Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2024 May 1.
Published in final edited form as: J Appl Dev Psychol. 2023 Mar 15;86:101530. doi: 10.1016/j.appdev.2023.101530

Socioeconomic Disparities, Nighttime Bedroom Temperature, and Children’s Sleep

Ben Hinnant a, Joseph A Buckhalt a, Emily F Brigham a, Brian T Gillis a, Mona El-Sheikh a
PMCID: PMC10081498  NIHMSID: NIHMS1884856  PMID: 37035841

Abstract

We investigated associations between family income-to-needs, nighttime bedroom temperature (NBT), and children’s sleep. Using a sample of 46 children (Mage = 11.5), we recorded NBT and objective sleep parameters via actigraphy nightly for one week to evaluate within- (night-to-night) and between-person associations. We found consistent evidence for a curvilinear association between NBT and sleep variables at the between-person level, indicating that children who slept in rooms that were “too hot” or “too cold” experienced poorer sleep. Moreover, children in lower income-to-needs families had more extreme NBTs. There was some evidence that family income-to-needs is indirectly related to sleep via NBT, but with interpretational caveats. These findings point to NBT as a potentially modifiable variable, which has implications for practical applications to mitigate effects of socioeconomic disparities on children’s sleep.

Sleep is a pivotal bioregulatory system that is a correlate and predictor of developmental outcomes in children, including cognitive functioning, academic achievement, and mental and physical health (Shochat et al., 2014; Spruyt, 2019). Thus, identification of variables that can affect sleep in youth is warranted (El-Sheikh & Sadeh, 2015). Sleep is best understood and operationalized as a multi-dimensional construct and is commonly measured via actigraphy (i.e., objective sleep data collected on a watch-like device). Important and frequently used actigraphy measures assessing different facets of sleep include sleep efficiency (i.e., a measure of sleep quality), sleep time, wake time, long wake episodes (i.e., sleep periods with more than five minutes of activity), and sleep activity (i.e., percentage of sleep periods with activity; El-Sheikh & Sadeh, 2015).

Although it is acknowledged that the physical sleep environment is associated with sleep (McWood et al., 2022), few studies have focused on the association between nighttime bedroom temperature (NBT) and objective measures of children’s sleep. Studies of experimental manipulations of temperature on sleep in adults have shown that sleep quality is degraded on either side of an ideal temperature range (Caddick et al., 2018; Joshi et al., 2016). Children have a much higher surface area to mass ratio than adults, which means greater heat absorption and heat loss in sleeping conditions that are too hot or too cold, respectively (Falk, 1998). As a result, children’s sleep may be especially sensitive to variation in NBT. If the curvilinear association between NBT and sleep is replicated in children, the question of what variables predict NBT and “too hot” and “too cold” sleeping conditions becomes important. In this investigation, we tested curvilinear associations between repeated measures of NBT and children’s objective sleep measures over one week at within- (i.e., night-to-night) and between-person levels of analysis. Moreover, we explored the link between socioeconomic disparities and NBT by testing the novel hypothesis that children in lower income-to-needs families would sleep at more extreme NBTs (i.e., too hot and too cold sleeping conditions relative to the between-person sample mean). Finally, we explored extreme NBTs as a possible mechanism linking socioeconomic disparities (operationalized as family income-to-needs) and children’s sleep in indirect effects models. The overall goals of this study are to highlight associations between NBT and children’s sleep, to point out the insidious and multifarious ways in which socioeconomic disparities may impact children’s sleep, and to call attention to improved sleeping conditions as achievable targets of holistic interventions aimed at mitigating the effects of socioeconomic disparities.

Temperature and Sleep

NBT is one part of the sleep environment, which also includes air flow, humidity, and personal thermal comfort (e.g., sleepwear and bedding; Caddick et al., 2018). The sleep environment and individual difference factors (e.g., biological thermal regulation) both contribute to a thermal comfort zone, the microclimate range in which a person feels comfortable while sleeping (Joshi et al., 2016). Experimental studies with adults under varying NBT and thermal comfort conditions have found temperatures outside of the thermal comfort zone negatively impact sleep (Bach et al., 1994; Togo et al., 2007). Similarly, within-persons, large changes in ambient temperature have been shown to negatively impact sleep (Lan et al., 2016; Togo et al., 2007). A few studies have established relations between temperature and sleep in children, though most studies have relied on parent- or child-reports through general sleep environment measures rather than objective measurement of NBT (Bagley et al., 2015; Rubens et al., 2018). Thermoregulatory capabilities of adults and children differ (Falk, 1998) and NBT is likely to be as important, if not more important, for children’s sleep. However, no studies have attempted to extend experimental results with adults indicating a curvilinear association between NBT and sleep measures to pediatric samples. Thus, the first aim of this study was to test for curvilinear associations between NBT and objective measures of children’s sleep at the within- and between-person levels of analysis. We hypothesized that at the within-person level, sleep would be poorer when NBT deviated in either direction relative to a child’s own mean NBT (i.e., a curvilinear, quadratic association between NBT and sleep). We made the same hypothesis at the between-person level, namely that children sleeping in NBTs outside of the sample’s mean NBT in either direction would have poorer sleep.

Socioeconomic Disparities and Sleep Environments

Disparities pervade the day-to-day lives of socioeconomically disadvantaged individuals and families, including their sleep environments. A growing literature elucidates in-home factors impacting sleep that are more likely to occur in lower income households, including indoor air temperature (Hoyniak et al., 2022). The homes of lower income individuals tend to hotter in the summer (Ruddell et al., 2010) and colder in the winter (San Miguel-Bellod et al., 2018). Limited ability to regulate indoor temperature to a comfortable level may be an important mechanism by which socioeconomic disparities affect sleep (Mezick et al., 2008) and health more broadly (Gronlund, 2014). We expected that children from socioeconomically disadvantaged families would sleep at temperatures that are too hot and too cold, an association that is neither linear nor curvilinear. Thus, we hypothesized that the between-person regression of NBT on income-to-needs would violate the linear regression assumption of homoscedasticity (i.e., that residuals are constant along the predictor-outcome slope) and that greater residuals, which reflect greater than expected deviation in NBT (i.e., too hot and too cold NBTs), would be more apparent for children in lower income-to-needs families. For simplicity and to differentiate them from observed NBTs related to the first study aim, we refer to too hot and too cold NBTs based on the residuals of the NBT on income-to-needs slope as extreme NBTs.

