Maternal Stress, Sleep, and Parenting

Abstract

Associations between stress, sleep, and functioning have been well-established in the general adult population, but not as well established in the specific sub-population of interest here, parents. To advance understanding of how maternal sleep is linked with both mothers’ experiences of stress and their parenting, this study used actigraphic and mother-report measures of sleep, observed and mother reports of parenting, and measures of multiple stressors of relevance. In a community sample of mothers of toddlers (N = 314, child age M = 2.60, SD = 0.07 years), maternal stress was indexed with a cumulative risk score that combined sociodemographic risks and common parent stressors, including household chaos, role overload, parenting hassles, child misbehavior, negative life events, and lack of social support. We found that mothers who experienced shorter, later, and more variable sleep experienced higher levels of stress as indexed by the cumulative risk index. In addition, those with higher stress required longer to fall asleep and perceived more sleep problems. We also found that actigraphic measures of poor and insufficient maternal sleep were associated with less observed positive parenting, even when controlling for the cumulative risk index and maternal age, employment, and family size. Mothers who required longer to fall asleep also reported more dysfunctional parenting, with the same statistical controls. The findings, coupled with research showing that sleep is amenable to intervention, suggest that parental sleep may ultimately prove to be a useful intervention target for promoting positive parent involvement and responsiveness.

Parenting is often considered to be the most important sub-system of the broader social developmental system (Bronfenbrenner, 1979). Positive parent behaviors, like warmth and responsiveness, have been shown to contribute to the development of positive child behavior (Kochanska & Aksan, 1995), whereas negative parent behaviors, such as overreactivity, have been shown to be associated with children’s externalizing behavior problems (O’Leary, Slep, & Reid, 1999). For the sake of supporting positive child development, it is important to identify factors that may influence variations in parenting behavior. One leading candidate is parental stress. Stressed parents tend to use more harsh parenting and less positive parenting (Matheny, Wachs, Ludwig, & Phillips, 1995). Prior research has also established associations between stress and sleep (Leproult, Copinschi, Buxton, & Van Cauter, 1997) and between sleep and cognitive functioning and emotion processing (Harrison & Horne, 2000). The associations between sleep and cognitive and emotional functions might generalize to the specific context of the functioning of parents. We would expect that sleep deprived parents may display harsher and more reactive parenting and less warm and responsive parenting. Even if parental sleep deficits are associated with parenting deficits, it would be important to know if links between sleep and parenting are independent of, or supplemental to, the general stress the parent experiences in accounting for parenting behavior. In a community sample of mothers of toddlers, the present study assessed mothers’ sleep with actigraphic and mother-reported measures. It also assessed mothers’ parenting with observed and mother-reported measures. The study aimed to 1) test the links between maternal sleep and stress, and 2) determine whether maternal sleep is linked with parenting in ways that are similar to, or even surpassing, maternal stress.

A Cumulative Risk Approach to Maternal Stress

Parenting young children can be a source of stress. Immediate stressors, such as a chaotic home or a demanding child, and background risk factors, such as single parenthood or low income, may lead parents to experience demands in excess of their resources (Hobfoll, 2001). Parents facing child demands and other parenting pressures, such as highly chaotic homes, tend to be more overreactive and less positively responsive with their children (Matheny et al., 1995). Child demands and daily pressures may be especially challenging in families with limited economic and social resources. Parents with economic hardships are at increased risk of experiencing other stressors, such as marital discord and intimate partner violence, as well as more child behavior problems (Conger, Rueter, & Conger, 2000). Increased demands with fewer resources could at least partially explain why low-income parents tend to show more ineffective parenting (Conger et al., 2000). Similarly, single parents often receive unhelpful social support and experience more aversive and fewer positive social events than parents with partners, along with feelings of incompetence, embarrassment, and stronger negative reactions to their children (Dumas, 1986), which may explain why single parenthood is also associated with more inconsistent and ineffective parenting (Bank, Forgatch, Patterson, & Fetrow, 1993).

To measure the stress of parenthood, we used a composite of stressors and risk factors. Cumulative risk indexes (CRI’s) have been shown to be more predictive of negative outcomes, relative to the study of a single risk factor (e.g., Sameroff, Seifer, Barocas, Zax, & Greenspan, 1987). Our CRI is not identical to previous CRI’s (e.g., Deater-Deckard, Dodge, Bates, & Pettit, 1998; Deater-Deckard, Li, & Bell, 2016; Sameroff, Seifer, Baldwin, & Baldwin, 1993), but ours has substantial overlap with these. It incorporates measures of parent employment and education, marital status, lack of social support, stressful life events, and household chaos, among other sources of stress. Although there is not consensus about which stress variables should be examined, previous research has shown that the precise combination of stressors may not be as important compared to the mere choice to consider risk in aggregate (Sameroff et al., 1993).

There are several ways to compute a CRI using dichotomized or dimensional risk indicators (Evans, Li, & Whipple, 2013). Because we did not have a rationale to dichotomize continuous stress measures, we used a dimensional CRI. Our composite measure aggregated stressful life events, child misbehavior, parenting pressures, and sociodemographic risks on a standardized (z-scored) metric. It is comparable to the three domains of the widely-used Parenting Stress Index (Abidin, 1990): feelings of distress, difficult child behavior, and dysfunctional parent-child interactions. This study extends prior research by considering a wide spectrum of stress experienced by mothers, and mothers’ sleep deficits, in relation to parenting.

