.08b0.12b0.14b0.26b0.33b0.60b0.15b0.22b0.36b0.13b0.18b0.32b0.11b0.10b0.13b0.35b0.50b aCorrelation is significant at 0.05 level bCorrelation is significant at 0.01 level Journal of Youth and Adolescence