Socioeconomic Disparities and Sleep

Sleep health disparities have been well documented in adults and children. In a recent meta-analytic review of 19 studies, Sosso et al. (2021) found consistent positive associations between multiple measures of socioeconomic status and objective, actigraphic measures of sleep duration and efficiency. These associations were not found to differ by age of study participants, suggesting that children experience the effects of socioeconomic disparities on sleep as strongly as adults. It is important to note that socioeconomic disparities may exert their effects on sleep via various mechanisms of risk including the physical sleep environment (Bagley et al., 2015; Hoyniak et al., 2022; Phelan et al., 2010; Rubens et al., 2019). Alhasan et al. (2022) and El-Sheikh et al. (2022) have highlighted how little is known about the link between socioeconomic disparities and sleep while providing recommendations for future research on the multilevel mechanisms connecting these constructs. Limited work has tested potential mechanisms linking socioeconomic resources to sleep (e.g., Fronberg et al., 2022), warranting further research (Sosso et al., 2022). Therefore, the third aim of this study was to examine NBT as a potential microsystem-level mechanism indirectly linking socioeconomic disparities, operationalized as family income-to-needs, to objective measures of children’s sleep. We hypothesized that children in lower income-to-needs families would sleep in environments characterized by more extreme NBTs (too hot or too cold) and that extreme NBTs would be related to poorer sleep.

The Current Study

The first aim of this study extends the literature on relations between NBT and sleep, testing the hypothesis that NBT would show a curvilinear association with objective sleep measures wherein children would sleep more poorly when NBT deviated in either direction relative to a child’s own mean NBT (within-person) or the sample NBT (between-person). The second aim of this study tests the hypothesis that children in lower income-to-needs families sleep at more extreme NBTs, while the third aim of this study was to test the hypothesis that extreme NBTs may be an indirect mechanism linking socioeconomic disparities to children’s sleep. Additionally, this study aims to highlight the potential of NBT, as well as sleep hygiene in general, as a potentially achievable target in practical interventions aimed at reducing the effects of socioeconomic disparities on children’s development. To address the study aims and hypotheses, we use a subsample of a larger ongoing, longitudinal study of child development that piloted the use of objective sleep environment data collection instruments in families’ homes. In addition to the use of actigraphy devices to collect sleep data over one week, thermometers were used to record NBTs in children’s sleeping areas. To maximize internal validity, we considered potential confounding variables of sex, race, age, body mass index (BMI), and season (i.e., fall, spring) as between-person covariates.

Method

Participants

Participants were 46 children aged 10 to 12.67 years (19 female, Mage = 11.52, SD = 0.65) and were part of the third wave of a larger longitudinal study examining biopsychosocial influences on health (Auburn University Sleep Study) taking place in the Southeastern United States. We selected participants for the present study if they completed measurements for NBT, the primary variable of interest (N = 52), and if participants had 5 or more nights of sleep data (N = 46). This resulted in a subsample that was approximately 14% of the sample at the third wave of the larger study. Independent samples t-tests indicated that there were no significant differences based on race, sex, income to needs ratio, or parental education between the subsample and the larger sample (see Table 1). The university IRB approved the study and protocols and parents provided written consent while children provided assent. Children were fifth- and six-graders who were racially representative of their community (69% White/European American, 30% Black/African American). Educational attainment data for mothers and fathers indicated that 56% and 73% were high school graduates and that 44% and 24% had a college degree, respectively. The income-to-needs ratio of the sample ranged from 0 to 7.86 (M = 1.54, SD = 1.09), with 24% of the sample living below the poverty line (income-to-needs < 1).

Table 1.

Independent Samples T-Tests Comparing Included (N = 46) and Excluded Sample (N = 292)

Variable t p Mean Diff df
Child Race 1.12 .27 .083 275
Child Sex −1.61 .11 −.124 275
Income-to-Needs 1.25 .21 .561 275
Mother Degree −.69 .48 −.139 183
Father Degree .15 .88 .030 177
BMI .45 .67 .085 249
Open in a new tab

Note. BMI = body mass index. T-tests were conducted in SPSS using pairwise deletion, resulting in the varying sample size.

Procedure

Thermometers and actigraphs were mailed to families. Parents were asked to place the thermometer in children’s room for seven consecutive nights during the academic school year and to secure the actigraph to the child’s non-dominant wrist before sleep. Children completed a sleep diary each night, which included reports of sleep and wake times to validate actigraphy data. Children had between 5 and 7 nights of actigraphy data (M = 6.10, SD = 0.82). Following the actigraphic assessment, families visited the lab for further data collection.

Measures

Nighttime Bedroom Temperature (NBT)

NBT data were collected at 10-minute intervals using the Extech Instruments Model RHT10 Temperature/Humidity Datalogger, which has an accuracy of less than ±1 °C for temperature and ±3.5% for relative humidity, and records at a resolution of .1°C and .1%, respectively (Extech Industries, Nashua, NH). Use of this device is supported by other studies measuring temperature using the same device in indoor environments, including school classrooms and office spaces (Katafygiotou & Serghides, 2014; Marchetti et al., 2015). In the present study, the device recorded temperature in the child’s bedroom during the 7 nights actigraphy data were collected. Parents were instructed to place the device next to their child’s sleeping area. Devices ran continuously and children’s reports of sleep and wake times were used to calculate measures of NBT, reported in degrees Celsius. Average NBT (i.e., the mean of all 10-minute measurements that occurred from sleep onset to wake) had good night-to-night consistency (α = .86). NBT was treated as random in analysis (i.e., varying both within- and between-persons).

Extreme NBTs.

