Day-by-day changes in the indoor environment and sleep health among African American adults

Abstract

Objectives

The study aimed to examine the association between day-by-day changes in the indoor physical environment and dimensions of sleep health among African American adults.

Methods

A total of 36 African American participants with at least five consecutive days of objective sleep measured by actigraphy (417 person-days) were included in the analysis. Indoor environmental factors, including temperature, carbon dioxide, noise, humidity, and barometric pressure, were objectively measured by an indoor environmental monitor. Linear mixed models were utilized to assess the within-person association between day-by-day changes in indoor environmental factors and sleep efficiency, wakefulness after sleep onset, sleep duration, and linear models for the association with sleep regularity, adjusting for sex, age, educational attainment, and the number of household members. The joint associations of all factors with sleep measures were assessed using quantile g-computation.

Results

The participants' mean (standard deviation) age was 55.3 (8.1), and most were female and had a bachelor's degree or higher. In complete case analysis (N = 29, 288 person-days), higher noise levels were associated with lower sleep efficiency (β = −1.81%, 95% CI: −2.00, −0.45), especially among those in the highest quartile vs. the lowest (β = −2.73%, 95% CI: −4.91, −0.54). Higher pressure levels were associated with longer sleep duration (β = 15.22 minutes, 95% CI: 1.11, 28.02) and more irregular sleep (β = 16.07 minutes, 95% CI: 1.40, 30.73). Having higher levels of all environmental factors was associated with lower sleep efficiency (β = −1.38%, 95% CI: −3.01, 0.26). Multiple-imputation analysis yielded consistent results.

Conclusions

Day-by-day changes in indoor physical environment were associated with sleep efficiency, duration, and irregularity among African American adults in the Atlanta Metro area.

Introduction

Sleep health is an important indicator of health. However, according to the National Health Interview Survey (NHIS), about 30% of adults in the United States (US) report sleeping less than the recommended minimum of 7 hours of sleep per day.¹ In addition, 15% of adults had trouble falling asleep, and 18% had trouble staying asleep.² Inadequate sleep has been found to be a significant contributor to various adverse health conditions, such as obesity,³ depression,⁴ and cardiovascular disease.⁵ Therefore, it is warranted to identify the determinants of inadequate sleep as potential targets to improve population sleep health and reduce subsequent health burdens.

Although inadequate sleep is a general health challenge in the US, it is disproportionately prevalent among racial and ethnic minoritized individuals. For example, African American adults have significantly worse sleep health compared with White adults, such as shorter sleep duration and poorer quality.^6,7 Given that inadequate sleep is a significant risk factor for adverse health outcomes, the health disparities among African Americans may be attributed to the sleep disparities that they experience.^8,9 Therefore, studying determinants of inadequate sleep among African American individuals is critical. As a result of structural barriers related to unfair treatment, African American populations have been disenfranchised from accessing adequate housing, education, employment, and other critical resources or opportunities for centuries. The profound impact of those barriers results in African American populations disproportionately having low socioeconomic status and being segregated into underfunded and neglected areas with poor environments and housing conditions.¹⁰ Residing in areas with adverse physical environments, such as excessive noise, extreme temperature and humidity, and severe air pollution, can substantially disrupt sleep.^11,12 Noise disrupts sleep by activating the auditory system to trigger stress reactions even while asleep.¹³ Humid heat could interfere with the cooling process of the core body temperature, which is a necessary step in falling asleep.¹⁴ High barometric pressure driven by high temperature and humidity has also been found to induce somatic symptoms, such as headache^15,16 and pain,^17,18 that may disrupt sleep. As an indicator of air quality, a high concentration of carbon dioxide (CO₂) inhibits breathing during sleep¹⁹ and could impair sleep quality.²⁰ Besides the experimental evidence, existing epidemiologic studies support that the aforementioned environmental factors have been found to interfere with sleep.¹¹ However, most epidemiologic studies targeted outdoor or neighborhood-level environmental exposure,^21-26 and attention to indoor environmental factors is lacking.

