Maternal emotions during the pre/postnatal periods and children’s sleep behaviors: the mediating role of children’s behavior

Abstract

Background:

We aimed to investigate the association of mother’s perceived levels of depression and happiness across all trimesters of pregnancy and after delivery on their children’s sleep problems, as well as examine the possible mediating effect of children’s behavioral problems.

Methods:

Participants included 1257 children (54% boys, mean age = 5.74±0.48 years). Pre- and postnatal maternal emotions were self-reported using a 5-point scale for happiness and a 3-point scale for depressive emotions. Children reported sleep and behavioral problems using the Child Behavior Checklist.

Results:

Adjusted models showed that children of women reporting depressive emotions during either the postnatal period (β=3.07, p=0.01) or both prenatal and postnatal periods (β=2.91, p=0.01) were more likely to report sleep disturbances. By contrast, children of women reporting higher levels of happiness in the second (β=−1.91, p=0.04) and third (β=−2.27, p=0.001) trimesters were less likely to report sleep problems.

Limitations:

Differences in maternal-report of children’s behaviors could reflect memory and recall bias, and maternal emotions were assessed by researcher-designed single item measures. All measures were completed by a single reporter. Other associated factors should be considered in clarifying the complex associations.

Conclusions:

Maternal depressive emotions during pre- and postnatal periods were associated with an increase in children’s sleep problems, while increased happiness during pregnancy was associated with a decrease in children’s sleep problems. Children’s behavioral problems significantly mediated these relationships. To our knowledge, this is the first study to examine the relationship between maternal pre- and postnatal emotions and children’s sleep behavior in an Asian sample.

1. Introduction

Maternal prenatal and postpartum mood disturbances have been considered early-life risk factors for predisposing to later emotional and behavioral problems in children (Hannigan et al., 2018; Leis, Heron, Stuart, & Mendelson, 2014). There is emerging evidence suggesting that maternal prenatal (Matenchuk et al., 2019) and postnatal (Piteo et al., 2013) anxiety and depression are associated with sleep problems among infants, measured as increased nighttime awakenings, decreased time in deep sleep, and greater indeterminate sleep. Beyond infancy, longitudinal research reports that maternal prenatal anxiety and depression predicted an increased likelihood of young children’s sleep problems at the ages of 18 months and 30 months (O’Connor et al., 2007) and during adolscence (Taylor et al., 2017). However, the impact of maternal perinatal mood on offspring’s sleep patterns at preschool age remains understudied. Research shows that sleep problems affect 15–25% of the preschool-age population (Owens, 2007) and are linked to subsequent impairments in children’s neurobehavioral growth and academic performances (Plancoulaine, Lioret, Regnault, Heude, & Charles, 2015). Thus, understanding associations between maternal mood and childhood sleep has profound implications for child health and development.

Previous research on maternal mood and children’s sleep behaviors has primarily focused on samples in English-speaking countries such as the United States (Swanson, Flynn, Wilburn, Marcus, & Armitage, 2010), Canada (Oberlander et al., 2010), the United Kingdom (O’Connor et al., 2007), and Australia (Johnson, McMahon, & Gibson, 2014). A cross-cultural study found that parental-perceived sleep problems are more prevalent among preschool-aged children in predominantly Asian countries (24.2%) compared to children in predominantly Caucasian countries (18.4%) (Mindell, Sadeh, Kwon, & Goh, 2013). The high risk for sleep-related impairments in Asian children warrants more research on this vulnerable population.

In contrast to negative mood, the extant literature has consistently reported an association between positive affect and better sleep outcomes, particularly in healthy populations (Fuligni & Hardway, 2006; Ong, Kim, Young, & Steptoe, 2017). John and colleagues found that prenatal maternal hardiness, a strong predictor of positive psychological adjustment to pregnancy, was associated with parenting cognition and behaviors around child sleep and, in turn, contributed to child sleep at 19 months postpartum (Johnson et al., 2014). Feelings of happiness have also been associated with increased self-reported sleep duration and decreased variability in sleep time (Fuligni & Hardway, 2006) in adolescents. However, little is known about the relationship between prenatal and postpartum happiness among mothers and sleep patterns in their children during early childhood.

Additionally, assessments of happiness are more culturally appropriate for Asian women, while depressive symptoms, which are often related to numerous mental disorders, are stigmatized. Specifically, the reported prevalence of depression is very low in China compared to other countries with similar economic development, such as Morocco, India, and Ukraine (Rai, Zitko, Jones, Lynch, & Araya, 2013). This low prevalence may possibly be a reflection of cultural “denial” of depression or a propensity to express depression somatically (Gordon Parker, Gemma Gladstone, & Kuan Tsee Chee, 2001). Therefore, assessing happiness may provide a complement to questions about negative mood that could reduce self-report bias due to cultural stigma.