NBT residuals from the regression of NBT on income-to-needs were estimated and saved as a new variable. Positive NBT residuals reflect “too hot” sleeping conditions (i.e., NBTs higher than would be expected based on family income-to-needs). Negative NBT residuals reflect “too cold” sleeping conditions (i.e., NBTs lower than would be expected based on family income-to-needs). Negative NBT residuals were multiplied by −1 so that extreme NBT scores captured both “too hot” and “too cold” conditions. This operationalization is, to our knowledge, a novel approach to evaluating the variation in NBT that may exist due to socioeconomic disparities. Because extreme NBT scores are the residuals from the between-person regression of NBT on income-to-needs, extreme NBT was treated as a between-person variable only.

Sleep Actigraphy

Octagonal Basic Motionloggers (Ambulatory Monitoring, Inc, Ardsley, NY, USA) measured motion in 1-min epochs using the zero-crossing mode. Analyses were conducted with the Octagonal Motionlogger Interface with ACTme software and the analysis software package (Action W2, 2000 Ambulatory Monitoring Inc.). The Sadeh scoring algorithm was used to determine whether children were awake or asleep (Sadeh et al., 1994). Nights on which children used medication for acute illnesses (e.g., cold) were excluded from analyses. On average, children had 6.1 nights of actigraphy data.

Assessing multiple sleep parameters is recommended (Sadeh et al., 2000), and thus the following well-established parameters were derived: (a) Sleep Efficiency (as a measure of sleep quality): percentage of epochs scored as sleep between onset and wake time; (b) Sleep Time: total minutes between sleep onset and wake time excluding periods of wakefulness; (c) Wake Time – total minutes during the night scored as being awake; (d) Long Wake Episodes: number of waking episodes ≥5 min, and (e) Sleep Activity: percentage of epochs with activity > zero (multiplied by 10 to scale the variance). Definitions of all sleep variables are consistent with the actigraphy manual. Notably, research has shown that the actigraphy instrument used to measure sleep in this study provides reliable and valid data at a normal range of sleep temperatures (Shin et al., 2015). There was good night-to-night reliability for these variables (α = .72-.92).

Because sleep efficiency is a percentile, its variance was very small. Thus, sleep efficiency scores were multiplied by 10 for data analysis purposes. Sleep time and wake time had very large variances and thus were divided by 60 to derive hours for analytic purposes.

Potential Covariates

Sex (coded as 0 for female and 1 for male), race (coded as 0 for White/European American and 1 for Black/African American), age, BMI, and season of data collection (coded as 0 for spring, 1 for fall) were considered as potential between-person covariates.

Results

Analysis Plan

Multilevel models were used to evaluate the non-independence of repeated measures of NBT and sleep variables and evaluate their associations at within- and between-person levels (Heck & Thomas, 2020). First, the total variances of NBT and sleep variables were partitioned into within- (night-to-night) and between-person levels in unconditional models and intraclass correlation coefficients (ICCs; the proportion of between-person variance in NBT and sleep variables) were estimated. We addressed our first hypothesis about curvilinear associations between NBT and sleep variables at within- and between-person levels in a set of conditional multilevel models. We addressed our second hypothesis that children in lower income-to-needs families would sleep at more extreme temperatures by regressing between-person NBT on income-to-needs and examining the residuals. Finally, we tested our third hypothesis that NBT may be an indirect mechanism linking socioeconomic disparities to children’s sleep in another set of indirect effect multilevel models, though the indirect association was only tested at the between-person level because income-to-needs was treated as time-invariant. We present both p values and the 90% confidence intervals of estimates in evaluating the hypotheses so that readers can make informed decisions about the meaningfulness of the results (du Prel et al., 2009).

Data analysis was conducted in Mplus version 8.1 (Muthén & Muthén, 1998–2017). In all models, predictor variables were grand mean centered. The curvilinear (i.e., quadratic) term for NBT was created by squaring that variable. Analytic models used maximum likelihood estimation (Enders & Bandalos, 2001) with robust standard errors (Maas & Hox, 2004), the default estimator in Mplus 8.1. Indirect associations were estimated with the default delta method (Bollen & Stine, 1990); bootstrapped indirect effects are not currently available in multilevel models.

Preliminary Analyses

Descriptive analyses of study variables were evaluated and appear in Table 2. Additionally, missingness was examined for the primary study variables. Among the 46 study participants, four were missing data on income-to-needs. There were no missing data for child age, NBT, or sleep variables. Independent samples t-tests indicated that participants missing income-to-needs data did not differ from participants with income-to-needs data on any variables, fitting the assumption that data were missing completely at random (i.e., MCAR). Sex, race, age, BMI, and season of data collection were considered as potential between-person covariates; however only age was consistently correlated with sleep variables and was retained as a covariate in the multilevel models.

Table 2.

Descriptive Statistics and Correlations Among Study Variables

Variable 1 2 3 4 5 6 7 8 9
1. Child age (years) -
2. Income-to-Needs −.08 -
3. NBT −.12 −.32* -
4. Extreme NBT .06 −.16* .05 -
5. Wake time .26 −.17 −.20 .06 -
6. Sleep time −.41** .10 .10 −.11 −.23 -
7. Sleep efficiency −.33** .20 .20 −.02 .97** .40* -
8. Long wake episodes .31* −.19 −.10 .01 .82** −.39* −.86** -
9. Sleep activity .20 .01 −.21 .09 .73** −.20 -.75** .63** -
Mean
(SD)		 11.52
(.65)		 1.54
(1.09)		 21.74
(2.57)		 1.23
(.09)		 0.74
(.54)		 7.15
(1.05)		 9.10
(.67)		 2.65
(2.43)		 36.29
(13.41)
Open in a new tab

Note. NBT = nighttime bedroom temperature.

*

p < .05.

**

p < .01.

Unconditional multilevel models showed that NBT and all sleep variables varied both within- and between-persons with intraclass correlation coefficients (ICCs) ranging from .34 to .71 and so were considered as random intercepts (i.e., allowed to vary between-persons). The ICC for NBT was .59, indicating that 59% of the variance was between-persons and 41% was within-persons. ICCs for sleep variables are included in Table 3. Variance estimates for NBT and sleep variables had p values less than .001 at the within-person level and less than .05 at the between-person level. Together, the ICCs and p values for the variance estimates of NBT and sleep variables suggest that there is meaningful variance for analysis at both within- and between-person levels.

