Among a cohort of women of older age living in the US, we estimated the prevalence of perceived indoor temperature extremes that disrupt sleep. We also characterized sleep disturbances among women who frequently vs. infrequently reported sleep-disrupting indoor temperature extremes (SDITE) and determined whether observed relationships differed across race and ethnicity as well as annual household income (AHI) categories. Overall, we found that reports of trouble sleeping due to ‘too hot’ temperatures were prevalent (9.0%) while reported trouble sleeping due to ‘too cold’ temperatures was less prevalent (1.4%). Consistent with our hypotheses, NH Black women were most likely to frequently report SDITE that were ‘too hot’; prevalence was six percentage points higher compared to NH White women. Inconsistent with our hypothesis, perceiving ‘too hot’ SDITE was most reported by women within the highest income category. Nonetheless, perceiving ‘too hot’ SDITE was associated with all poor sleep dimensions measured in this study except short sleep duration even after adjustment for trouble sleeping due to other factors (e.g., pain, inability to breathe comfortably), other relevant confounders, and multiple comparisons; associations were strongest with LSOL and daytime dysfunction – defined as trouble staying awake while driving, eating meals, or engaging in social activity. Moreover, the apparent additive influences of ‘too hot’ SDITE and self-identifying as Latina as well as of ‘too hot’ SDITE and low income (≤$49,999) on associations with LSOL were larger than the sum of each. Also consistent with our hypothesis, Latina women and women within the lowest AHI category most often reported SDITE that were ‘too cold’; Latina women had three times the prevalence observed among NH White women, and women with AHI ≤$49,999 versus ≥$100,000 had two times the prevalence. These SDITE perceived as ‘too cold’ were associated with poorer sleep dimensions, especially with a higher prevalence of LSOL, daytime dysfunction, and HPD sleep apnea – an outcome in which associations were strongest among Latina women. Notably, groups with either the highest burdens or strongest relationships (e.g., Latina, low-income) are populations more likely to experience energy insecurity or difficulty meeting household energy needs.⁴³

We also explored whether relationships varied by two potentially important effect modifiers, geographic region of residence and menopausal status. Aligned with observations of stronger additive associations with LSOL among Latina women, frequently vs. infrequently perceiving SDITE was more strongly associated with short sleep duration (‘too hot’), long sleep duration (‘too hot’), and HPD sleep apnea (‘too cold’) among women in Puerto Rico, the majority of which are Latina. Menopausal status did not act as an effect modifier; however, additive associations between perceiving SDITE that were ‘too cold’ with LSOL appear stronger among postmenopausal compared to menopausal women.

Explicitly, after adjustment and multiple comparison correction, women who reported frequent versus infrequent ‘too hot’ SDITE also reported longer sleep duration (>9 hours), LSOL, poor sleep maintenance, sleep medication use, daytime dysfunction, and HPD sleep apnea. Furthermore, after adjustment and multiple comparison correction, women who reported frequently versus infrequently perceiving ‘too cold’ SDITE had a higher prevalence of all poor sleep dimensions. For each temperature extreme, the highest magnitudes of association were with LSOL and daytime dysfunction. Overall, the findings suggest that women with perceived temperature extremes – whether too hot or too cold – are especially more likely to report delayed sleep onset and sleep that is generally more disturbed and non-restorative, given observed associations with daytime dysfunction.

The general relationships observed between perceived SDITE and sleep are mostly consistent with prior literature. Except for a few prior studies,^29,36 studies generally suggest higher indoor temperatures as associated with LSOL, poor sleep quality, and sleep disturbances;^{26–28,32,37,41,60–63} thus prior literature related to hot indoor temperatures and poorer sleep health mostly agree with our findings. In fact, our findings related to ‘too hot’ temperatures align with a recent systematic review that showed higher indoor temperatures as persistently associated with degraded sleep quality, whether measured subjectively or objectively.²² However, our results suggesting no associations between perceived ‘too hot’ temperatures and short sleep duration add to the mixed evidence to date, with some studies suggesting an association and other suggesting no association.^{22,29,32,61,64–70} Lastly, our findings related to perceiving ‘too cold’ SDITE are consistent with prior reviews and original research in which the majority reported cold indoor temperatures as associated with LSOL, insomnia symptoms, poor sleep quality, and lower time in slow wave sleep, which correspond with our observed associations.^{23,28,33–35,38,41}

Regarding differences by race and ethnicity as well as SES, our hypotheses were supported for certain groups and sleep dimensions. The most consistent differences in additive relationships were between perceived ‘too hot’ SDITE and LSOL in which both Latina ethnicity and ‘lower’ compared to ‘high’ AHI acted synergistically, suggesting potentially larger public health impacts on these groups. Moreover, associations between frequently versus infrequently perceiving ‘too hot’ SDITE and both short and long sleep duration were much stronger in Puerto Rico, a region which largely comprised Latina women, more so than in other regions.

Additionally, associations between frequently perceiving ‘too cold’ temperature extremes and a higher prevalence of HPD sleep apnea were stronger among Latina women compared to NH White women as well as in Puerto Rico. In combination, results suggest that sleep health among Latina women, particularly Latina women residing in Puerto Rico, may be particularly vulnerable to temperature extremes. Regarding synergism between ‘too hot’ SDITE and AHI, results were plausible even though women with ‘high’ AHI had higher unadjusted prevalence of perceiving ‘too hot’ SDITE. Factors like poorer sleep health even without the exposure among women with ‘low’ AHI and other SES-related differences may have amplified worse sleep outcomes. Lastly, for perceived sleep-disrupting ‘too cold’ temperatures in relation to sleep medication use, there was antagonistic interaction among women with middle versus high AHI; it is possible that higher income individuals may more frequently use sleep medication to assuage sleep disturbances related to sleep environments perceived as ‘too cold’.