Few studies have examined the biobehavioral underpinnings of the relationship between maternal emotions and child development. Given a combination of “programming effects” on the developing fetus, as well as social learning and attachment problems, environmental stress underpinnings, and reciprocal effects, prenatal and postpartum exposure to maternal negative emotions has been considered an environmental risk factor for both internalizing and externalizing behavioral difficulties in childhood (Cuijpers, Weitz, Karyotaki, Garber, & Andersson, 2015; Goodman & Gotlib, 1999). Although the causal relationship between behavioral problems and child sleep remains unclear, emerging evidence has further suggested that emotional and behavioral function contributes to subsequent sleep health in children (Quach, Nguyen, Williams, & Sciberras, 2018; Shanahan, Copeland, Angold, Bondy, & Costello, 2014). Specifically, research shows that behavioral problems assessed by the Child Behavior Checklist (CBCL) factor scores predicted 17% to 22% of the variance in sleep scores that reflect parasomnias, tiredness, and insomnia among school-aged children (Stein, Mendelsohn, Obermeyer, Amromin, & Benca, 2001). Thus, behavioral problems in children may, at least, partially account for the link between maternal emotions and child sleep. However, very few studies have explored the mediating role of child behavioral characteristics.

The primary aim of this study was to investigate the relationship between prenatal and postnatal emotions in women and their children’s sleep problems (parental report) at ages 5 and 6 years in a healthy Chinese sample. We hypothesize that 1) children of mothers with prenatal and/ or postnatal depressive mood will exhibit an increased risk for sleep problems, and 2) children of mothers with higher self-perceived happiness during pregnancy will exhibit a decreased risk for sleep problems. As an exploratory aim, we examined whether children’s behavioral problems mediated the relationship between maternal emotions and children’s sleep. To our knowledge, this is one of the first studies to investigate children’s sleep behaviors in relation to maternal prenatal and postnatal emotions in an Asian sample. A better understanding of the role of maternal perinatal emotions in relation to their children’s sleep will ultimately inform primary and secondary prevention efforts in reducing childhood sleep disturbances and subsequent impairment in health and daily functioning.

Methods

2.1. Setting, participants, and study design

The current study was a part of the Jintan Child Cohort Study in China (J. Liu et al., 2015), which aims to examine how early health risk factors predict cognitive and behavioral development in childhood and adolescence. Detailed information on the Jintan Cohort Profile, including subjects, recruitment, and procedures, has been reported elsewhere (Jianghong Liu et al., 2011; J. Liu et al., 2010). As shown in the participant flowchart (Figure 1), this retrospective longitudinal study used a multiple-stage sampling method and enrolled 1656 Chinese children (55.5% boys, 44.5% girls) who represented preschoolers from different school districts in Jintan City in 2004. In China, preschool spans three years when children are ages 3–6 years. According to their year (1^st, 2^nd, and 3^rd) in preschool, children were classified into lower, middle and upper cohort. The present study only focused on Wave I follow-up, in which data was collected in 2005−2007 for 1385 children when each grade cohort were in the last month of their preschool program (year 3) (J. Liu et al., 2010). The current study used a sample of 1257 children (682 boys, 575 girls). Children’s sleep and behavior were reported by mothers when children were age 5–6 years and in their last month of preschool. At the same time, mothers’ emotional state during the prenatal and postnatal period were retrospectively reported using a prenatal questionnaire. Child and family characteristics of the sample are summarized in Table 1. Institutional Review Board approval was obtained from the University of Pennsylvania and the Ethical Committee for Research at Jintan Hospital, and informed consent was obtained via written agreement.

Figure 1.

Participant flowchart

Table 1.

Sample characteristics and associated sleep scores (n=1257).

	M ± SD/n (%)	Child sleep (t-transformed score) (M ± SD)	t/F/r	p
Child age	5.74±0.48	49.89± 9.93	0.05a	0.09
Sex			1.68	0.20
Boys	682 (54.26)	50.22±9.83
Girls	575 (45.74)	49.49±10.05
Residence			6.63	0.001 b **
Rural	252 (20.05)	48.92±9.87
Suburban	503 (40.02)	49.13±9.07
Urbans	502 (39.94)	51.13±10.65
Mother’s education level			2.24	0.11
≤ middle school	552 (44.02)	49.28±9.79
high school	308 (24.56)	49.98±10.12
≥ college	394 (31.42)	50.66±9.98
Father’s education level			3.05	0.05
≤ middle school	414 (33.04)	49.02±9.87
high school	366 (29.21)	49.85±10.30
≥ college	473 (37.75)	50.67±9.68
Cesarean birth / birth complication			1.35	0.24
No	708 (57.10)	49.86±10.20
Yes	532 (42.90)	50.53±9.70
Child behavior (raw score)	30.83±19.55	49.89± 9.93	0.62 a	<0.001**

Note:

^a.correlation coefficient=0.05 between age and sleep score, and 0.62 between child behavior and child sleep.

