Longitudinal relations between sleep and cognitive functioning in children: Self-esteem as a moderator

Summary

Short and poor-quality sleep disrupt cognitive functioning, yet associations vary across studies, underscoring the importance of examining individual differences and moderators of risk. Utilizing a multi-method, two-wave longitudinal design, we examined self-esteem as a moderator of relations between actigraphy-derived sleep duration (minutes) and quality (efficiency, long-wake episodes) and children’s cognitive functioning 1 year later. During the first study wave (T1), participants were 243 children (47% female) with a mean age of 10.4 years (SD = 8.0 months). The sample was representative of its community, with 37% identifying as Black/African American and 63% White/European American. Children completed a self-esteem measure and wore actigraphs for seven consecutive nights. Participants returned to the lab 1 year later and completed a standardized assessment of cognitive functioning. Results indicated that self-esteem moderated longitudinal associations between sleep quality and cognitive functioning. Specifically, children with both better sleep quality and higher self-esteem performed better relative to other children in the sample.

1 |. INTRODUC TION

Multiple reviews and meta-analyses have shown that short and poor-quality sleep are associated with lower cognitive performance in children (Astill et al., 2012; Short et al., 2018). While the majority of studies have been cross-sectional, limited longitudinal research suggests that poor sleep at an earlier age is predictive of worse cognitive outcomes over time (Bub et al., 2011).

While many studies have shown significant associations between sleep and cognitive functioning, the magnitude of relations varies, suggesting that individual differences may be operating as moderators. In the literature, individual and group differences have been examined as moderators of risk, including socioeconomic status (SES), and psychological factors including coping and emotion responses (Staton et al., 2014; Wang & Yip, 2020), suggesting that poor sleep is especially detrimental for some individuals.

Self-esteem and other related constructs (self-concept; self-efficacy; self-confidence) have been associated with positive outcomes, including cognition and academic achievement (Alves-Martins et al., 2002). Sleep is directly and indirectly related to self-esteem (Lemola et al., 2013), which may enhance individuals’ sense of competence and protect children from the detrimental effects of poor sleep. Moderation models expand upon direct links between sleep and cognitive functioning, and examination of conjoint interactive effects may identify vulnerability and protective factors. For example, poor subjective sleep quality over time is associated with worse cognitive functioning; however, this risk is particularly detrimental to girls and African American children (Bub et al., 2011). Additionally, optimal physiological regulation and greater sleep efficiency together protected low-income children from worse cognitive outcomes (Staton et al., 2014). These studies highlight the need to investigate interactions between sleep and individual differences as predictors of cognitive outcomes.

We examined self-esteem as a moderator of relations between children’s sleep and their cognitive functioning 1 year later, and expected that higher self-esteem would be protective against poor cognitive functioning otherwise associated with short and poor-quality sleep.

2 |. METHODS

2.1 |. Participants

Participants partook in a longitudinal investigation of sleep and health in youth (Auburn University Sleep Study). Data were collected during the second and third waves of the larger study in 2010–2011 (referred to as T1 in this paper) and 2011–2012 (T2) with excellent retention (98%). Families were recruited through letters sent home from rural and semi-rural schools in Alabama, USA. Based on exclusion criteria, children were not diagnosed with a sleep or developmental disorder according to mothers’ reports; children with attention deficit/hyperactivity disorder were excluded from analysis (n = 37). At T1, the analytic sample included 243 children (M_age = 10.4 years; SD = 0.66), was heterogeneous with regard to sex (47% female), race (37% Black/African American, 63% White/European American) and SES, and was demographically representative of the recruitment area. SES was assessed using income-to-needs ratio (U.S. Department of Commerce, 2010). At T1, 66% of families in the study were living in or near poverty, 25% were lower middle class, and 9% were middle class and above.