Table 3.

Two-level Analysis of Sleep Parameters Within- and Between-Individuals on Nighttime Bedroom Temperature

Sleep Efficiency
ICC = .53		 Sleep Time
ICC = .34		 Wake Time
ICC = .52		 Long Wake Episodes
ICC = .39		 Sleep Activity
ICC = .71
Level 2: B SE B SE B SE B SE B SE
Between 90% CI 90% CI 90% CI 90% CI 90% CI
Intercept 9.20**
[9.02, 9.37]		 .10 7.51**
[7.30, 7.71]		 .12 .66**
[.53, .80]		 .08 2.32**
[1.84, 2.80]		 .29 33.08**
[29.54, 36.62]		 2.15
Linear NBT .08
[−.01, .18]		 .06 .03
[−.11, .17]		 .09 −.07
[−.14, .01]		 .04 −.25
[−.51, .01]		 .16 −1.84†
[−3.58, −.11]		 1.06
Quadratic NBT −.03**
[−.04, −.01]		 .01 −.03*
[−.05, −.01]		 .01 .02**
[.01, .03]		 .01 .07**
[.03, .12]		 .03 .46**
[.18, .74]		 .17
Age −.18
[−.39, .03]		 .13 −.17
[−.45, .11]		 .17 .15
[−.02, .31]		 .10 .47
[−.07, 1.00]		 .33 1.65
[−2.87, 6.16]		 2.74
R 2 .22 .20 .21 .23 .14
Level 1: B SE B SE B SE B SE B SE
Within 90% CI 90% CI 90% CI 90% CI 90% CI
Linear NBT −.01
[−.05, .02]		 .02 −.01
[−.09, .06]		 .05 .01
[−.02, .04]		 .02 .10
[−.02, .21]		 .07 .71*
[.19, 1.24]		 .32
Quadratic NBT .01
[−.01, .02]		 .01 .01
[−.01, .02]		 .01 −.01
[−.01, .01]		 .01 −.01
[−.03, .01]		 .01 .08
[−.01, .16]		 .05
R 2 <.01 <.01 <.01 <.01 .03
Open in a new tab

Note. Sleep time and wake time are in hours.

†

p < .10.

*

p < .05.

**

p < .01.

Primary Analyses

Nighttime Bedroom Temperature and Sleep

Within-Person Associations in Nighttime Bedroom Temperature and Sleep.

Sleep variables were regressed on the linear and quadratic NBT variables at the within- and between-person levels and age at the between-person level (Table 2; note that after entering NBT variables into the model, age was no longer associated with any sleep variables). The within-person effects evaluate night-to-night associations between NBT and sleep variables (i.e., relative to an individual’s own mean sleep across nights, whether intraindividual variation in NBT is related to sleep) while the between-person effects evaluate individual difference associations between NBT and sleep variables (i.e., relative to the sample’s mean, whether interindividual variation in NBT is related to sleep). Findings indicated that the only relationship between NBT and sleep variables at the within-person level was a linear association with sleep activity (i.e., on nights when the room was warmer, children experienced more sleep activity).

Between-Person Associations in Nighttime Bedroom Temperature and Sleep.

There were no linear associations between NBT and sleep parameters at the between-person level. However, the quadratic term for NBT was a predictor of all sleep parameters. This relationship is shown for sleep efficiency in Figure 1 and suggests that children’s sleep suffers in “too hot” or “too cold” nighttime conditions. Notably, peak sleep efficiency across children is observed at about 22–23C (71–73F). Supplemental Figures 1–4 depict the additional quadratic effects on sleep parameters. Thus, our first hypothesis that NBT would have curvilinear associations with sleep variables was supported at the between-person level but not at the within-person level of analysis.

Figure 1. Nonlinear association between nighttime bedroom temperature and sleep efficiency at the between-person level.

Figure 1.

Open in a new tab

Income-to-Needs and Nighttime Bedroom Temperature

NBT was regressed on family income-to-needs at the between-person level to examine whether the assumption of homoscedasticity was violated and whether the residuals of the regression slope indicated that children in lower income-to-needs homes slept at more extreme temperatures (hypothesis two). Income-to-needs was negatively associated with NBT, indicating that children from lower income-to-needs families tended to sleep at warmer temperatures, B = −.73, SE = .14, p < .01, 90% CI [−.97, −.50]. Inspection of the residuals, shown in Figure 2, indicated that there was greater deviation in observed NBT values at lower levels of family income-to-needs. The bivariate correlation between the two indicated that income-to-needs was negatively associated with more extreme NBTs, r = −.16, SE = .07, p = .01, 90% CI [−.27, −.05]. These findings support our second hypothesis that children from lower income-to-needs families sleep at temperatures that are higher and lower than would be expected, relative to the sample mean.

Figure 2. Residual NBT scores across levels of family income-to-needs.

Figure 2.

Open in a new tab

Indirect Associations via Extreme Nighttime Bedroom Temperature

Because family income-to-needs did not vary within-persons, indirect effect models were tested at the between-person level. All results are presented in Table 4. Family income-to-needs was not directly related to sleep measures. Family income-to-needs was negatively related to extreme NBTs; children from lower income-to-needs families tended to sleep at more extreme NBTs, though the regression coefficients varied slightly across models with different sleep variables as outcomes. Extreme NBTs were related to all sleep measures in expected directions (i.e., negatively related to sleep efficiency and sleep time and positively related to wake time, long wake episodes, and sleep activity); children sleeping at more extreme NBTs tended to have poorer sleep. There was evidence to varying degrees for indirect associations between family income-to-needs and sleep efficiency, wake time, and long wake episodes via extreme NBTs but not for sleep time or sleep activity. The strongest indirect association between income-to-needs and sleep variables was found for long wake episodes but there was also some evidence for indirect associations with sleep efficiency and wake time.

Table 4.