^b.Post hoc analysis results: urban vs rural, t=3.05, p=0.007, urban vs suburban, t=3.39, p=0.003 M=mean, SD=standard deviation. t value represents the calculated difference represented in units of standard error; F stands for a test statistic for the ratio of two estimates of the same variance, and r stands for the correlation coefficient.

^*p<0.05

^**p<0.01

2.2. Variables and measurements

Maternal emotions.

Maternal emotions, including perceived feelings of depressive mood and happiness, were assessed retrospectively through researcher-designed questionnaires when children were 5–6 years old. Specifically, women self-reported prenatal (Item 1) and postnatal (Item 2) depressive emotions as follows: 1= “Feel Very Depressed,” 2= “Feel Somewhat Depressed,” and 3= “Feel Not At All Depressed.” We specified that “prenatal” referred to during pregnancy, while “postnatal” referred to 3 months post-pregnancy. As there were only 3 individuals who reported feeling “very depressed” during the prenatal or postnatal period, we categorized participants into 2 groups: depressive and non-depressive emotions. Based on the perinatal/postnatal assessment, a combined variable was generated to reflect perinatal depressive emotions, including levels of non-depressive emotions, perinatal depressive emotions only, postnatal depressive emotions only, and depressive emotions at both assessments. Regarding happiness, women were asked to rate feelings of happiness during each trimester of pregnancy as follows: −2= “Very Unhappy,” −1= “Unhappy,” 0= “Neither Unhappy Nor Happy (i.e., “Neutral”),” 1= “Happy,” and 2= “Very Happy.” We recorded the category scores as 1, 2, 3, 4 respectively.

Sleep and behavioral problems.

Both children’s sleep and behavioral problems were assessed with the Chinese version of the Child Behavior Checklist (CBCL) for Preschool Children (Achenbach & Rescorla, 2000), which was completed by each child’s mother. The CBCL contains 7 sleep items, : unwillingness to sleep alone, difficulty initiating sleep (DIS), having nightmares, resisting going to bed at night, sleeping less than most children (lack of sleep), talking or crying out in sleep (sleep talking/crying), and difficulty maintaining sleep (DMS). Items were rated) for current frequency and frequency within the past 2 months (0= “No,” 1= “Sometimes,” 2= “Often”. The total score of the 7 items was translated into the standard T score, and higher scores indicated greater sleep problems (J. Liu et al., 2012). Children’s overall behavioral problems, including both internalizing and externalizing behaviors, were calculated by the sum of raw scores for two behavioral domains which was equivalent to CBCL total score minus sleep raw score. Higher scores indicate greater behavioral problems. The CBCL has demonstrated satisfactory psychometric properties in the assessment of behavioral and emotional problems in Chinese children (J. Liu, Cheng, & Leung, 2011). The sleep syndrome derived from the CBCL has been previously reported in our sample (J. Liu et al., 2012).

Covariates.

Sociodemographic information collected included the child’s sex, age, residence (e.g., urban, suburban, or rural), cesarean birth / birth complications (e.g. preterm) and maternal and paternal education level. These demographic variables were selected on the basis of literature reviews and/or our preliminary analyses indicating that these variables were associated with children’s sleep problems (J. Liu et al., 2012).

2.3. Statistical analyses

Sample characteristics were summarized by descriptive statistics, including mean, standard deviation, and percentage (Table 1). At the bivariate level, we examined differences in sleep scores between groups by sex, maternal and paternal education level, cesarean birth / birth complications, home residence, and depressive emotions using chi-squared tests. For bivariate analyses that showed overall differences, post hoc analyses were followed using pairwise comparisons with a Tukey-Kramer adjustment. Correlations of sleep scores with child age and CBCL behavior score and happiness ratings were made using Spearman correlation tests. In addition, we used generalized linear modeling to test the adjusted associations between child sleep, prenatal and postnatal depressive emotions, and maternal happiness during pregnancy. Depressive emotions during the prenatal period, postnatal period, and both periods combined, as well as happiness during each trimester, were entered into the regression models separately. We also repeated the analysis on prenatal emotions (depressive emotions and happiness) with postnatal depressive emotions treated as a covariate. To address the secondary aim, we followed Baron & Kenny’s procedures to estimate the mediating effect of children’s behavioral problems (Baron & Kenny, 1986; MacKinnon & Dwyer, 1993): (1) testing the direct effect (c) of maternal emotions on child sleep; (2) testing the effect (a) of maternal emotions on child behavior ; and (3) testing the effect (c′) of maternal emotions on child sleep controlling for child behavior. In each model set, we controlled for children’s sex, age, grade, home residence, and maternal and paternal education level (Model 1) before adding the child behavioral score as a covariate into the model (Model 2). Mediating effect was determined by the following criteria: first, child behavior (mediator) was a significant predictor of child sleep; second, maternal emotion was a significant predictor of child behavior (mediator) and child sleep (outcome); third, the coefficient (c′) of maternal emotion was reduced in absolute size compared to it is in the first model (c). We clustered standard errors at the cohort level. Data were analyzed using STATA 14.0 for Windows, and a p value of <.05 was considered significant.