2.2 |. Procedures

The university’s institutional review board approved study procedures, and children and their parents gave assent and consent to participation. At T1, children were asked to wear an actigraph on their non-dominant wrist for seven consecutive nights during the school year. Children wore actigraphs an average of 5.8 nights (SD = 1.5), excluding nights on which they took medication for allergy or acute illness. Participants then visited a campus laboratory, where parents reported on demographics and child health, and children reported on their self-esteem. During the visit, children’s height and weight were recorded using a wall-mounted stadiometer and Tanita digital scale, and body mass index (zBMI) was derived. Approximately 1 year later (M = 336.2 days; SD = 33.9), children revisited the laboratory to complete a test of cognitive functioning.

2.3 |. Measures

2.3.1 |. Sleep (T1)

Sleep was examined with Octagonal Basic Motionlogger actigraphs (Ambulatory Monitoring), and was scored using the Sadeh algorithm (Sadeh et al., 1994) in Action W2 (Ambulatory Monitoring). Three parameters were derived and assessed; both duration and quality of sleep and their definitions are consistent with the manual that accompanied the software. Minutes comprises the total number of minutes slept. Sleep efficiency represents the percentage of epochs scored as sleep during the total sleep period from actigraphically determined sleep onset until wake. Long-wake episodes (LWE) are a count of the number of wake episodes ≥ 5 min. Following common principles, actigraphy data were included in analyses for participants who had at least 5 nights of data after exclusion of nights they took medication. Data were treated as missing for participants with fewer than 5 nights (14.8% of sample); children with missing actigraphy data were not removed from the analytic sample, and missing data were handled statistically. Sleep minutes, efficiency and LWE were stable across nights (α = 0.78; 0.89; 0.87, respectively); the mean across nights for each variable was used in analyses.

2.3.2 |. Self-esteem (T1)

Children reported their self-esteem using the six-item global self-esteem subscale of Harter’s (1985) Self-Perception Profile for Children, which is a validated measure (α = 0.70 with the current sample). Items assessed how much children agreed with statements about themselves, using a four-point scale with opposing statements (1 = really true on initial statement to 4 = really true on opposing statement), such as, “Some kids like the kind of person they are” versus “Other kids wish they were different”.

2.3.3. |. Cognitive functioning (T2)

Children’s cognitive functioning was assessed using the Brief Intellectual Ability (BIA) scale of the Woodcock Johnson Tests of Cognitive Abilities III, a widely used measure of cognitive functioning (Woodcock et al., 2001). The BIA, a measure of overall cognitive functioning tapping into both crystallized and fluid intelligence, is based on an assessment of verbal comprehension, fluid and categorical reasoning, and perceptual processing. The W score, which provides an individual’s standardized, theory-scaled deviation from a criterion score, was used in analyses.

2.3.4 |. Control variables (T1)

Covariates included zBMI and parent-reported child sex, race and SES.

2.4 |. Plan of analysis

Values for study variables that exceeded 4 SDs were recoded to the highest or lowest values within 4 SDs to reduce outlier effects. None of the primary variables was skewed.

A series of path models was fit in AMOS 23 (Arbuckle, 2014), and full-information maximum likelihood was used to handle missing data. Missing data on primary study variables were relatively low (10%−18%) and within acceptable parameters (Enders & Bandalos, 2001).

Models were fit separately for each sleep variable to investigate unique moderation effects of self-esteem on cognitive functioning. Potential confound variables (race, sex, zBMI, SES) that were significantly associated with primary study variables were covaried. Mean-centred covariates, self-esteem and sleep parameters were examined for associations with cognitive outcomes. Two-way interactions between self-esteem and individual sleep parameters were added to the models. Significant interactions were plotted at high and low (±1 SD) levels of sleep and self-esteem, and simple slopes were tested to examine whether they were significantly different from zero. Models were considered an acceptable fit if they satisfied at least two of the three following criteria: χ²/df < 3, CFI > 0.95 and RMSEA ≤ 0.06 (Hu & Bentler, 1999).