Two-level Analysis of Indirect Associations between Family Income-to-Needs, Extreme Nighttime Bedroom Temperature, and Sleep Parameters

Extreme NBT
ICC = .41		 Sleep Efficiency
ICC = .53		 Sleep Time
ICC = .34		 Wake Time
ICC = .52		 Long Wake Episodes
ICC = .39		 Sleep Activity
ICC = .71
Level 2: B SE B SE B SE B SE B SE B SE
Between 90% CI 90% CI 90% CI 90% CI 90% CI 90% CI
Intercept .04
[−.24, .32]		 .17 9.02**
[8.87, 9.17]		 .09 7.96**
[7.55, 8.37]		 .25 .39**
[.15, .64]		 .15 1.15
[−.02, 2.32]		 .71 27.33**
[21.43, 33.23]		 3.59
Inc.-to-needs −.29†
[−.57, −.01]		 .17 −.06
[−.17, .06]		 .07 .01
[−.17, .19]		 .11 .05
[−.03, .14]		 .05 .31†
[.02, .60]		 .18 2.38
[−.16, 4.92]		 1.54
Extreme NBT - - −.30**
[−.43, −.17]		 .08 −.35**
[−.54, −.15]		 .12 .22**
[.11, 32]		 .07 .87*
[.24, 1.50]		 .38 4.56**
[2.23, 6.88]		 1.41
Age - - −0.20
[−.42, .02]		 .13 −.15
[−.41, .12]		 .16 .17
[−.01, .34]		 .10 .55
[−.01, 1.09]		 .33 2.46
[−2.29, 2.46]		 2.89
Indirect Effect .09†
[.01, .17]		 .05 .10
[−.02, 21]		 .07 −.06†
[−.12, −.01]		 .04 −.25*
[−.47, −.03]		 .13 −1.32
[−2.79, .16]		 .90
R 2 .26 .31 .24 .33 .15
Level 1:
Within		 B
90% CI 		SE B
90% CI 		SE B
90% CI 		SE B
90% CI 		SE B
90% CI 		SE
Extreme NBT −.01
[−.08, .05]		 .04 −.02
[−.13, .09]		 .07 .01
[−.04, .06]		 .03 −.08
[−.27, .11]		 .12 .56
[−.14, 1.25]		 .42
R 2 <.01 <.01 <.01 <.01 <.01
Open in a new tab

Note. Sleep time and wake time are in hours.

†

p < .10.

*

p < .05.

**

p < .01.

Discussion

There is strong meta-analytic evidence for an association between socioeconomic resources and sleep disparities; those with fewer economic resources tend to have poorer sleep (Sosso et al., 2021). However, more research is needed on the mechanisms that explain the link between socioeconomic disparities and sleep in order for intervention efforts to be able to effectively target these mechanisms. This study partly addressed this gap in knowledge by pointing out NBT as a physical sleep environment characteristic that could be targeted as part of interventions aimed at reducing the effects of socioeconomic disparities on children’s development. This study used a sub-sample (n = 46) of a larger longitudinal study of child development that piloted more intensive measurement of sleep environments, including repeated measures of NBT over one week. The first aim of this study was to examine curvilinear associations between NBT and multiple sleep parameters in children to evaluate whether children slept more poorly when NBT deviated on either side of their own within-person (i.e., night-to-night) or between-person (i.e., sample) means. We found robust between-person curvilinear associations between NBT and all objective measures of children’s sleep; children slept more poorly at temperatures on either side (i.e., “too hot” and “too cold”) of the sample’s average NBT of 22–23C (71–73F). The curvilinear associations with poorer sleep seemed to be more pronounced at temperatures that were colder than, relative to warmer than, the sample’s mean NBT. These findings mirror, in part, existing research on effects of temperature on sleep in adults (Caddick et al., 2018; Joshi et al., 2016). However, we also expected to find the same curvilinear association at the within-person level of analysis in that children would sleep more poorly on nights when NBT deviated in either direction from a given child’s own mean NBT. Although there was meaningful variance in NBT and all sleep measures at both within- and between-person levels of analysis, our hypothesis was only supported at the between-person level. These findings would seem to contradict results of within-person manipulations of sleep conducted with adult samples (Lan et al., 2016; Togo et al., 2007), however these experiments focused on the effects of within-night variations in temperature and limited thermal comfort options as part of their designs. In the context of night-to-night variation in NBT and its association with sleep in this correlational study, children may adjust their thermal comfort on an as needed basis to the degree that they are able (e.g., kicking off blankets on nights that they are too hot or adding extra clothing on cold nights). The capability of children to adjust their own thermal comfort to NBT conditions or for parents to provide ideal NBT conditions is likely linked to socioeconomic resources, the key construct in addressing study aims two and three.

The second aim of the study was to investigate whether children from families with lower income-to-needs ratios sleep at more extreme NBTs. To operationalize extreme NBTs, we regressed NBT on family income-to-needs at the between-person level of analysis and used the residuals to capture “too hot” and “too cold” NBTs (i.e., deviation scores in NBT as a function of income-to-needs). The linear regression of NBT on family income-to-needs indicated that children in families with lower income-to-needs ratios tended to sleep at warmer temperatures. This result could speculatively be attributed to the geographic region where the study was conducted, the Southeastern United States where temperatures tend to be higher than in some other regions of the country. More importantly, extreme NBT (i.e., residuals capturing “too hot” or “too cold” NBTs) was negatively associated with family income-to-needs. Supporting the second hypothesis, children from families with lower income-to-needs ratios tended to have more extreme deviations from the sample mean in their NBT, indicating that they experience more extreme NBTs in both directions (i.e., “too hot” and “too cold” relative to the average NBT). The third study aim was to test whether extreme NBTs may be an indirect mechanism by which family income-to-needs is associated with children’s sleep. The strength of the indirect associations between family income-to-needs and sleep variables via extreme NBTs varied by sleep outcome with three of five sleep variables (sleep efficiency, wake time, and long wake episodes) showing some evidence for an indirect association. Thus, there was mixed support for our third hypothesis. In considering these mixed findings for indirect paths in more detail, it is clear from the results that the links between extreme NBTs and all sleep variables were robust (i.e., children sleeping at more extreme NBTs had poorer sleep as measured by all the objective sleep variables). The more modest association in the indirect paths was in the link between family income-to-needs and extreme NBTs. Given that the indirect effect estimate is calculated as some variation of the product of those two regression coefficients (depending on the specific indirect effect estimation approach used), the mixed findings could be attributed to the less robust link in the chain, possibly in combination with the modest sample size and larger standard errors that are tied to smaller sample sizes. If our results are valid, research with larger samples might yield similar estimates but with higher power to detect true indirect associations. Alternatively, the relatively weaker link between family income and extreme NBTs may indicate that there are other important predictors of NBT to consider.