Adult Stress, Sleep Deficits, and Functioning

Bidirectional links between adult stress and sleep deficits are well established. Experimentally induced stress and real-world stressors can reduce sleep duration and quality (Akerstedt et al., 2002). Poor sleep, in turn, increases morning cortisol and cholesterol levels, reflecting a physiological stress response involving over-activation of the major neuroendocrine stress systems, the autonomic sympatho-adrenal system and the hypothalamic-pituitary-adrenal axis (Ekstedt, Akerstedt, & Soderstrom, 2004; Meerlo, Sgoifo, & Suchecki, 2008). Sleep loss also reduces the rate at which cortisol concentrations decline throughout the day, resulting in elevated levels of cortisol in the evening (Leproult et al., 1997). When a sleep deprived adult faces demanding tasks, and is not allowed to rest, the activity of the neuroendocrine systems does not return to normal throughout the day (Meerlo et al., 2008).

Research on sleep deprived medical professionals responding to disruptive and unexpected events in their work days suggests that the heightened activation of stress systems following sleep loss is associated with stronger reactions to challenges compared to milder appraisals of these challenges when well-rested (Zohar, Tzischinsky, Epstein, & Lavie, 2005). Sleep deficits also limit the regulatory functioning of the prefrontal cortex, resulting in increased negative emotionality, impulsivity, and sensitivity to low-level stressors (Minkel et al., 2012), as well as poorer performance on tasks requiring flexible thinking and management of competing demands (Harrison & Horne, 2000). By extension, it would be reasonable to expect that when sleep deprived parents encounter stressors, they may be more likely to perceive them as stressful, and they may struggle to flexibly adapt to child needs and respond in effective ways. In this study, sleep deficits are regarded as part of a stress cycle in which other forms of stress could impede sleep, and disrupted sleep could increase subsequent perceptions of stress in response to challenges and demands, potentially resulting in more negative and less positive parenting.

Maternal Sleep Deficits and Parenting

There is abundant and important research on the sleep and functioning of mothers of infants (Sinai & Tikotzky, 2012) and mothers of chronically ill children (Meltzer & Walsh, 2013). Research on the sleep of parents of typically developing children after infancy is comparatively scarce, but growing (Meltzer & Montgomery-Downs, 2011; Meltzer & Westin, 2011). We know that child night awakenings are related to maternal night awakenings, which predict maternal negative affect, distress in the parenting role, and caregiving overload (Meltzer & Mindell, 2007; Mindell, Sadeh, Kwon, & Goh, 2015). We also know that maternal insomnia is linked with more household chaos and child neglect (Gregory et al., 2012). This research using parent-report measures suggests that sleep deficits may be linked with stress and parenting. The present study extends this work by using actigraphy to measure sleep.

The Present Study

We think of sleep deficits as part of a stress process, proximally affecting daily functioning (e.g., parenting) in a way that is similar to, or even surpasses, other standard stress processes. To test this, we collected multiple measures of stress, sleep, and parenting in a community sample of mothers of toddlers. Toddlerhood is an ideal period for this research because it is characterized by increasingly challenging child behavior. With increased mobility and independence, toddlers can substantially add to parent stress (Fagot & Kavanagh, 1993). Toddlerhood is also characterized by changes in sleep patterns, including decreases in daytime sleep (Acebo et al., 2005) and the total amount of sleep (Galland, Taylor, Elder, & Herbison, 2012), which are likely associated with changes in maternal sleep (Mindell et al., 2015).

We expected maternal sleep deficits and stress to be positively correlated. Additionally, given the proximal effect sleep deficits have been shown to have on cognitive and emotional functioning, we hypothesized that parental sleep deficits would be associated with less positive parenting and more dysfunctional parenting, even when the CRI and other covariates were statistically controlled. Just as mothers returning home from stressful work days show less positive parenting compared to the amount of positive parenting shown on less stressful days (Repetti & Wood, 1997), we hypothesized that mothers who experienced poor or insufficient sleep would similarly show less positive parenting compared to their more well-rested peers. Notably, positive parenting was observed in the present study during the hour before the toddler’s bedtime, a period that is commonly challenging for parents and is associated with fewer child sleep difficulties if mothers are emotionally available and sensitive (Teti, Kim, Mayer, & Countermine, 2010). We expected sleep deprived mothers to report more dysfunctional parenting and show less positive parenting during the critical bedtime period. A sole focus on dysfunctional parenting without consideration of supportive parenting could be uninformative because supportive parenting has been shown to have a positive influence on child sleep (Teti et al., 2010), and on child adjustment and academic performance, even when harsh parenting is controlled (Pettit, Bates, & Dodge, 1997).

Method

Participants

We studied two community samples as the initial part of a multi-site longitudinal study, The Toddler Development Study. Recruitment was primarily through a database using county birth records and community outreach efforts, such as through the local Head Start agency and the Housing Authority. Advertisements were also used (e.g., postcard mailings, flyers in common areas, and social media postings). Given this extensive recruitment procedure, computing the number of families recruited was not feasible. All families who were willing to participate were included. The number of families who declined was not recorded, but our recruiters typically received consent from the mothers once contact was made. Compensation ($100 total, approximately $25 per hour of participation) was provided, and transportation offered. The local Institutional Review Boards approved the study (protocol # 0811000120), and mothers provided informed consent. The sample included 314 mother–child pairs. Table 1 shows the sample sizes for each site. Eight percent of the full sample were unwilling to have observers visit their home so the observed positive parenting index was missing for that portion of the sample, and 16% of the full sample was missing actigraphy data due to refusal to wear actigraphs (n = 6), actigraphic equipment failure (n = 42), or lost actigraphs (n = 2). For the 84% of the sample with actigraphic data, mothers provided, on average, 6.76 days of data (SD = 1.46 days), in accordance with guidelines to collect at least five days of actigraphy data for reliable estimates (Acebo et al., 1999). Mothers with usable actigraphic data had significantly higher SES ratings (48.51 ± 13.13) compared to mothers without actigraphy data (43.30 ± 13.86), t (314) = −2.45, p = 0.02. However, there was no significant difference on SES for mothers with and without parenting observed before the child’s bedtime (t (314) = −1.56, p = 0.12).