Our results may be explained by synthesizing and connecting results from prior studies, including studies of objective sleep that demonstrated how ambient temperatures impact sleep architecture. Prior literature suggests intensified temperature extremes indoors compared to outdoors²⁵ and demonstrates strong relationships between perceived and actual temperatures.^32,41 Perceptions of temperature extremes that disrupt sleep may accurately reflect reality. Additionally, studies demonstrate that women are particularly susceptible to temperature extremes compared to men,⁴⁴ and that socially-relegated populations may encounter more temperature extremes than other populations due to factors such as energy insecurity,⁴² a social determinant of health related to low socioeconomic status that reflects residents’ limited ability to control indoor environments. The thermal environment is recognized as one of the primary causes of sleep disturbances, hypothesized as acting through biologic pathways, such as impacting receptors, neurotransmitters, neurons, and hormones that are involved in sleep and circadian rhythm activity.^40,71 Cold temperatures have been found to increase the number and duration of wakefulness periods and lighter sleep stages, and hot or high temperatures can reduce duration of deep sleep and increase sleep onset latency as well as wakefulness.^13,71 Therefore, the observed linkage between perceived temperature extremes and poor sleep – namely LSOL and daytime dysfunction may be explained by this accumulation of findings about temperatures and sleep architecture from prior literature.

While our findings related to effect modification and additive interaction extend the prior literature given limited prior assessments, they are generally congruent with prior literature. To our knowledge, only one prior study investigated both hot and cold indoor temperature extremes (versus one category alone) in a multiethnic US population while considering socioeconomic status – albeit using objective temperature measurements in a pediatric sample.³⁰ Consistent with our current study, there were U-shaped relationships between indoor temperature and actigraphy-assessed poor sleep dimensions. Partially in contrast to our study, indoor temperature extremes were more prevalent among children in families with lower SES while we observed only colder perceived temperatures to be more prevalent among women with low AHI. In another study using an unsupervised machine learning approach, women and low SES respondents were mostly likely to report uncomfortable temperatures, and uncomfortable temperatures were associated with self-reported sleep problems.³¹ Additionally, similar to our observations of effect modification, reviews of the literature on outdoor temperatures and sleep suggest stronger impacts on women, marginalized and socially-relegated populations, and individuals residing in the warmest regions (i.e., Puerto Rico in our study).^16,17,22 Additional studies of social characteristics as potential effect modifiers of the indoor temperature and sleep relationship are warranted.

There are limitations and strengths worthy of consideration. The cross-sectional design limits causal interpretations. Additionally, exposure misclassification is possible. We investigated perceived temperature extremes considered to cause the participants trouble sleeping.

Therefore, sleep disturbances are expected to be higher in the exposed compared to the unexposed group. However, all models were adjusted for trouble sleeping due to other reasons, which was endorsed by most participants. Thus, the modeling approach allowed for the potential isolation of sleep disturbances due to temperature. Additionally, our use of perception better approximates thermal comfort, which is an important yet understudied factor in temperature-sleep associations.^13,37,72,73 Nonetheless, additional studies that include both objective and subjective indoor temperatures are necessary.²² Further studies inclusive of indoor environmental factors like humidity, air flow, and bed microenvironment (e.g., bedding, air conditioning), which impact thermal comfort,^37,40 are also necessary. Our study included subjective sleep assessments, which are moderately correlated with objective sleep measurements;⁷⁴ therefore, additional studies using objectively-measured sleep are also warranted. There were also small samples sizes who reported some exposures (e.g., participants who reported ‘too cold’) and outcomes (e.g., daytime dysfunction), which resulted in imprecise estimates. Unmeasured confounding related to factors like seasonality is also possible. Additionally, ineligible Sister Study participants were more likely to have characteristics associated with both perceived SDITE and poor sleep; therefore, current results may be underestimations. Moreover, the Sister Study sample consisted of older women with higher SES than the general US population, thus limiting generalizability. However, we expect the mechanisms linking temperature to sleep to be applicable across other populations.

Nevertheless, additional studies with various racial and ethnic, sex, age, and SES groups are warranted. Despite the limitations, strengths include the sample heterogeneity in racial, ethnic, and geography of US women along with the large sample size, which allowed for robust investigation of important potential modifiers. Additionally, we captured multiple sleep dimensions with certain dimensions assessed using a validated instrument (i.e., PSQI).

Furthermore, by investigating perceived temperature extremes and sleep disturbances in the context of factors like SES and menopausal status, our findings contribute to the literature by considering physical and psychosocial characteristics of people sleeping in bedrooms, which has been identified as needed to guide bedroom design interventions to improve sleep.⁷⁵

In conclusion, we characterized sleep dimensions among an older-aged population of Latina, NH Black, and NH White US women who reported perceived indoor temperatures.

Women who reported frequent trouble sleeping due to being ‘too hot’ were more likely to identify as NH Black while women who reported frequent trouble sleeping due to being ‘too cold’ were more likely to identify as Latina; these SDITE were consistently strongly associated with LSOL and daytime dysfunction. Additionally, certain associations were generally stronger, and burdens were higher among Latina women, including Latina women in Puerto Rico. Moreover, we assessed perceived indoor temperature using the established PSQI, which can be used in many existing studies to capture thermal comfort coupled with the collection of objectively measured temperature and other features of the indoor/bedroom microenvironment to build upon our approach. Although more studies are warranted, optimizing indoor temperatures can plausibly improve sleep health and serve as a target for interventions to address racial and ethnic sleep disparities among older-aged US women.