Nonetheless, our study extends the literature in important ways through both methods and findings. With regard to the former, our first aim used a rigorous repeated-measures multilevel design, allowing us to test within- and between-person associations, which is under-utilized in developmental science (Rush & Hofer, 2017). Our second aim then modeled extreme NBTs in a novel way by examining the distribution of residual values of extreme NBTs (regardless of hot/cold) to determine covariance with family income. Our third aim evaluated indirect associations to identify a specific mechanism linking socioeconomic advantage to sleep disturbance in children, thus highlighting an important aspect of the “socioeconomic model of sleep,” which provides a roadmap for understanding the effects of socioeconomic position on sleep (Sosso et al., 2021). Our in-home, objective assessment of temperature (across multiple nights) is not often seen in this literature, but is a needed addition, given that many studies to date have only relied on parent-report of room temperature in general using single-items about sleeping rooms being either too hot or too cold (Johnson et al., 2018; Lee et al., 2018). Finally, our findings link NBT to sleep in middle childhood, contributing to a literature that primarily has examined such associations in preschool (Hoyniak et al., 2022) and adult (Caddick et al., 2018) samples.

Limitations and Future Directions

The present study, while contributing important findings to the study of sleep-related disparities in socioeconomically disadvantaged youth, is limited in its translation from analysis to implications. By this, we refer to the presumed process-based analysis for testing our third hypothesis that sleeping at extreme NBTs would be the indirect link between family income-to-needs and sleep. Testing cause-effect relationship(s) with cross-sectional data is impossible, thus we have studiously avoided using the term mediation along with its causal implications. However, indirect associations from a statistical perspective do not necessarily have to imply that a causal mediation process is being tested. Replication and extension of these findings with longitudinal data will thus be key to either supporting or contradicting our results.

A consideration of the study, if not a limitation, is its measurement of extreme NBT. The analytic approach used is, to our knowledge, a novel way of operationalizing extreme NBTs in order to capture unexpectedly high or low NBTs in a single variable. Thus, the validity of this method of measuring extreme NBTs is limited to its predictive validity (i.e., it was related to family income-to-needs in the expected direction) and its statistical validity (i.e., the use of residuals to capture unexpectedly hot or cold NBTs is a logical solution that could not be addressed with linear or polynomial regression). There are, however, alternate analytic approaches that could be used to address the association between socioeconomic disparities, NBT, and sleep in a similar way. We address some of these below.

These findings based on data from a subsample of a larger study. Participants in this study, however, did not differ from the full sample on key characteristics, and the full sample is generally representative of the socioeconomic and racial characteristics of the region from which it was drawn. Thus, there is some evidence for external validity in the study, at least regionally. These findings may not generalize as well to children in other geographic regions who live in different climates. For example, children living in very cold climates might only show a positive linear association between NBT and sleep quality or only a linear negative association between socioeconomic disparities and NBT. Further, the associations may differ by season with one set of associations when hot temperatures prevail and another set of associations when low temperatures prevail. Although season was not related to NBT or sleep variables in this study, it and other relevant climate variables are important to consider as control or antecedent variables.

A clear study limitation was the relatively small sample size for between-person analyses and statistical power to evaluate the roles of multiple between-person macro and micro sleep environment variables (e.g., outdoor temperature, thermal comfort of bedding and sleepwear, ambient noise, presence of siblings sharing the room). These would be relevant and important variables to consider in attempts to understand the multiple processes by which socioeconomic disparities are related to sleep. Some, such as ambient noise or access to thermal comfort, may be indirect mechanisms of these associations whereas others, such as outdoor temperature, may involve conditional mechanisms. For example, macro-level variables such as seasonal temperature variations may filter down to directly impact NBT, and this may be especially the case for children from lower socioeconomic backgrounds where families cannot afford to maintain a constant, ideal NBT (i.e., a mediated process that is moderated by socioeconomic disparities). Thus, studying microclimate variables, such as bedding, and macro-level variables, such as seasonal temperature variations and socioeconomic disparities, in conjunction with NBT remains an important future direction for researchers who aim to clarify the role that environments play in children’s sleep. Future applied research in this area may find success in exploring novel, cost effective, and efficient ways of regulating NBT and other sleep environment-related variables as part of holistic intervention to significantly benefit children and their development.

The broader goal of this research is to bring attention to the insidious and pervasive effects of socioeconomic disparities on children’s development and their developmental environments (Barros et al., 2010; Lawson et al., 2018) with an emphasis on children’s sleep environments as a potentially overlooked target of intervention efforts. There is evidence that programmatic intervention can improve children’s sleep and mitigate the risk of negative developmental outcomes (Blake et al., 2017; Gruber et al., 2016). Such holistic interventions might also target environmental factors such as NBT that are likely to influence children’s sleep, which the literature also has called for (Johnson et al., 2018). In particular, our finding that children in families with lower income-to-needs ratios are more likely to sleep at NBTs outside of the optimal range for the highest quality sleep suggests that these children may benefit from interventions that address bedroom temperature in addition to other aspects of the sleep environment. Children and families experiencing socioeconomic disadvantage face many structural barriers (Williams, 1999) including those related to the sleep environment (e.g., poorer quality housing; Johnson et al., 2018; 2021). Prevention and intervention efforts to improve sleep environments and facilitate children’s sleep can happen at institutional and societal levels as well as individual and family levels, thus contributing to the reduction of sleep-related health disparities along socioeconomic gradients (El-Sheikh et al., 2022; Sosso et al., 2021).

Supplementary Material

1

Highlights.