Table 1.

Descriptives for Stress, Sleep, and Parenting Composites

Site	Composite	N (%)	M	SD	Lower	Upper
Site 1a	Cumulative Risk Index (z)	207 (99)	0.02	1.00	−1.89	3.83
	General Sleep Deficit (z)	175 (84)	−0.08	0.99	−3.05	3.98
	Observed Positive Parenting (z)	195 (93)	−0.07	1.00	−4.09	1.12
Site 2b	Cumulative Risk Index (z)	105 (100)	−0.01	0.94	−2.34	2.53
	General Sleep Deficit (z)	89 (85)	0.16	1.01	−1.58	3.16
	Observed Positive Parenting (z)	94 (90)	0.15	0.98	−4.00	1.15
Totalc	Cumulative Risk Index (z)	312 (99)	0.01	0.98	−2.34	3.83
	General Sleep Deficit (z)	264 (84)	0.00	1.00	−3.05	3.98
	Observed Positive Parenting (z)	289 (92)	0.00	1.00	−4.09	1.15

Note. “Lower” is the lower bound of the 95% confidence interval; “Upper” is the upper bound of the 95% confidence interval. General sleep deficit is the composite of short, late, and variable sleep. Observed positive parenting is the composite of maternal responsivity, involvement, and responsiveness observed during the hour before the target child’s bedtime.

^aSite one denominator for % completion was 209.

^bSite two denominator for % completion was 105.

^cTotal sample was 314.

Across the two samples, mothers were 21 to 50 years old (M = 32.67, SD = 5.04), and the target children were, on average, 2.60 years old (SD = 0.07; range = 2.45 – 2.89; 53% female). Based on the Hollingshead (1975) index, the sample was largely middle class (range = 13 – 66, M = 47.71, SD = 13.36). The two sites did not significantly differ in terms of maternal age [M (SD) = 32.68 (4.68) and 32.74 (5.64)] or family SES [M (SD) = 47.39 (13.95) and 48.36 (12.11) for Sites 1 and 2, respectively]. Thirty-one percent of mothers did not work outside the home, and 28% worked 40 or more hours per week (M (SD) = 19.41 (18.01) hours). Ninety percent of mothers were Caucasian, 4% were Hispanic, 3% were African American, 1% were Asian American, and 2% identified as mixed race, American Indian, or other. Eighty-five percent were married, remarried, or cohabitating with their romantic partner, and the remaining 15% were single/never married, divorced, or separated. In 30% of the families, the target child was the only child. Sixty percent had one or two other children in the home, while 10% had four or more children. In terms of psychological adjustment, 5% of the mothers endorsed clinical levels of depressive symptoms on the Center for Epidemiological Studies Depression Scale (CES-D; Radloff, 1977) by scoring at or above the stringent cut-off of 27 used by Zich and colleagues (1990; M = 17.58, SD = 5.21).

Procedure

Mothers completed questionnaires on demographics, stress, and parenting, and mother-child dyads participated in two home visits and a lab visit. At the initial home visit, the mother and the target child were each given actigraphs to wear for one week. The present study used only the maternal sleep data. Within the next several days, research assistants conducted a second home visit to observe parent-child interaction in the hour prior to the child’s bedtime. One week later, mothers visited the lab to complete a battery of tasks.

Measures

Stress.

Mothers completed the Changes and Adjustments Questionnaire (CAQ; Dodge, Pettit, & Bates, 1994) to rate a list of 16 stressful life events potentially experienced in the past year, such as moving or death in the family, as either 0 = did not happen, 1 = happened and had a minor effect on the family, or 2 = happened and had a major effect on the family. The list of stressful life events is based on previous scales (e.g., Johnson & McCutcheon, 1980), with particular emphasis on those that would be most relevant for families of young children. The sum was used in all analyses (M = 2.34, SD = 2.27, α = .55).

Mothers completed the Social Support scale (Procidano & Heller, 1983) to rate experiences of support and companionship, as well as lack of support, on a 0 to 4 Likert scale, ranging from never to often. The 11-item lack-of-support scale was used to assess the frequency of unwanted advice or intrusion, the failure of others to provide help, others’ unsympathetic or insensitive behavior, and experiences of social rejection or neglect. The average lack of support per item was used in all analyses (M = 1.00, SD = 0.63, α = .89).