3. Results

3.1. Sample characteristics

Table 1 shows sample characteristics and associations with sleep scores. The mean (standard deviation) age of our sample was 5.74 (0.48) years. The average raw sleep score was 2.77±2.14 and the mean t-transformed sleep score was 49.89± 9.93. There were no significant overall differences in sleep problems (p’s>0.05) between groups by children’s sex, parental education level, or cesarean birth / birth complications, but home residence was significantly correlated with sleep problems (F=6.63, p=0.001). Post hoc analyses found that children living urban areas had significantly greater sleep problems (higher scores) than those in the rural (t=3.05, p=0.007) and suburban areas (t=3.39, p=0.003). Sleep scores were positively correlated with children’s behavioral problems (r=0.62, p<0.001) but were not significantly correlated with children’s age (r=0.05, p=0.09). The prevalence of maternal depressive emotions is detailed in Table 2. Thirty-one percent (n=141) of women reported feeling “somewhat” or “very” depressed in the prenatal period, and 20.48 % (n=128) reported any level of postnatal depressive emotions. Among those with both prenatal and postnatal data available (n=437), 12% (n=52) had persistent depressive emotions during both periods of time. As shown in Table 3, feelings of happiness had a positive trajectory throughout all trimesters, and the highest average score of happiness was reported in the third trimester.

Table 2.

Self-reported depressive emotions and associated sleep and behavior scores

Depressive emotions	N (%)	Child Sleep
		t-transformed score	t/F	p value
Prenatal (n=451)
No	310 (68.74)	49.33±9.41	1.44	0.23
Yes	141 (31.26)	50.50±9.19
Postnatal (n=625)
No	497 (79.52)	49.17±9.37	12.34	0.001**
Yes	128 (20.48)	52.63±10.42
Combined1 (n=437)
No	268 (61.33)	49.08±9.21	2.29	0.08
Prenatal only	83 (18.99)	49.80±8.96
Postnatal only	34 (7.78)	52.29±10.49
Both period	52 (11.90)	52.10±9.51

Note:

^1.Combined depressive emotions were recorded based on prenatal and postnatal depressive emotions. It contains missing data.

^2.t value represents the calculated difference represented in units of standard error; F stands for a test statistic for the ratio of two estimates of the same variance.

^**p<0.01

Table 3.

Self-reported happiness and correlations with children’s sleep (r)

	M ± SD	Happiness (Trimester 2)	Happiness (Trimester 3)	Child sleep (t-transformed score)
Happiness
Trimester 1 (n=1224)	4.92±1.19	0.60**	0.46**	−0.13**
Trimester 2 (n=1216)	5.05±0.98		0.68**	−0.13**
Trimester 3 (n=1218)	5.25±0.99			−0.13**

Note: higher average happiness scores showed that women felt happier. ** p<0.01 r =correlation coefficient. ² M=mean, SD=standard deviation. Higher average happiness scores showed that women felt happier.

^**p<0.01

3.2. The relationship between maternal prenatal and postnatal depressive emotions and children’s sleep problems

Table 2 shows the associations between perinatal depressive emotions and children’s sleep problems by ANOVA. Whereas children of mothers who percieved postnatal depressive emotions had significantly more severe mother-reported sleep problems (F=12.34, p=0.001), children’s sleep problems did not differ significantly between groups with and without prenatal depressive emotions (p’s>0.05). After adjustment for children’s age, sex, residence, and parental education level, the generalized linear model (Table 4) showed no significant association between maternal prenatal depressive emotions and the risk for children’s sleep problems (β=0.99, p=0.42). On the other hand, children of women with postnatal depressive emotions tended to exhibit more sleep problems (β=3.67, p=0.04) versus the group without postnatal depressive emotions. The combined periods of depressive emotions showed similar results in Model 1. Compared with the condition of no depressive emotions at all, maternal depressive emotions in the postnatal period only (β=3.07, p=0.01) or in both prenatal and postnatal periods (β=2.91, p=0.01) significantly predicted greater sleep problems in children. However, the magnitudes of these associations were low with standardized betas between 0.09–0.15 (Table 4). Additionally, there was no association between prenatal depressive emotions only (without postnatal depressive emotions) and child sleep problems (β=0.65 p=0.70).

Table 4.