Assessing indoor environmental factors for sleep health is essential because the indoor environment is potentially modifiable and can vary by individual behaviors and household structure independent of the variation of the outdoor environment,^27,28 which can extend the knowledge on the environmental contribution to sleep health. Indoor environmental measures may be more accurate in reflecting the individual’s actual environmental exposures and the impact on sleep.¹² The suggested ranges of indoor environmental factors that can benefit sleep health are: CO₂ ≤ 1000 ppm,²⁹ temperature 18 – 24 °C,³⁰ humidity 30%-50%,³¹ and noise ≤45 dBA.³² The currently limited studies on indoor environmental factors have examined air quality,^33,34 noise, ^33,35-38 temperature,^33,39-41 humidity,^33,34 and barometric pressure.²⁹ The findings generally suggest that indoor excessive noise, poorer air quality, and high temperature and humidity are important contributors to poorer sleep health, whereas evidence is limited in barometric pressure. However, most of those studies only consider a single environmental factor, which prevents the assessment of the multidimensional impact of the indoor environment on sleep. They are also limited by the lack of objective measurement of environmental factors or sleep, as well as repeated measures to account for the individual variability of daily environmental exposure and sleep. Currently, only one study has assessed multiple indoor environmental factors simultaneously and used repeated objective measures on the indoor environment and sleep.³³ Therefore, more comprehensive and accurate assessments of indoor environments and sleep health are essentially needed to extend the current evidence.

To address the above gaps, this study examined the objectively measured day-by-day changes in indoor environment factors using an indoor environmental monitor, including temperature, CO₂, noise, humidity, and barometric pressure, associated with objectively measured dimensions of sleep health via actigraphy (e.g., sleep efficiency, wakefulness after sleep onset, duration, and regularity). We hypothesized that higher levels of temperatures, CO₂, noise, humidity, and barometric pressure would be associated with poorer individual nights of sleep.

Participants and methods

Study population

We conducted the analysis among participants in the Sleep Home Environmental Exposures and Blood Pressure study (SHEEP). SHEEP was a pilot study designed to determine the feasibility of identifying social and environmental risk factors for insufficient sleep and sleep disorders, and examine the association between insufficient sleep and blood pressure among African American adults in Atlanta, Georgia. Participants of SHEEP (N = 42) were aged 18-70 and were recruited from existing research studies in Atlanta [the Myocardial Infarction and Mental Stress (MIMs) Study], and the Morehouse-Emory Cardiovascular Center for Health Equity (MECA),⁴² and referrals from current study participants. Institutional Review Board (IRB) approval was obtained from the Emory University IRB.

We contacted potential participants by phone or e-mail (N = 166) with an invitation to participate. Eligibility screeners were self-administered or conducted over the phone (N = 87), and those who we were unable to contact or were not interested in participating were not screened. Eligible participants were those who self-identified as African American or Black, lived in the Atlanta Metro area, spoke English, and were over the age of 18 years, under a BMI of 45 kg/m², currently not pregnant, had no physician-diagnosed history of cardiovascular disease, congestive heart failure, stroke, HIV or AIDS, liver disease, renal/kidney disease, or history of alcoholism and drug use (e.g., cocaine, heroin, and methamphetamines). Participants were excluded from the study if they were shift workers (working outside of a typical work schedule) (N = 71). Written and electronic informed consent was obtained from all participants (N = 68). Of those who consented, 42 participants were enrolled in the SHEEP study.

The first phase of the study occurred between March 2019 and November 2020, and trained research assistants completed the visits in the homes of participants. During the in-person visit, the research assistants obtained consent from participants, administered questionnaires, conducted an environmental audit of the home, and instructed participants on how to complete the sleep diaries, wear the sleep monitors (actigraphy and in-home polysomnography) and the ambulatory blood pressure device, and environmental monitors were placed in the homes. We also conducted second visits among a subset of participants enrolled pre COVID from March 2019-March 2020. Questionnaires were administered via phone or videoconferencing (i.e., Zoom), and participants were instructed to wear the ActiWatch, which was delivered to their homes.

The second phase occurred from January 2021-April 2021, and the protocol was altered due to the COVID-19 pandemic. During social distancing regulations, we contacted potential participants via phone/videoconferencing (i.e., Zoom) to administer questionnaires and instruct them on how to use the devices and complete the sleep diaries. Members of the research team dropped off devices at participants’ homes using contactless and social distancing guidelines.

Overall, participants completed 10-15 days of actigraphy, environmental monitoring, and sleep diaries. Some participants had additional records after day 15 due to the delay in retrieving the device. We kept these records in the following analyses, as the delay happened at random, and we do not expect it to cause bias in this study.