To measure child misbehavior that would likely contribute to parental stress, the externalizing behavior scale of the Child Behavior Checklist (CBCL 1 ½ -5; Achenbach & Rescorla, 2000), the intensity scale of Eyberg Child Behavior Inventory (ECBI; Robinson, Eyberg, & Ross, 1980), and the Child Misbehavior Scale (Pinderhughes, Dodge, Bates, Pettit, & Zelli, 2000) were used. The CBCL externalizing scale sums aggression and attention problems, with items rated as 0 = not true, 1 = somewhat or sometimes true, and 2 = very true or often true (M=11.86, SD=7.29, α = .91). The ECBI intensity scale’s 36 items assessed the frequency of child oppositional behaviors, such as “dawdles in getting dressed” and “argues with parents about rules,” on a 1 to 7 Likert scale, ranging from never to always (M=103.93, SD=22.79, α = .89). The Child Misbehavior Scale sums the frequency of 12 types of child misbehavior, including “tantrums” and “hits/bothers adults”, in the past six months. Mothers were asked to state whether each behavior 0 = does not happen, 1 = happens sometimes, or 2 = happens a lot (M = 7.31, SD=3.01, α = .72). Two percent of the sample surpassed the CBCL clinic cut-off (T = 70), and 10% surpassed the ECBI clinical cut-off (T = 60), which is below the expected percentage of cases falling within clinical range.

Mothers also completed the Parenting Daily Events scale (Crnic & Greenberg, 1990), from which we used the sum of the 16 parenting tasks subscale items. Mothers rated the frequency (1 = never to 5 = constantly) and intensity (1 = no hassle to 5 = big hassle) of 8 common parenting hassles, such as sibling conflicts and errands (M=40.31, SD=8.39, α = .81). As a measure of household disorder, mothers completed the Confusion, Hubbub, and Order Scale (CHAOS, Matheny et al., 1995), which sums 12 binary (1 = yes/ 0 = no) items, such as “You can’t hear yourself think in our home” and “It’s a real zoo in our home” (M=3.56, SD=2.87, α = 0.79). Mothers’ perceived role overload was measured with the revised, 6-item Reilly Role Overload Scale (Thiagarajan et al., 2006). Items included “I need more hours in the day to do all the things that are expected of me” and “There are times when I cannot meet everyone’s expectations”, all using a 7-point Likert-scale, ranging from never to always. The average per-item rating was used in subsequent analyses (M=4.39, SD=0.99, α = .80).

Cumulative risk index (CRI).

Because these stress measures were significantly correlated (r values ranged from .13 to .62, see Supplement) and because we aimed to measure stress in aggregate, a CRI was formed. Z-scores from each stress variable (CAQ stressful life events, Hollingshead SES (reversed), single parent status (coded 0 = partnered, 1 = single), lack of social support, CBCL externalizing problems, ECBI intensity scale, the Child Misbehavior Scale, hassles from parenting tasks, household chaos, and role overload) were averaged and re-standardized to form a dimensional stress measure (M = 0.01, SD = 0.98, range = −2.34 to 3.83).

Sleep.

Mothers wore a watch-like actigraph on their non-dominant wrist. The MicroMini Motionlogger from Ambulatory Monitoring, Inc. (AMI; Ardsley, NY) recorded minute-by-minute patterns of motor activity. Actigraph data were scored with the Motionlogger Analysis Software Package Action W-2 software (AW2, version 2.6.92), which is the clinical sleep estimation and database program from AMI. The Cole-Kripke algorithm, which has been validated for adults and shown to provide reliable estimates of sleep indexes when averaged over seven nights, was used to reduce the motion data into meaningful sleep variables (Cole, Kripke, Gruen, Mullaney, & Gillin, 1992; Sadeh, Sharkey, & Carskadon, 1994). Mothers also completed a diary record of bedtime, night awakenings, and rise times. Minutes asleep while in bed were based on the bedtime reported in the diary and actigraphic determined end of sleep. Variables concerning activity and awakenings after sleep onset were based on motion recorded by the actigraph using the zero-crossing mode and a moderate sensitivity threshold (Meltzer, Montgomery-Downs, Insana, & Walsh, 2012). A night waking was scored when the activity count was above threshold (50 crossings) for at least five minutes.

Based on analyses with a separate sample (Staples, Bates, Petersen, McQuillan, & Hoyniak, under review), a Principal Components Analysis (PCA) with oblique rotation was used to extract four overarching sleep components: 1) Duration, 2) Variability, 3) Activity, and 4) Timing. Oblique rotation was used because sleep components were intentionally allowed to correlate with each other. Given the shared method variance among the actigraph variables, an oblique rotation allowed the covariance between components to be directly assessed. The four components represent broad dimensions of actigraphy that are often examined in sleep research using single indexes without taking advantage of the benefits of aggregated measures (Meltzer, Montgomery-Downs, Insana, & Walsh, 2012). The sleep components have proved useful, so far, in a study of toddlers’ cognitive development (Hoyniak et al., 2018). We chose to use composites when examining our actigraphy data for several reasons. First, a survey of the sleep literature observed wide differences across studies in the selection and use of actigraph variables (Meltzer et al., 2012). We sought to systematically examine multiple dimensions of sleep without re-running analyses for each of the numerous variables exported from AW2. Additionally, it is well known that aggregate measures provide important measurement advantages over single-variable analysis (Rushton, Brainerd, & Pressley, 1983). Using these composites allowed us to parsimoniously examine important sleep features without having to run, and correct for, as many tests as we would in a comprehensive single variable analysis.