Adjusted relationship between children’s sleep problems and pre- and postnatal depressive emotions

	Child sleep problems (t-transformed score)
	Model 1			Model 2
Depressive emotions	B (Robust SE)	P	Standardized B	B (Robust SE)	P	Standardized B
Prenatal (n=423)
Prenatal depressive emotions a	0.94(0.78)	0.35	0.05	0.18 (0.68)	0.81	0.003
Child behavioral problems	0.32(0.02)	0.004**	0.62	0.31 (0.02)	0.006**	0.66
Postnatal (n=588)
Postnatal depressive emotions b	3.41 (0.72)	0.04*	0.14	1.37 (0.92)	0.27	0.06
Child behavioral problems	0.32(0.02)	0.004**	0.62	0.31 (0.02)	0.004**	0.62
Combined (n=411)
Combined depressive emotions c
Prenatal Only	0.74 (1.39)	0.65	0.03	−0.006 (1.44)	0.99	0.001
Postnatal Only	3.11(0.16)	0.003*	0.09	0.84 (0.65)	0.33	0.03
Both periods	2.63 (0.52)	0.04*	0.09	0.74(0.53)	0.29	0.03
Child behavioral problems	0.31(0.02)	0.005**	0.62	0.31 (0.02)	0.006**	0.65
Prenatal (n=423)
Prenatal depressive emotions a	0.99(0.97)	0.42	0.05	0.06(0.78)	0.95	0.003
Child behavioral problems	0.32(0.01)	0.001**	0.62	0.31(0.02)	0.006**	0.66
Postnatal (n=588)
Postnatal depressive emotions b	3.67(0.72)	0.04*	0.15	1.53(1.01)	0.27	0.06
Child behavioral problems	0.32(0.01)	0.001**	0.62	0.31(0.02)	<0.01**	0.62
Combined (n=411)
Combined depressive emotions c
Prenatal Only	0.65(1.48)	0.70	0.03	−0.06(0.97)	0.96	0.002
Postnatal Only	3.07(0.26)	0.01*	0.09	0.71(0.71)	0.42	0.02
Both periods	2.91(0.29)	0.01*	0.09	0.75(0.45)	0.24	0.03
Child behavioral problems	0.32(0.01)	0.001**	0.62	0.30(0.02)	0.006**	0.65

Note:

^1.Model 1 adjusted for child age, sex, home residence, and mother’s and father’s education, birth complications; and prenatal depressive emotion and child behavior entered the model separately. Model 2 adjusted for variables in Model 1 plus children’s behavioral problems. Standard errors were clustered at the cohort level. Sample sizes represent numbers of complete cases of each model. The full models accounted for 41.55% −43.26% of the variance.

^2.B=coefficient, SE=standard error, p=p values

^3.The function of Standardized coefficient cannot be combined with cluster adjustment in STATA. Thus, the standardized B reported here were from models without cluster adjustment. Statistical significance was similar with and without cluster function.

Reference group:

^a.no prenatal depressive emotions;

^b.no postnatal depressive emotions;

^c.no perinatal depressive emotions.

^*p<0.05;

^**p<0.01

3.3. The relationship between maternal happiness during pregnancy and children’s sleep problems

As shown in Table 3, child sleep scores showed significant inverse correlations with happiness ratings at each trimester (r=−−0.14, p’s<0.01), suggesting that increased maternal happiness was associated with fewer sleep problems in children. Table 5 shows this association after adjusting for covariates. Maternal happiness during pregnancy was significantly associated with children’s sleep problems. Specifically, as women’s self-reported happiness levels increased during the first (β=−1.71, p<0.001), second (β=−1.91, p=0.04) and third trimesters (β=−2.27, p=0.001), their children tended to have lower sleep scores, which reflected fewer sleep problems. The standardized coefficients ranged from −0.11 to −0.13. We repeated the analysis by adding postnatal depressive emotions as a covariate and found similar results (results not presented).

Table 5.

Adjusted relationship between children’s sleep problems and prenatal happiness

	Child sleep problems (t-transformed score)
Happiness	Model 1			Model 2
	B (Robust SE)	P	Standardized B	B (Robust SE)	P	Standardized B
First trimester(n=1216)
Happiness	−1.71(0.39)	<0.001**	−0.13	−0.65 (0.42)	0.26	−0.05
Child behavioral problems	0.32(0.01)	0.001**	0.62	0.31(0.02)	0.003**	0.61
Second trimester(n=1209)
Happiness	−1.91(0.40)	0.04*	−0.11	−0.62 (0.40)	0.13	−0.04
Child behavioral problems	0.32(0.01)	0.001**	0.62	0.31(0.01)	<0.001**	0.61
Third trimester (n=1211)
Happiness	−2.27 (0.53)	0.001**	−0.13	−0.38 (0.37)	0.40	−0.02
Child behavioral problems	0.32(0.01)	0.001**	0.62	0.31(0.02)	0.003**	0.62

Note:

^1.Model 1 adjusted for child age, sex, home residence, and mother’s and father’s education, birth complications; and prenatal happiness and child behavior entered the model separately.. Model 2: adjusted for variables in Model 1 plus children’s behavioral problems. Sample sizes represent numbers of complete cases of each model. The full models accounted for 39.87%−40.27% of the variance.