  • We evaluated links between family income, nighttime bedroom temperature, and children’s sleep

  • Children experience poorer sleep when nighttime bedroom temperatures are “too hot” or “too cold”

  • Children from lower income families are more likely to sleep at “too hot” or “too cold” conditions

  • There was some evidence that family income is linked to children’s sleep via nighttime bedroom temperature

References

  1. Alhasan DM, Gaston SA, & Jackson CL (2022). Investigate the complexities of environmental determinants of sleep health disparities. Sleep, 45(8), 145, 10.1093/sleep/zsac145. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Bach V, Maingourd Y, Libert JP, Oudart H, Muzet A, Lenzi P, & Johnson LC (1994). Effect of continuous heat exposure on sleep during partial sleep deprivation. Sleep, 17(1), 1–10. 10.1093/sleep/17.1.1 [DOI] [PubMed] [Google Scholar]
  3. Bagley EJ, Kelly RJ, Buckhalt JA, & El-Sheikh M (2015). What keeps low-SES children from sleeping well: The role of presleep worries and sleep environment. Sleep Medicine, 16(4), 496–502. 10.1016/j.sleep.2014.10.008 [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Blake MJ, Sheeber LB, Youssef GJ, Raniti MB, & Allen NB (2017). Systematic review and meta-analysis of adolescent cognitive-behavioral sleep interventions. Clinical Child and Family Psychology Review, 20(3), 227–249. 10.1007/s10567-017-0234-5 [DOI] [PubMed] [Google Scholar]
  5. Barros FC, Victora CG, Scherpbier R, & Gwatkin D (2010). Socioeconomic inequities in the health and nutrition of children in low/middle income countries. Revista de Saúde Pública, 44, 1–16. 10.1590/S0034-89102010000100001 [DOI] [PubMed] [Google Scholar]
  6. Bollen KA, & Stine R (1990). Direct and indirect effects: Classical and bootstrap estimates of variability. Sociological Methodology, 20, 115–140. 10.2307/271084 [DOI] [Google Scholar]
  7. Caddick ZA, Gregory K, Arsintescu L, & Flynn-Evans EE (2018). A review of the environmental parameters necessary for an optimal sleep environment. Building and Environment, 132, 11–20. 10.1016/j.buildenv.2018.01.020 [DOI] [Google Scholar]
  8. Cumming G (2008). Replication and p intervals: P values predict the future only vaguely, but confidence intervals do much better. Perspectives on Psychological Science, 3(4), 286–300. 10.1111/j.1745-6924.2008.00079.x [DOI] [PubMed] [Google Scholar]
  9. Du Prel JB, Hommel G, Röhrig B, & Blettner M (2009). Confidence interval or p-value? Part 4 of a series on evaluation of scientific publications. Deutsches Ärzteblatt International, 106(19), 335–339. 10.3238/arztebl.2009.0335 [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. El-Sheikh M, Gillis B, Saini E, Erath S, & Buckhalt JA (2022). Sleep and disparities in child and adolescent development. Child Development Perspectives, 16(4), 200–207. 10.1111/cdep.12465 [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. El‐Sheikh M, & Sadeh A (2015). I. Sleep and development: Introduction to the monograph. Monographs of the Society for Research in Child Development, 80(1), 1–14. 10.1111/mono.12141 [DOI] [PubMed] [Google Scholar]
  12. Enders CK, & Bandalos DL (2001). The relative performance of full information maximum likelihood estimation for missing data in structural equation models. Structural Equation Modeling, 8(3), 430–457. 10.1207/S15328007SEM0803_5 [DOI] [Google Scholar]
  13. Falk B (1998). Effects of thermal stress during rest and exercise in the paediatric population. Sports Medicine, 25(4), 221–240. 10.2165/00007256-199825040-00002 [DOI] [PubMed] [Google Scholar]
  14. Fronberg KM, Bai S, & Teti DM (2022). Household chaos mediates the link between family resources and child sleep. Sleep Health, 8(1), 121–129. 10.1016/j.sleh.2021.10.005 [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Gronlund CJ (2014). Racial and socioeconomic disparities in heat-related health effects and their mechanisms: A review. Current Epidemiology Reports, 1(3), 165–173. 10.1007/s40471-014-0014-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Gruber R, Somerville G, Bergmame L, Fontil L, & Paquin S (2016). School-based sleep education program improves sleep and academic performance of school-age children. Sleep Medicine, 21, 93–100. 10.1016/j.sleep.2016.01.012 [DOI] [PubMed] [Google Scholar]
  17. Heck RH, & Thomas SL (2020). An introduction to multilevel modeling techniques: MLM and SEM approaches (4th ed.). Routledge. [Google Scholar]
  18. Hoyniak CP, Bates JE, Camacho MC, McQuillan ME, Whalen DJ, Staples AD, Rudasill KM, & Deater-Deckard K (2022). The physical home environment and sleep: What matters most for sleep in early childhood. Journal of Family Psychology, 36(5), 757–769. 10.1037/fam0000977 [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Johnson DA, Billings ME, & Hale L (2018). Environmental determinants of insufficient sleep and sleep disorders: implications for population health. Current Epidemiology Reports, 5(2), 61–69. 10.1007/s40471-018-0139-y [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Johnson DA, Jackson CL, Guo N, Sofer T, Laden F, & Redline S (2021). Perceived home sleep environment: Associations of household-level factors and in-bed behaviors with actigraphy-based sleep duration and continuity in the Jackson Heart Sleep Study. Sleep, 44(11), zsab163. 10.1093/sleep/zsab163 [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Johnson DA, Thorpe RJ, McGrath JA, Jackson WB, & Jackson CL (2018). Black–white differences in housing type and sleep duration as well as sleep difficulties in the United States. International Journal of Environmental Research and Public Health, 15(4), 564. 10.3390/ijerph15040564 [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Joshi SS, Lesser TJ, Olsen JW, & O’Hara BF (2016). The importance of temperature and thermoregulation for optimal human sleep. Energy and Buildings, 131, 153–157. 10.1016/j.enbuild.2016.09.020 [DOI] [Google Scholar]