The daily diary and actigraphic variables that compose each of the four sleep components were each averaged across the week of data collection. The weekly-averages of each variable were then standardized (z-scored) based on sample values. The standardized variables were then aggregated together to form the sleep component (e.g., Sleep Duration, which indexes the average duration of sleep across the week of data collection). The component representing Sleep Duration is composed of the mean of z-scored actigraph variables including average time the mother spent in bed each night, the time the mother spent in bed after sleep onset, and the time the mother spent asleep each night, excluding night awakenings (M = −.05, SD = .86). The component representing Sleep Variability is composed of the mean of z-scored actigraph variables including the night-to-night standard deviations of: time of sleep onset, duration of time spent in bed at night, time spent asleep at night not including night awakenings, and the midpoint of sleep (halfway between actigraphically determined true sleep onset and wake time; M = .06, SD = .85). The component representing Sleep Activity is composed of the mean of z-scored actigraph variables including the average time awake after sleep onset, the average minute-by-minute activity level, the average number of night awakenings lasting at least five minutes, the average duration of the longest wake episode after sleep onset, and the average percent of active epochs after sleep onset (M = −.07, SD = .83). The component representing Sleep Timing is composed of the mean of z-scored variables including the average time of midsleep, average time of sleep onset, and the average bedtime reported on the sleep diary (M = .01, SD = .96). These four sleep components explained 82% of the variance in 17 actigraphy variables (Staples et al., under review). The four sleep components demonstrated strong internal consistency (average α = .92, ranging from .89 to .93 across composites). Sleep onset latency, a single variable with low loadings on the four components, was retained as an additional sleep index, measuring the amount of time between diary-reported bedtime and actigraphically determined true sleep (M = 32.58 minutes, SD = 32.50).

Based on the correlations between Sleep Duration, Timing, and Variability (absolute r values ranged from .32 to .53, see Table 2), we sought to create a higher-order composite to potentially reduce the number of analyses with sleep, stress, and parenting. We fit a Confirmatory Factor Analysis (CFA) model to test whether a higher-order construct reflecting general sleep deficit would be justified. We set the mean and variance of the latent factor to zero and one, respectively, which allowed the factor loadings to vary freely. The CFA model converged and fit well, RMSEA = .04 and CFI = .94. The factor loadings with the latent General Sleep Deficit construct were β = .46 for Sleep Variability, β = - .70 for Sleep Duration, and β = .76 for Sleep Timing. The higher-order composite, General Sleep Deficit, was therefore the average of the following sleep components: Sleep Variability, Sleep Duration (reversed), and Sleep Timing. Higher scores reflected worse sleep (more variable, shorter, and later). We averaged the three associated sleep components together, and standardized the average score again to form the General Sleep Deficit higher-order composite.

Table 2.

Bivariate correlations for sleep, stress, parenting, and potential confounding variables

	1	2	3	4	5	6	7	8	9	10	11	12	13
1. General Sleep Deficit	1.00
2. Sleep Duration	−.80**	1.00
3. Sleep Variability	.72**	−.32**	1.00
4. Sleep Timing	.81**	−.53**	.35**	1.00
5. Sleep Activity	.13**	−.16*	.13*	.80**	1.00
6. Sleep Onset Latency	.26**	−.08	.34**	.21**	.37**	1.00
7. PSQI Global Sum	.10	−.01	.12	−.04	.15*	.12	1.00
8. CRI	.32**	−.19**	.31**	.24**	.05	.22*	.34**	1.00
9. Obs. Pos. Parenting	−.23**	.19*	−.28**	−.14*	−.16**	−.29**	−.12	−.28**	1.00
10. PS Dysfunctional	.14*	−.11	.24**	.07	.09	.26**	.15*	.36**	−.18**	1.00
11. Maternal Age	−.09	.03	−.03	−.16*	−.06	−.09	−.10	−.14*	.06	−.01	1.00
12. Hours Worked	.01	−.10	.07	−.14*	−.07	.01	−.02	.07	−.09	.08	.14*	1.00
13. Number of Siblings	.08	−.11	.01	.07	.02	.05	−.12*	.09	−.14*	.06	.09	−.30**	1.00

Note. N range = 240 – 294. Significant associations are bolded.

^**.Correlation is significant at the 0.01 level (2-tailed).

^*.Correlation is significant at the 0.05 level (2-tailed). General Sleep Deficit refers to the higher-order composite of sleep duration, timing, and variability. CRI refers to the dimensional cumulative risk index, comprised of all traditionally studied stressors. PSQI = Pittsburgh Sleep Quality Index. Obs. Pos. Parenting = Observed Positive Parenting composite. PS = Parenting Scale (self-report). Hours Worked refers to the number of hours the mother worked outside of the home each week, ranging from 0 to 60 in this sample.

Mothers also completed the Pittsburgh Sleep Quality Index (PSQI; Buysse, Reynolds, Monk, Berman, & Kupfer, 1989) to assess their own perception of sleep patterns and problems, and use of sleep aids, with a scale ranging from 0 = not during the past month, 1 = less than once a week, 2 = once or twice a week, to 3 = three or more times a week. The global sum, computed from mothers’ report on their low subjective sleep quality, long sleep latency, short sleep duration, low sleep efficiency, high sleep disturbance, use of sleeping medication, and high daytime dysfunction, was used in the present analyses (M = 6.05, SD = 2.43, α = .74). Fifty-five percent of the sample were classified as poor sleepers using the recommended cut-off score of 5.

We had seven total sleep indexes: (1) Sleep Duration, (2) Sleep Timing, (3) Sleep Variability, (4) the General Sleep Deficit composite, (5) Sleep Activity, (6) Sleep Onset Latency, and (7) PSQI global sum. Although all seven indexes were analyzed in bivariate correlations (see Table 2), only the last four indexes were used in the final regression models for parsimony.