Objective sleep measure

Participants underwent wrist actigraphy to identify sleep/wake times, sleep timing, and quality of sleep. Actigraphy was collected using a Philips ActiWatch Spectrum Plus device, which participants were instructed to wear on their nondominant wrist while concurrently completing a brief daily sleep diary. The Cole-Kripke algorithm was used to differentiate waking and sleep, and a trained research assistant further defined the start and end of the sleep interval based on activity counts, lux levels, event markers, and the sleep diary. In this study, we analyzed sleep quality as sleep efficiency (percentage of time spent asleep/time in bed [%]), wakefulness after sleep onset (WASO, minute), and sleep duration (min/h) per main sleep on each day. We also measured sleep regularity (standard deviation [SD] of sleep duration, minute) across the study period, in which higher SD indicated more irregular sleep.

Physical indoor environmental exposure

Day-by-day temperature (°C), CO₂ (ppm), noise (dBA), humidity (%), and barometric pressure (pressure, hPa) were recorded by Netatmo Environmental Monitor (Boulogne-Billancourt, France). The monitor was placed in the sleeping environment of the participant (e.g., bedroom) away from a heat/cooling source or a consistent source of noise. CO₂, temperature, and relative humidity were measured continuously in real time during a 10- to 15-day period. The 5-minute averages of noise levels were also recorded. The normal or recommended levels for indoor environmental factors are CO₂ ≤ 1000 ppm;²⁹ temperature 18 – 24 °C;³⁰ humidity 30%-50%;³¹ noise ≤45 dBA;³² and pressure 1009 - 1023 hPa (Atlanta [sea level adjusted]). In this study, the daily midnight-to-midnight averages of all environmental elements per day were used in the following analyses.

Statistical analysis

Due to the nonadherence during the actigraphy monitoring periods (10-15 days), not all participants had valid objective sleep data on each day. Therefore, participants with at least five consecutive days of valid objective sleep data were included in the analytic sample (N = 36). Descriptive statistics, mean (standard deviation [SD]) or median (interquartile range [IQR]) for continuous variables, and number (percentage [%]) for categorical variables, were performed to display the distribution of participants’ characteristics, including age (years), sex (female, male), education attainment (no bachelor’s degree, bachelor’s degree or above), household income (<$50,000, $50,000-$99,999, ≥$100,000), financial difficulty in the past year (yes, no), household size, and previously diagnosed sleep disorders (Sleep Apnea, Insomnia, and Restless Leg Syndrome).

In the primary analysis, we conducted a complete case analysis (N = 29) (i.e., no missing data in all included variables) fitting a linear mixed model to assess the association between day-by-day changes in indoor environmental factors and sleep efficiency/WASO/sleep duration using the formula below:

$$Slee{p}_{ij}={\beta }_0+{\beta }_{1}Indoo{r}_{ij}+BX+b_i+ϵ$$

In this formula, $Slee{p}_{ij}$ represented the average of sleep efficiency/WASO/sleep duration of the i^th subject on day j. $Indoo{r}_{ij}$ represented the average of each indoor environmental factor (temperature, CO₂, noise, humidity, and pressure) of the i^th subjects on day j. All factors were standardized to mean = 0 and SD = 1. Due to the skewed distribution of the CO₂ level, it was first log-transformed before the standardization. $X$ represented covariates adjusted in this model, including sex, age, educational attainment, and the number of household members. $b_i$ represented the random intercept for the subject. The model estimated the association within individuals before averaging across them, representing the average association between day-by-day changes in indoor environmental factors and sleep health of an individual. For the outcome of sleep regularity, we fitted a linear regression model by an average indoor environmental factor across the days of the study and included covariates. Because these environmental factors co-occur and are correlated, estimating their joint contribution to different domains of sleep health is of interest. To assess the joint effect of all indoor environmental exposures, we applied the quantile g-computation approach with bootstrapping within clusters (i.e., per participant) to obtain a valid estimate of the confidence interval.⁴³ Quantile g-computation is a mixture method that estimates the joint effect of all included exposures by one quantile, using the g-computation estimator.^43,44 Each environmental factor was transformed into quartiles under the original scale before entering the quantile g-computation, and the interpretation for the estimated association would be the change in outcome for every quartile increase in all included factors. No meaningful difference in demographic characteristics, environmental factors, or sleep measures was observed across the complete case sample (N = 29), the full analytic sample (N = 36), and the study population (N = 41) (Supplementary Table S1).