3 |. RESULTS

Descriptive statistics and correlations among variables are reported in Table 1. Independent samples t-test indicated that White children have higher cognitive functioning scores than Black children (t₁₉₇ = 3.86, p < .001; M_White = 509.7, SD = 9.7; M_Black = 503.8, SD = 11.6). No significant sex differences were found for primary variables.

TABLE 1.

Descriptive statistics and correlations among study variables

	1	2	3	4	5	6	7	8	9
1. Sex	—
2. Race/ethnicity	−0.05	—
3. SES	0.12	−0.41*	—
4. zBMI	0.02	0.15*	−0.09	—
5. Sleep minutes	−0.09	−0.08	0.07	−0.19*	—
6. Sleep efficiency	−0.09	0.09	−0.003	−0.10	0.73*	—
7. LWE	0.07	−0.09	0.03	0.09	−0.61*	−0.94*	—
8. Self-esteem	−0.11	0.05	0.02	−0.03	−0.02	0.07	−0.12	—
9. Cognitive functioning	−0.01	−0.27*	0.26*	0.01	0.13	0.16*	−0.17*	0.20*	—
Mean	—	—	1.60	0.67	446.18	88.75	3.40	19.59	507.30
(SD)	—	—	(0.96)	(1.16)	(46.33)	(6.62)	(2.15)	(3.78)	(10.90)

Sex (0 = girl, 1 = boy). Race/ethnicity (0 = White, 1 = Black). 446.18 min = 7.44 hr. Cognitive functioning based on Woodcock Johnson-III Test of Cognitive Functioning (represented in W scores) at T2; all other variables are based on data collected at T1.

LWE, long-wake episodes; SES, socioeconomic status; zBMI, standardized body mass index.

^*p < .05.

^**p < .01.

^***p ≤ .001.

Findings from final path models are presented in Table 2. Covariates accounted for 9.7% of the unique variance in cognitive functioning. Sleep quality and self-esteem, but not sleep minutes, were associated with cognitive functioning.

TABLE 2.

Unstandardized and standardized coefficients of associations linking sleep, self-esteem and cognitive functioning in children

	Sleep minutes				Sleep efficiency				LWE
	B	SE	β	R 2	B	SE	β	R 2	B	SE	β	R 2
Sex	−0.52	1.43	−0.02	9.7%	−0.62	1.41	−0.03	9.7%	−0.87	1.41	−0.04	9.7%
Race	−4.33*	1.64	−0.19		−5.38*	1.63	−0.24		−5.43*	1.63	−0.24
SES	1.90*	0.87	0.17		1.94*	0.85	0.17		1.99*	0.85	0.18
zBMI	0.43	0.66	0.05		0.47	0.62	0.05		0.50	0.62	0.05
Sleep	0.02	0.02	0.10	14.1%	0.22*	0.12	0.14	16.9%	−0.76*	0.35	−0.15	16.9%
Self-esteem	0.60*	0.19	0.21		0.52*	0.19	0.18		0.52*	0.19	0.18
Sleep x Self-esteem	0.006	0.004	0.10	15.4%	0.07*	0.03	0.15	19.2%	−0.22*	0.09	−0.17	19.4%

Path coefficients reported are from the final model. SE = standard error. Sex (0 = girl, 1 = boy). Race/ethnicity (0 = White, 1 = Black). Cognitive functioning based on Woodcock Johnson-III Test of Cognitive Functioning (represented in W-scores) at T2; all other variables are based on data collected at T1.

LWE, long-wake episodes; SES, socioeconomic status; zBMI, standardized body mass index.

^*p ≤ .05.

^**p < .01.

^***p < .001.