Parenting.

Mothers completed the widely-used Parenting Scale (Arnold, O’Leary, Wolff, & Acker, 1993) to rate discipline behaviors on 30, 7-point Likert scaled items, yielding 3 factor-based scales, laxness, overreactivity, and verbosity, and a composite scale, dysfunctional parenting, which was the average of all items (M = 2.75, SD = 0.51, α = .70). Sample items include “When my child misbehaves, I get so frustrated or angry that my child can see I’m upset” for overreactivity, “I threaten to do things that I know I won’t actually do” for laxness, and “If saying no doesn’t work right away, I keep talking and try to get through to my child” for verbosity. The overarching dysfunctional parenting scale was used in all analyses. Two percent of the sample reported problematic levels of dysfunctional parenting, based on a stringent clinical cut-off of 3.85 used in previous research (Freeman & DeCourcey, 2007).

Observed parenting measures were collected during the second home visit when two trained research assistants observed the hour before the target child’s bedtime. All research assistants were thoroughly trained on the observation procedure using a comprehensive observation manual, training sessions with lab coordinators before visiting homes, and a pairing procedure so that novice observers were accompanied by advanced observers for additional training purposes on home visits. Observers completed a Post-Observation Questionnaire (POQ) and two scales from the Home Observation for Measurement of the Environment inventory (HOME; Bradley & Caldwell, 1979). The POQ used a 5-point Likert scale to assess responsiveness via 9 items, including “to what extent did this mother seem to enjoy accommodating herself to the child’s needs and wants” and “to what extent did both mother and child contribute reciprocally to their interactions”. The two HOME subscales used were responsivity (11 items) and parental involvement (5 items). Each subscale had binary yes (1) /no (0) and not applicable responses, and the sum of responses divided by the number of applicable items was used for each. Sample items include “mother responds to child’s vocalizations with a verbal response” for responsivity and “mother tends to keep child within visual range and to look at him/her often” for involvement. The two observers’ ratings on the HOME and POQ subscales were averaged, and a single composite index of observed positive parenting was formed because all three scales were conceptually related and significantly correlated (r’s ranged from .34 to .65). Z-scores of the three scales were averaged and standardized to yield a composite z-score (range = −4.09 to 1.15). The Cronbach’s α for the full composite was 0.82. Inter-rater agreement on the 3-in-1 composite was acceptable, with r = 0.57. Mothers with z-scores at or below −.28 were in the lowest quartile of the sample.

Data Analysis Plan

All analyses were conducted in SPSS Version 24 (IBM Corp, 2016), except for the CFA, which was conducted with Mplus 7.2, using full information maximum likelihood estimation to account for missing data (Muthén & Muthén, 2014). We first computed bivariate correlations to determine the association between the CRI, maternal sleep, and parenting. We also included three possible confounding variables – maternal age, number of hours worked outside the home, and number of siblings of the target child – because each of these are conceptually distinct from the CRI variables, but potentially associated with stress, sleep, and parenting. Finally, we conducted regressions to (1) determine whether sleep indexes were associated with concurrent observed positive parenting, above and beyond the CRI and relevant covariates, and (2) to test whether sleep was associated with mother-reported dysfunctional parenting, above and beyond the CRI and the covariates.

Results

Descriptives of the stress, sleep, and observed parenting composites by site, as well as for the full sample, are listed in Table 1. Bivariate correlations between sleep, the CRI, parenting, and potential confounding variables are presented in Table 2. All of the sleep indexes except Sleep Activity were associated with the CRI, meaning that mothers who experienced shorter, more variable, and later sleep, as well as those who required longer to fall asleep and who reported more sleep problems on the PSQI, had significantly higher stress scores on the CRI. Interestingly, Sleep Activity was the only actigraphic sleep index linked with PSQI (r = .15) and not linked with the CRI (r = .05). The significant associations between the CRI and the other sleep indexes generally held even when relevant covariates (maternal age, number of children, and number of hours worked outside of the home) were statistically controlled (β = 0.27, p < 0.01 for General Sleep Deficit predicting CRI, β = 0.37, p < 0.01 for PSQI global sum predicting CRI, and β = 0.12, p < 0.10 for sleep onset latency predicting CRI).

Greater cumulative risk was associated with more mother-reported dysfunctional parenting and less observed positive parenting. Observed positive parenting was significantly associated with each of the actigraphic measures of sleep in the expected directions, such that mothers who experienced shorter, more variable, later, or more fragmented sleep, or who required longer to fall asleep, showed less positive parenting during the hour before the target child’s bedtime. However, observed positive parenting was not significantly associated with maternal report of sleep problems on the PSQI. The PSQI rating was only associated with mother-reports of dysfunctional parenting. Mothers’ reports of dysfunctional parenting were also linked with actigraphic measures of more variable sleep and longer sleep onset latency, as well as the General Sleep Deficit composite. In summary, the General Sleep Deficit composite, Sleep Variability, and sleep onset latency were each associated with both parenting measures, but Sleep Timing and Activity were only linked with lower levels of observed positive parenting.

Table 2 also shows the associations between maternal stress, sleep, and parenting with possible confounds: maternal age, hours worked, and family size. Three main patterns emerged. First, older mothers tended to go to bed earlier, work more hours outside of the home, and to have lower scores on the CRI. Additionally, mothers who worked more hours outside of the home tended to go to bed earlier. Third, mothers with more children worked fewer hours outside the home, showed less positive parenting, and reported fewer sleep problems on the PSQI.