^2.B=coefficient, SE=standard error, p=p values

^*p<0.05;

^**p<0.01

3.4. The mediating effects of child behavioral problems

Based on the basic procedure of the Baron & Kenny’s (1986) mediation analyses and also described in Liu & Ulrich (2015), our results showed that children’s behavioral problems mediated the association between maternal emotions and children’s sleep problems. This was demonstrated by the following steps. First, as described earlier, postnatal depressive emotion, combined perinatal depressive emotion and happiness in three trimesters were significantly associated with children’s sleep scores when not including child behavior in the multivariable model. Second, as shown in Table 6, maternal emotions, including prenatal and postnatal depressive mood and self-perceived happiness during each trimester, significantly predicted children’s behavioral problems as measured by the CBCL scores (p’s<0.05). Third, children who reported greater behavioral problems (β=3.02, p<0.01) were more likely to have increased risks of sleep problems (Table 4 and 5). This trend remained significant even after controlling for maternal emotions (Tables 5 and 6). Finally, after controlling for children’s behavioral problems, the magnitudes of the associations between of maternal emotions and children’s sleep problems decreased and the corresponding p-values were no longer significant (Tables 5 and 6), indicating child behavior functioning as a significant mediator.

Table 6.

Adjusted associations between maternal emotions and child behavioral problems

	Child behavioral problems
Depressive emotions	B (Robust SE)	p
Prenatal	2.63 (0.47)	0.03*
Postnatal Combined	6.44 (0.88)	0.02*
Prenatal only	2.25 (0.18)	0.001*
Postnatal only	7.19 (2.20)	0.08*
Both period	5.98 (0.30)	0.003**
Happiness	B (Robust SE)	p
Happiness (trimester 1)	−2.04 (0.44)	0.04*
Happiness (trimester 2)	−2.92(0.33)	0.01*
Happiness (trimester 3)	−3.16 (0.37)	0.01*

Note: B=coefficient, SE=standard error, p=p values

^*p<0.05;

^**p<0.01

4. Discussion

This study examined maternal emotions during the prenatal and postnatal periods and their association with mother-reported sleep behaviors in children at ages 5–6 years. Additionally, the study explored the mediating role of children’s behavioral problems that were shown to underlie these relationships. Our major findings include: (1) women reported feeling more depressed during pregnancy (31%) than after delivery (20%), and perceived happiness increased across trimesters; (2) women’s perceived postnatal, but not prenatal, depressive emotions were associated with greater sleep problems in their children; (3) women’s perceived happiness during the second and third trimesters was related to decreased sleep problems in their children; and (4) children’s behavior problems significantly mediated the association between maternal perinatal emotions and children’s sleep problems. The present study is among the first to examine both negative and positive maternal perinatal emotions in relation to sleep quality in children beyond infancy. Findings of this study have potential clinical implications for enhancing maternal health care and associated child well-being.

Compared to women without perinatal depressive emotions, we found that depressive emotions during the postnatal period or persistent from prenatal to postnatal period were significantly related to children’s overall sleep problems, which was reflected by the difficulty maintaining sleep (i.e., sleeping less, sleep resistance, and unwillingness to sleep alone). Our results are consistent with previous studies that have documented the association between maternal perinatal depression and preschool children’s difficulty maintaining sleep (Swanson et al., 2010) as well as between postpartum depression and increased infant nighttime awakenings in a predominantly Hispanic sample (Gress-Smith, Luecken, Lemery-Chalfant, & Howe, 2012). More recently, a study by Taylor and colleagues (2017) showed that depressive emotions during the postnatal period significantly increased the risk of sleep disturbances in children at ages 16 and 18 years, suggesting the importance of postnatal depressive emotions in predicting offspring sleep beyond infancy. Regarding emotions during the prenatal period, while previous findings suggest that higher levels of prenatal mood disturbance predicted more sleep problems in infants and toddlers (O’Connor et al., 2007), the prediction of children’s sleep quality from prenatal depressive emotions was less robust in our sample. However, we found a significant trend towards poor sleep in children in a subsample of women who had persistent depressive emotions from prenatal to postpartum period. Together with the significant associations with postnatal depressive emotions alone, exposure to maternal negative emotions in infancy may have a more profound impact on child sleep.