Supportive of moderation effects, interactions between sleep quality (efficiency, LWE) and self-esteem predicted cognitive functioning. Children with low self-esteem had similar cognitive performance scores regardless of their sleep quality (Figure 1; model fit: χ²/df = 1.1, CFI = 0.97, RMSEA = 0.02, ns). However, children with higher self-esteem in conjunction with greater sleep efficiency had significantly higher cognitive performance scores than other children in the sample. At higher levels of self-esteem, BIA scores were higher for children with higher sleep efficiency (predicted M = 514.6) compared with those with lower sleep efficiency (M = 508.3), a difference of 0.57 SD.

FIGURE 1.

Sleep efficiency predicting cognitive functioning 1 year later at lower and higher levels of self-esteem (±1 SD). Cognitive functioning is based on Woodcock Johnson-III Test of Cognitive Functioning (represented in W-scores)

An interaction between LWE and self-esteem yielded a similar pattern of effects (Figure 2; model fit: χ²/df = 1.2, CFI = 0.94, RMSEA = 0.03, ns). For children with higher self-esteem, those with fewer LWE had BIA scores (M = 514.9) 0.62 SD higher than their counterparts with more LWE (M = 508.1). Children with lower self-esteem had relatively lower levels of cognitive functioning irrespective of their LWE. Models for sleep efficiency and LWE explained 19% of the variance in BIA scores, of which about 2% was contributed uniquely by interactions of sleep and self-esteem. Finally, self-esteem did not moderate relations between sleep minutes and cognitive functioning (model fit: χ²/df = 0.78, CFI = 1.00, RMSEA = 0.00, ns).

FIGURE 2.

Long-wake episodes (LWE) predicting cognitive functioning 1 year later at lower and higher levels of self-esteem (±1 SD). Cognitive functioning is based on Woodcock Johnson-III Test of Cognitive Functioning (represented in W-scores)

4 |. DISCUSSION

Contributing to the growing literature on sleep and cognitive functioning, we examined whether children’s self-esteem at age 10 years moderated associations between objective assessments of children’s sleep duration and quality and their cognitive functioning 1 year later. Consistent with expectations, higher self-esteem in conjunction with more optimal sleep quality (efficiency, LWE) was predictive of the best performance on cognitive functioning over time. At less optimal sleep quality, cognitive functioning was lower for all participants regardless of their self-esteem. Findings support the importance of examining individual differences towards a better understanding of relations between sleep and well-being in youth.

Results regarding main effects are consistent with previous studies that have identified associations among multiple indices of actigraphic sleep quality and cognitive functioning (Bub et al., 2011; Staton et al., 2014). The lack of main or interactive effects of sleep duration underscores the need to assess the role of sleep quality parameters.

Results should be interpreted within the context of the sample, including the age of participants, their locale and high level of economic disadvantage. SES is a particularly influential variable, as findings of this study indicate; note, however, that SES was covaried in all models. These factors need to be considered as limitations to generalizability and should be considered tentative until replicated. Although the longitudinal design of the study expands on relations between sleep, individual differences and cognitive outcomes, inclusion of prior cognitive performance in models would strengthen causal interpretations.

The full models accounted for 19% of the variance in cognitive functioning. Moderation effects for sleep quality parameters accounted for 2% of unique variance in cognitive functioning, a magnitude consistent with other effects reported in the literature (Astill et al., 2012; Short et al., 2018). These interactions indicate that youth with better quality sleep and greater self-esteem benefited in cognitive functioning scores that were 0.6 SD (6.3–6.8 points) higher than youth with worse sleep quality, as well as lower self-esteem.

Recent research suggests that sleep (Gruber et al., 2016) and self-affirmation (Cohen et al., 2009) interventions are effective at improving academic outcomes. Our study findings suggest that incorporating self-esteem with sleep intervention protocols could be beneficial in improving youth well-being.

Abbreviations:

BIA

Brief Intellectual Ability

LWE

long-wake episodes

SES

socioeconomic status

zBMI

standardized body mass index

DATA AVAILABILITY STATEMENT

The data are not publicly available due to privacy or ethical restrictions. The data that support the findings of this study are available from the corresponding author upon reasonable request.