Finally, multiple regressions tested whether maternal sleep deficits were incrementally associated with parenting, above and beyond the CRI and covariates (maternal age, number of children in the home, and number of hours worked outside of the home). Table 3 demonstrates that the three actigraphic sleep indexes (General Sleep Deficit, Sleep Activity, and sleep latency) each predicted less observed positive parenting, above and beyond the CRI and covariates, but mother-reported sleep problems on the PSQI did not. Table 4 shows the multiple regressions for mother-reported dysfunctional parenting. Only the actigraphic sleep index of sleep onset latency was significantly, positively associated with dysfunctional parenting when controlling for the CRI and relevant covariates, suggesting that mothers who required longer to fall asleep tended to report more frequent use of dysfunctional parenting.

Table 3.

Summary of Hierarchical Regression Analysis for Variables Predicting Observed Positive Parenting

	Model 1			Model 2			Model 3			Model 4
	General Sleep Deficit			Sleep Activity			Sleep Onset Latency			PSQI Global Sum
Variable	B	SE B	β	B	SE B	β	B	SE B	β	B	SE B	β
Number of Siblings	−.17	.06	−.20**	−.19	.06	−.23**	−.18	.06	−.21**	−.19	.06	−.22**
Hours Worked	−.01	.00	−.15*	−.01	.00	−.17*	−.01	.00	−.16*	−.01	.00	−.14*
Maternal Age	.01	.01	.08	.01	.01	.07	.01	.01	.06	.01	.01	.04
CRI	−.12	.07	−.13^	−.17	.07	−.18**	−.13	.06	−.14*	−.13	.07	−.13^
Sleep Index	−.17	.06	−.19**	−.19	.07	−.17*	−.01	.00	−.23**	−.03	.03	−.07
R2	.11			.10			.13			.08
F for change in R2	8.16**			6.82*			12.64**			.96

Note. Hours Worked refers to the number of hours the mother worked outside of the home each week, ranging from 0 to 60 in this sample. CRI refers to the dimensional cumulative risk index, comprised of all traditionally studied stressors. Model 1 (N = 221) used the General Sleep Deficit sleep index, which is the higher-order composite of sleep duration, timing, and variability. Model 2 (N = 220) used the Sleep Activity actigraphic component. Model 3 (N = 217) used the sleep onset latency actigraphic sleep index. Model 4 (N = 244) used the PSQI global sum sleep index. For all models, step 1 included the relevant covariates and the CRI, and the sleep index was entered in step 2.

^{^}p < .10.

^*p < .05

^**p < .01.

Table 4.

Summary of Hierarchical Regression Analysis for Variables Predicting Mother-Reported Dysfunctional Parenting

	Model 1			Model 2			Model 3			Model 4
	General Sleep Deficit			Sleep Activity			Sleep Onset Latency			PSQI Global Sum
Variable	B	SE B	β	B	SE B	β	B	SE B	β	B	SE B	β
Number of Siblings	.03	.03	.07	.04	.03	.08	.03	.03	.06	.02	.03	.05
Hours Worked	.00	.00	.12^	.00	.00	.12^	.00	.00	.10	.00	.00	.06
Maternal Age	.00	.01	−.02	.00	.01	−.02	.00	.01	.01	.00	.01	.00
CRI	.17	.04	.30**	.17	.04	.31**	.16	.04	.30**	.20	.04	.35**
Sleep Index	.03	.03	.06	.03	.04	.05	.00	.00	.21**	.00	.01	.02
R2	.11			.11			.15			.12
F for change in R2	.92			.52			10.62**			.10

Note. Hours Worked refers to the number of hours the mother worked outside of the home each week, ranging from 0 to 60 in this sample. CRI refers to the dimensional cumulative risk index, comprised of all traditionally studied stressors. Model 1 (N = 224) used the General Sleep Deficit sleep index, which is the higher-order composite of sleep duration, timing, and variability. Model 2 (N = 223) used the Sleep Activity actigraphic component. Model 3 (N = 220) used the sleep onset latency actigraphic sleep index. Model 4 (N = 250) used the PSQI global sum sleep index. For all models, step 1 included the relevant covariates and the CRI, and the sleep index was entered in step 2.

^{^}p < .10

^*p < .05

^**p < .01.

Discussion

The present study considered whether mothers’ sleep was linked with stress and parenting. To our knowledge, this is the first study to examine this among mothers of typically developing toddlers using actigraphic measures of sleep and observed and mother-report measures of parenting. The present study advances knowledge of (1) the extent to which mother sleep deficits relate to other parent stressors studied cumulatively, and (2) the links between sleep deficits and parenting, even when controlling for stressors and relevant covariates.

We found significant associations between maternal sleep difficulties (i.e., long sleep onset latencies; variable, short, and late sleep) and our CRI, which included stressors like single parenthood, low SES, child misbehavior, and household chaos. This is consistent with prior work showing links between sleep and stress in general (Akerstedt et al., 2002; El-Sheikh, Keiley, Bagley, & Chen, 2015), and links between maternal insomnia and household features (e.g., chaos, stimulation, and neglect) in particular (Gregory et al., 2012). Using actigraphic sleep indexes and comprehensive measures of stress in a large community sample of mothers of toddlers, we found that maternal sleep deficits were indeed associated with other forms of stress, and that these links were not driven by the mother’s age, the number of children she had, or the number of hours she worked outside of the home.