In the secondary analysis, we assessed the dose-response pattern of the association by categorizing each indoor environmental factor into quartiles under the original scale and estimating the quartile-specific association with each sleep measure. We also performed sensitivity analyses by (1) assessing the association between the average indoor environmental factors and sleep measures averaged across the days of the study, (2) using noon-to-noon averages of environmental factors (N = 16) to rerun the primary analysis, and (3) imputing the missing values in indoor environmental factors and covariates using multiple imputation in the full analytic sample (N = 36) with 10 imputed datasets and rerunning the primary and secondary analyses. There were 129 person-days with missing information on temperature, noise, CO2, and humidity, and 160 person-days with missing pressure. No person-day had a missing value in adjusted covariates. We conducted a priori power calculation and found that, with an effect size of 0.3, we would have 85% power to detect the association under the complete case analysis (N = 29) and 90% under the imputed analysis (N = 36). Estimated coefficients (β) and 95% confidence intervals (CI) were reported. All analyses were performed in R (version 4.3.2).

Results

A total of 36 participants were included in the analytic sample, with a total of 417 person-days. The median duration of study was 11 days, with a minimum of 6 days and a maximum of 18 days. The participants' mean (SD) age was 55.3 (8.1) (Table 1). Most participants were females and had a bachelor's degree or above. They had more than two persons in the household on average. The average levels (mean [SD]) of indoor environmental factors over time were temperature 22.6 (2.9) °C; CO₂ (median [IQR]) 671.9 (308.4) ppm, noise 41.6 (3.3) dBA, humidity 53.0 (11.0) %, and pressure 1016.2 (6.0) hPa (Table 2). The average levels of sleep measures were sleep efficiency 86.3 (5.4) %, WASO 52.2 (26.8) minutes, sleep duration 405.0 (110.4) minutes (i.e., 6.8 [1.8] hours), and sleep regularity 74.4 (31.7) minutes.

Table 1.

Participant demographics for the total sample, N = 36

Characteristic	N = 36
	n (%)a
Age, mean (SD)	55.3 (8.1)
Sex
Female	29 (80.6)
Male	7 (19.4)
Education attainment
No Bachelor's degree	13 (36.1)
Bachelor's degree or above	23 (63.9)
Household income
<$50,000	13 (38.2)
$50,000 - $99,999	12 (35.3)
≥$100,000	9 (26.5)
Financial difficulty in the past year
Yes	29 (82.9)
Household size, mean (SD)	2.5 (1.4)
Previously self-reported sleep disorders
Sleep Apnea	6 (16.7)
Insomnia	5 (13.9)
Restless Leg Syndrome	4 (11.1)

^aThe number in each characteristic may not add up to the total sample size due to missing data.

Table 2.

Distribution of indoor environmental factors and sleep measures by quartiles

	Mean (SD)a	Median (IQR)
Temperature (°C)
All	22.6 (2.9)	22.5 (3.6)
≤ 20.8	18.9 (1.3)	19.1 (2.2)
20.8 - 22.5	21.7 (0.5)	21.6 (0.8)
22.50- 24.4	23.3 (0.6)	23.3 (1.0)
≥ 24.4	26.2 (1.9)	25.7 (1.7)
CO2 (ppm)
All	727.0 (244.1)	671.9 (308.4)
≤ 548.2	480.8 (36.3)	483.1 (50.7)
548.2 - 671.9	605.2 (39.3)	603.2 (68.5)
671.9 - 856.6	754.4 (52.1)	737.7 (94.8)
≥ 856.6	1067.8 (200.7)	1024.4 (165.8)
Noise (dBA)
All	41.6 (3.3)	41.4 (4.1)
≤ 39.2	37.9 (1.0)	38.2 (1.5)
39.2 - 41.4	40.3 (0.6)	40.3 (0.9)
41.4 - 43.3	42.3 (0.5)	42.3 (0.8)
≥ 43.3	46.1 (2.5)	45.1 (4.0)
Humidity (%)
All	53.0 (11.0)	52.9 (13.7)
≤ 45.8	38.9 (4.6)	39.5 (7.30)
45.8 - 52.9	49.9 (2.1)	50.3 (3.5)
52.9 - 59.4	56.1 (1.9)	56.2 (3.3)
≥ 59.4	67.0 (6.1)	65.2 (12.0)
Pressure (hPa)
All	1016.2 (6.0)	1015.9 (7.9)
≤ 1012.4	1008.7 (2.8)	1009.0 (4.7)
1012.4 - 1015.9	1014.2 (1.0)	1014.3 (1.7)
1015.9 - 1020.2	1017.9 (1.2)	1017.5 (2.1)
≥ 1020.2	1023.9 (3.2)	1022.6 (4.7)
Sleep efficiency (%)
All	86.3 (5.4)	86.9 (7.2)
≤ 83.2	78.9 (4.0)	80.2 (3.7)
83.2 – 86.9	85.2 (1.1)	85.2 (1.8)
86.9 – 90.3	88.6 (0.9)	88.6 (1.5)
≥ 90.3	92.4 (1.5)	92.3 (1.8)
WASO (min)
All	52.2 (26.8)	47.8 (31.9)
≤ 34.8	23.8 (8.1)	26.2 (12.6)
34.8 – 47.8	40.3 (3.5)	39.5 (5.0)
47.8 – 66.6	56.9 (5.8)	56.0 (10.2)
≥ 66.6	88.0 (22.6)	80.8 (23.2)
Sleep duration (min)
All	405.0 (110.4)	405.8 (139.1)
≤ 340.0	263.0 (70.0)	290.0 (85.6)
340.0 – 405.8	374.4 (18.4)	375.8 (29.8)
405.8 – 479.12	442.1 (22.7)	444.0 (44.6)
≥ 479.12	540.6 (45.4)	533.0 (52.2)
Sleep regularity (min)
All	74.4 (31.7)	68.1 (68.1)
≤ 54.2	39.0 (9.6)	36.2 (36.2)
54.2 – 68.1	62.4 (5.3)	65.5 (65.5)
68.1 – 89.6	82.0 (5.3)	78.6 (78.6)
≥ 89.6	119.4 (16.2)	110.8 (110.8)