  23. Katafygiotou MC, & Serghides DK (2014). Thermal comfort of a typical secondary school building in Cyprus. Sustainable Cities and Society, 13, 303–312. 10.1016/j.scs.2014.03.004 [DOI] [Google Scholar]
  24. Lan L, Lian ZW, & Lin YB (2016). Comfortably cool bedroom environment during the initial phase of the sleeping period delays the onset of sleep in summer. Building and Environment, 103, 36–43. 10.1016/j.buildenv.2016.03.030 [DOI] [Google Scholar]
  25. Lee S, Ha JH, Moon DS, Youn S, Kim C, Park B, Kim MJ, Kim HW, & Chung S (2019). Effect of sleep environment of preschool children on children’s sleep problems and mothers’ mental health. Sleep and Biological Rhythms, 17(3), 277–285. 10.1007/s41105-019-00209-0 [DOI] [Google Scholar]
  26. Maas CJ, & Hox JJ (2004). Robustness issues in multilevel regression analysis. Statistica Neerlandica, 58(2), 127–137. 10.1046/j.0039-0402.2003.00252.x [DOI] [Google Scholar]
  27. Marchetti N, Cavazzini A, Pasti L, Catani M, Malagù C, & Guidi V (2015). A campus sustainability initiative: Indoor air quality monitoring in classrooms. XVIII AISEM Annual Conference. IEEE. 10.1109/AISEM.2015.7066774 [DOI] [Google Scholar]
  28. McWood LM, Zeringue MM, Piñón OM, Buckhalt JA, & El-Sheikh M (2022). Linear and nonlinear associations between the sleep environment, presleep conditions, and sleep in adolescence: Moderation by race and socioeconomic status. Sleep Medicine, 93, 90–99. 10.1016/j.sleep.2021.11.004 [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Mezick EJ, Matthews KA, Hall M, Strollo PJ Jr., Buysse DJ, Kamarck TW, Owens JF, & Reis SE (2008). Influence of race and socioeconomic status on sleep: Pittsburgh Sleep SCORE project. Psychosomatic Medicine, 70(4), 410–416. 10.1097/PSY.0b013e31816fdf21 [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Muthén LK, & Muthén BO (1998–2017). Mplus User’s Guide.Eighth Edition. Los Angeles, CA: Muthén & Muthén. [Google Scholar]
  31. Phelan JC, Link BG, & Tehranifar P (2010). Social conditions as fundamental causes of health inequalities: Theory, evidence, and policy implications. Journal of Health and Social Behavior, 51(1), S28–S40. 10.1177/0022146510383498 [DOI] [PubMed] [Google Scholar]
  32. Rubens SL, Miller MA, & Zeringue MM (2019). The sleep environment and its association with externalizing behaviors in a sample of low‐income adolescents. Journal of Community Psychology, 47(3), 628–640. 10.1002/jcop.22142 [DOI] [PubMed] [Google Scholar]
  33. Ruddell DM, Harlan SL, Grossman-Clarke S, & Buyantuyev A (2010). Risk and exposure to extreme heat in microclimates of Phoenix, AZ. In Showalter P & Lu Y (Eds.), Geospatial techniques in urban hazard and disaster analysis (pp. 179–202). Springer. [Google Scholar]
  34. Rush J, & Hofer SM (2017). V. Design‐based approaches for improving measurement in developmental science. Monographs of the Society for Research in Child Development, 82(2), 67–83. 10.1111/mono.12299 [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Sadeh A, Raviv A, Gruber R (2000). Sleep patterns and sleep disruptions in school-age children. Developmental Psychology, 36(3), 291–301. 10.1037/0012-1649.36.3.291 [DOI] [PubMed] [Google Scholar]
  36. Sadeh A, Sharkey KM, Carskadon MA (1994). Activity-based sleep-wake identification: An empirical test of methodological issues. Sleep, 17(3), 201–207. 10.1093/sleep/17.3.201 [DOI] [PubMed] [Google Scholar]
  37. San Miguel-Bellod J, González-Martínez P, & Sánchez-Ostiz A (2018). The relationship between poverty and indoor temperatures in winter: Determinants of cold homes in social housing contexts from the 40s–80s in Northern Spain. Energy and Buildings, 173, 428–442. 10.1016/j.enbuild.2018.05.022 [DOI] [Google Scholar]
  38. Shin M, Swan P, & Chow CM (2015). The validity of Actiwatch2 and SenseWear armband compared against polysomnography at different ambient temperature conditions. Sleep Science, 8(1), 9–15. 10.1016/j.slsci.2015.02.003 [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Shochat T, Cohen-Zion M, & Tzischinsky O (2014). Functional consequences of inadequate sleep in adolescents: A systematic review. Sleep Medicine Reviews, 18(1), 75–87. 10.1016/j.smrv.2013.03.005 [DOI] [PubMed] [Google Scholar]
  40. Sosso FAE, Holmes SD, & Weinstein AA (2021). Influence of socioeconomic status on objective sleep measurement: A systematic review and meta-analysis of actigraphy studies. Sleep Health, 7(4), 417–428. 10.1016/j.sleh.2021.05.005 [DOI] [PubMed] [Google Scholar]
  41. Sosso FAE, Kreidlmayer M, Pearson D, & Bendaoud I (2022). Towards a socioeconomic model of sleep health among the Canadian population: A systematic review of the relationship between age, income, employment, education, social class, socioeconomic status and sleep disparities. European Journal of Investigation in Health, Psychology and Education, 12(8), 1143–1167. 10.3390/ejihpe12080080 [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Spruyt K (2019). A review of developmental consequences of poor sleep in childhood. Sleep Medicine, 60, 3–12. 10.1016/j.sleep.2018.11.021 [DOI] [PubMed] [Google Scholar]
  43. Togo F, Aizawa S, Arai JI, Yoshikawa S, Ishiwata T, Shephard RJ, & Aoyagi Y (2007). Influence on human sleep patterns of lowering and delaying the minimum core body temperature by slow changes in the thermal environment. Sleep, 30(6), 797–802. 10.1093/sleep/30.6.797 [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Williams DR (1999). Race, socioeconomic status, and health the added effects of racism and discrimination. Annals of the New York Academy of Sciences, 896(1), 173–188. 10.1111/j.1749-6632.1999.tb08114.x [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

1
NIHMS1884856-supplement-1.docx (75.9KB, docx)

RESOURCES