As we considered potentially confounding maternal characteristics, some noteworthy patterns emerged. We found that older mothers tended to go to sleep earlier, work more hours, and experience less stress compared to younger mothers. This confirms the abundance of previous research showing the sequelae of risks associated with young maternal age (e.g., Deal & Holt, 1998). We also found that mothers who worked more outside of the home tended to go to sleep earlier, perhaps due to their fixed work schedules and need to rise early for work each day. Mothers with more children worked fewer hours outside of the home and tended to show less positive parenting during the hour before the target child’s bedtime. This result is unsurprising given that mothers whose attention may be divided among several children may be less able to be highly involved and responsive to a single child during the bedtime routine. However, it was surprising to find that mothers with more children reported fewer sleep problems on the PSQI. It is unclear why this result emerged.

Each actigraphic sleep index (General Sleep Deficit, Sleep Activity, and sleep onset latency) was associated with less observed positive parenting, even after statistically accounting for individual differences in the CRI and mothers’ age, number of children, and employment. These associations have important implications because positive parenting is linked with better sleep for children (Teti et al., 2010) and positive child adjustment (e.g., Kochanska & Aksan, 1995). The present findings also add to intervention findings, which show that increases in the use of positive parenting techniques are linked with reductions in child behavior problems (Kaminski & Claussen, 2017), suggesting that efforts to promote positive parenting may have important downstream effects for child development. Although we think it is possible that mothers’ sleep is part of a chain of influences on parenting and child social development, we recognize that the empirical evidence for that chain, which is a relatively few correlational and intervention studies, is currently thin.

Mother-reported sleep problems were not significantly predictive of observed positive parenting when other forms of stress were controlled. Prior research has shown that actigraphy shows greater correspondence with polysomnography (the most precise method) than self-report measures, but actigraphy has low specificity for detecting wakefulness (Sadeh, 2011). Given the strengths and limitations of actigraphy, it is crucial that complementary measures of sleep (e.g., diary and PSQI) are also examined. The present findings suggest that mother reports of sleep deficits seem to operate in tandem with other reports of stress in their relations with parenting. The evidence for greater correspondence between actigraphy and polysomnography may at least partially explain why mothers’ actigraphically measured experiences of poor and insufficient sleep showed stronger associations with positive parenting than did mothers’ perceptions of sleep concerns, but future research is needed to replicate and explain these findings.

We also examined predictors of maternal reports of dysfunctional parenting. In bivariate correlations, higher levels of dysfunctional parenting were associated with more mother-reported sleep problems, higher scores on the General Sleep Deficit actigraphic composite, and longer time required to fall asleep. However, once the CRI and covariates were statistically controlled, only sleep onset latency was significantly predictive of dysfunctional parenting, such that mothers who required longer to fall asleep tended to report more regular use of overreactive, verbose, or lax parenting. Most of the variance in dysfunctional parenting seemed to be explained by the CRI, which seems plausible given previous research showing robust links between stress and dysfunctional parenting (Reitman, 2001), and given the shared method variance between the stress and parenting measures.

Insufficient and poor quality maternal sleep explained substantial variance in positive parenting, beyond other forms of stress, in this sample of relatively psychologically healthy families. We speculate that parental sleep may ultimately prove to be a useful intervention target. Behavioral interventions have been shown to successfully improve children’s sleep, with concomitant improvements in parental sleep (Mindell et al., 2011) and stress (Wolfson, Lacks, & Futterman, 1992). Interventions highlighting parent sleep and establishing consistent bedtime routines for the broader family system might improve parent sleep and result in more positive exchanges between parents and their children, but this has not yet been empirically tested.

This study had shortcomings that limit our conclusions but point to future research directions. First, the cross-sectional design meant we could not estimate potential causal directions in associations among stress, sleep, and parenting over time. Second, our findings demonstrate the importance of using actigraphic measures of sleep and observed measures of positive parenting, but it would be useful for future research to incorporate observed measures of both positive and dysfunctional forms of parenting and mother-reports of both positive and dysfunctional forms of parenting to comprehensively assess multiple domains of parenting. To get a valid, observational measure of dysfunctional parenting, it may be necessary to observe for much longer periods than we did. It was very rare to see truly negative parenting behavior, such as harsh scolding or spanking, during the home visits. Third, although the community sample did have a meaningful range of stress, it was not a sample of highly sleep deprived and otherwise highly stressed mothers, such as might be seen in a clinic for child behavior problems. Fourth, as in all studies with community samples, the assumption that families who participated represent those who did not has to be provisional, awaiting more advanced and expensive sampling.

Sleep Activity was the only actigraphic index linked with mother-reported sleep problems on the PSQI, and not linked with the CRI. The former point suggests that the PSQI global sum may primarily reflect poor sleep quality, rather than insufficient sleep or long sleep onset latencies. Future research is needed to advance understanding of the nature of parent sleep activity to explain these findings. Although actigraphy shows high minute-by-minute agreement with polysomnography, it also has been shown to overestimate night awakenings, potentially due to sensitivity to small movements (Sadeh, 2011). Future research could delineate various aspects of sleep activity, such as general activity during the sleep period versus awakenings lasting five or more minutes, to clarify the findings.

The current study advances understanding of stress and parenting by incorporating the associated influence of sleep deficits for mothers of toddlers. Maternal sleep, particularly when measured with actigraphs, was found to be significantly associated with other forms of stress as well as qualities of parenting. The findings support interest in parent sleep as a possible intervention target to promote positive parenting practices, which are associated with socially valued outcomes for children.