While prior research has primarily focused on negative perinatal emotions in women, we extended the relationship between perinatal emotions and offspring sleep to positive affect (in particular, perceived happiness) across trimesters. Women who felt greater levels of happiness during the second and third trimesters were more likely to have children with fewer sleep problems when they reached kindergarten age. The findings may not seem consistent with the results that sleep problems in children did not statistically differ by mother’s prenatal depressive emotions. One possible explanation could be that compared with happiness questions (n=1216–1224), the response rate of prenatal depressive emotion was low (n=423), which may lead to insufficient statistical power. Assessing happiness may provide a complement to depressive emotion questions, which are subjected to social stigma and have low response rates. To our best knowledge, there is no comparable evidence currently in the literature. Our findings, to some extent, parallel previous studies that reported associations of perceived happiness with higher sleep efficiency among adolescents (Fuligni & Hardway, 2006) and fewer sleep disturbances among middle-aged women (Troxel, Buysse, Hall, & Matthews, 2009). Additionally, positive affect has been postulated to promote health directly through physiological effects and indirectly through stress-buffering (Ong et al., 2017). Although 31% of women in our sample self-reported depressive emotions during pregnancy, a large portion of them (53%−71%) also reported feelings of happiness during each trimester. Future research is needed to examine the main effects and interactions between maternal depressive emotions and happiness on children’s sleep. The magnitudes of the associations between maternal emotions and child sleep were small in our sample, which may attenuate the great clinical significance. Nevertheless, this study represents the initial step to investigate the dynamic associations among positive/negative maternal emotions, child behaviors and sleep patterns during early childhood.

The mechanisms underlying the associations between maternal prenatal and postpartum emotions and offspring sleep during childhood remain unclear. Maternal mood disturbances may increase the likelihood of insecure attachment bonds between mother and infant , which contribute to the onset and perpetuation of sleep disorders in early childhood (Benoit, Zeanah, Boucher, & Minde, 1992). In contrast, positive psychological adjustment during pregnancy may contribute to enhanced resilience, parenting cognition and bedtime involvement of their children, thus predicting better child sleep during early childhood (Johnson et al., 2014). Postpartum depressive emotions are also associated with poor sleep quality in women (Park, Meltzer-Brody, & Stickgold, 2013). Given the behavioral and neural concordance in parent-child dyadic sleep patterns(Lee, Miernicki, & Telzer, 2017), maternal emotions may exert an indirect impact on child sleep.

Our study explored the role of children’s behavioral problems as a novel mediator between the association between maternal emotions and child sleep. We found that maternal perinatal depressive emotions and decreased happiness levels significantly predicted overall internalizing and externalizing behavioral problems in children around ages 5–6 years. In turn, more behavioral problems among children significantly increased the risk for sleep problems in the present study. This newly identified pathway is supported by prior findings in the literature. Perinatal exposures to maternal mental health have been documented to correlate with the later development of emotional and mental pathologies in children (Lewis, Galbally, Gannon, & Symeonides, 2014). Specifically, maternal prenatal emotions have a direct impact on offspring behavioral outcomes probably through intergenerationally shared genes (Hannigan et al., 2018) and fetal programming effect (Golub et al., 2016). Because of social learning and attachment problems and environmental stress, postnatal exposure to maternal depressive emotions may also increase the risk for both internalizing and externalizing behavioral problems in childhood (Cuijpers et al., 2015; Goodman & Gotlib, 1999). Consistent with our findings, cross-sectional and longitudinal studies suggest that emotional and behavioral problems, particular externalizing difficulties, further predict sleep problems in children from preschool age to early adolescence (Quach et al., 2018; Stein et al., 2001). Additionally, Netsi, Van, and colleagues (2015) reported improvements in infant sleep following mitigated postnatal depression in women and this relationship was partially explained by infant temperament (specifically, high reactivity). Despite the differences in populations and research design, the study by Netsi et al. (2015) supports our conceptualization of children’s behavioral problems as a potential mediator. However, the directionality of the association between child sleep problems and internalizing and externalizing behavioral problems are not conclusive. Using a longitudinal design, researchers found a bidirectional association between sleep and emotional and behavioral problems across childhood and early adolescence (Quach et al., 2018; K. E. Williams, Berthelsen, Walker, & Nicholson, 2017). Additionally, studies have found that maternal mood and child behavioral problems may have a reciprocal relationship and influence each other (Nicholson, Deboeck, Farris, Boker, & Borkowski, 2011). While Nicholson et al’s (2011) study, which included cross-sectional assessments of maternal emotions and children’s behavior, results in uncertain directionality of the maternal emotion – children’s behavior relationship, our maternal postnatal emotions refers to the first three months after childbirth, and child behavior assessments 5 years later. Thus, such temporal ordering may suggest a directionality of maternal emotions on children’s behavior. Nevertheless, results of this study need to be interpreted with caution. Future research is needed to understand the dynamic interplay among dyadic emotions, sleep and behaviors.

Several important limitations of the current study should be noted. First, differences in maternal reporting of children’s behaviors could reflect memory and recall bias since women’s mood and affect were queried approximately 4 to 6 years after the index pregnancy and delivery. In addition, women’s current depressive emotions were not measured in this study, although current depression might be a significant factor in maternal responses about remembered antenatal and postpartum mood, particularly in women whose children were experiencing behavioral difficulties during the time when the study was done. There is strong evidence that depression is associated with biases in explicit memory. Specifically, depressed individuals often report overly general and more negative memories, as opposed to specific and positive memories, compared with non-depressed individuals (A. D. Williams & Moulds, 2007). Thus, women with persistent or worsening depressive symptoms may have been more prone to inaccurate and biased recall in this study. Moreover, a persistently low mood has been shown to affect ratings of infant temperament in a negative way, since women may find caring for a young infant difficult (Netsi et al., 2015). Furthermore, women whose children have sleeping problems may remember their experiences during and after pregnancies more negatively than those whose children are sleeping well. Thus, ascertainment bias may be a factor in this study. Second, maternal emotions were assessed by researcher-designed single item measures which, while practical and efficient, lacked established psychometric properties such as validity and reliability. However, previous studies have demonstrated high correlations of single-item scales of happiness and depression with established instruments of the same constructs (Abdel-Khalek, 2006; St John & Montgomery, 2009; Zimmerman et al., 2006). Future research using validated instruments is warranted. Third, all measures were completed by a single reporter, with no objective verification of sleep patterns and sleep disruptions. Depressed mood was also assessed based on the subjective report. Future studies should include an objective measure of sleep patterns (e.g., actigraphy) and depression using a clinical diagnostic tool. Fourth, the causal pathway of the link between maternal mood and children’s sleep problems is unclear. Maternal emotional wellbeing and children’s sleep patterns are interconnected, which complicated the analysis of potential contributors and mechanisms of action. Sleep disruptions and behavioral problems in infants and children have been found to predict maternal sleep quality depressive symptoms, greater parenting stress, and poorer day-time functioning (Hughes, Gallagher, & Hannigan, 2015; Meltzer & Mindell, 2007; Nicholson et al., 2011). Additionally, the relationship between child sleep and behavioral problems may be bidirectional (Quach et al., 2018). However, the current study was not designed to discern the direction of the dynamic relationships among maternal mood, child behavioral problems and sleep disruptions. Fifth, we found worse child sleep among those living in urban areas, indicating a potential cultural phenomenon of child sleep. We did not examine the role of urbanicity in the relationship between maternal mood and child symptoms due to the scope of this study. Other limitations include not having information on a variety of factors that are likely to contribute to children’s sleep and behaviors. Factors include diagnostics on autism and neurodevelopmental disorders, and sibling sleep information. However, for the latter, the sibling rate is low given the One Child Policy in China at the time. We also did not specify the type of birth complications. Future research should consider these factors to clarify these complex associations.

To our knowledge, this is the first study to be conducted in China that examines the association between maternal mood and children’s sleep behavior. It is also among the few studies to examine maternal postnatal distress and sleep in childhood rather than in infancy. Poor sleep is a prevalent problem in Chinese children (Li et al., 2010), and understanding possible contributors could inform prevention and intervention efforts (Hiscock et al., 2014). Our results suggest that maternal unhappiness during pregnancy and postnatal depressive emotions are potential risk factors for children’s sleep problems in the Chinese population. These results supplement emerging research that demonstrates that maternal mood disturbances may have detrimental effects on children’s growth and development (Avalos, Flanagan, & Li, 2019) and may lead to more serious problems like childhood psychopathology (O’Donnell, Glover, Barker, & O’Connor, 2014) and behavioral problems (Leis et al., 2014). The findings also suggest that a child’s poor sleep patterns can potentially have a negative impact on or worsen a mother’s extant depressive symptoms, which may create or maintain a cyclical relationship in which both the mother and the child negatively affect one another (Swanson et al., 2010). Psychosocial interventions can be effective in preventing and treating postnatal depression (Ride et al., 2016) and improving children’s sleep behaviors, which in turn, could improve maternal mood (Kempler, Sharpe, Miller, & Bartlett, 2016). Hence, it is imperative to provide psychosocial support to expecting and new mothers to maintain emotional well-being and potentially optimize children’s neurobehavioral outcomes, including sleep. Given that the category of childhood sleep problems is a risk factor for many negative outcomes, including childhood obesity and behavioral problems (Jianghong Liu et al., 2016), it is important to develop policies to support robust maternal physical and emotional health during the prenatal and postnatal periods. Primary, secondary, and tertiary prevention programs, may likewise reduce the likelihood of maternal mood disturbances (Hendricks & Liu, 2012). Participation in these programs may prevent childhood sleep problems. Incorporating emotional support as part of prenatal and postnatal care may promote stability in maternal emotional functioning and help mitigate sleep problems in early childhood.

Highlights.

• Maternal depressive emotions during the pre- and postnatal periods were associated with an increase in children’s sleep problems

• Increased maternal happiness during pregnancy was associated with a decrease in children’s sleep problems

• Children’s behavioral problems significantly mediated the aforementioned relationships

• Maternal unhappiness during pregnancy and postnatal depressive emotions are potential risk factors for children’s sleep problems in the Chinese population