Abbreviation: WASO, wakefulness after sleep onset.

^aSuggested range of CO₂ ≤ 1000 ppm; temperature 18 – 24 °C; humidity 30%-50%; noise ≤45 dBA; and pressure 1009 - 1023 hPa (Atlanta [sea level adjusted])

The complete case analysis included 29 participants with 288 person-days. Every one-SD increase in day-by-day indoor noise was associated with a 1.81% decrease in sleep efficiency (95% CI: −2.00, −0.45) (Table 3). Temperature, CO₂, humidity, and pressure were not associated with sleep efficiency. Consistently, individuals exposed to louder day-by-day noise had more WASO (β = 2.78 minutes, 95% CI: −1.06, 6.67). Other factors were not found to be associated with WASO. Every one-SD increase in day-by-day pressure was associated with a 15.22-minute longer sleep duration (95% CI: 1.11, 28.02) and a 16.07-minute increase in sleep irregularity (95% CI: 1.40, 30.73). No other factors were found to be associated with sleep duration and regularity. Jointly exposed to all day-by-day indoor factors followed the trend of being associated with decreased sleep efficiency (β = −1.38%, 95% CI: −3.01, 0.26) (Table 4). No evidence was found for the joint association between indoor factors and other sleep measures.

Table 3.

Association between indoor environmental factors and sleep measures in the linear dose-response pattern

	Sleep efficiency (%)a	WASO (min)a	Sleep duration (min)a	Sleep regularity (min)b
	β (95% CI)c	β (95% CI)	β (95% CI)	β (95% CI)
Temperature (°C)	−0.04 (−1.10, 1.01)	1.19 (−3.46, 5.85)	−6.87 (−23.90, 10.17)	−6.79 (−18.58, 5.00)
CO2 (ppm)	−0.30 (−1.06, 0.47)	0.56 (−3.68, 4.80)	0.76 (−14.77, 16.29)	−0.65 (−13.96, 12.65)
Noise (dBA)	−1.10 (−1.78, −0.43)*	2.83 (−0.61, 6.28)	−5.05 (−17.32, 7.23)	−1.79 (−13.16, 9.59)
Humidity (%)	0.32 (−0.50, 1.13)	−1.06 (−5.53, 3.41)	−0.14 (−12.46, 12.18)	−5.61 (−16.83, 5.60)
Pressure (hPa)	−0.20 (−0.93, 0.52)	1.63 (−2.00, 5.26)	9.25 (−0.96, 19.46)	14.21 (2.76, 25.67)*

Abbreviation: WASO, wakefulness after sleep onset.

^aLinear mixed models were used to assess the day-by-day association between indoor environmental factors and sleep measures (except sleep regularity) using data of participants’ person-days

^bLinear models were used to assess the association between average levels of environmental factors across the days of the study and sleep regularity

^cCI not including 0 suggested that the estimate was statistically significant.

^*Statistically significant, p-value <0.05

Table 4.

Joint association between environmental factors and sleep measures using quantile g-computation

	β (95% CI)c
Sleep efficiency (%)a	−1.38 (−3.02, 0.27)
WASO (min)a	4.67 (−5.71, 15.05)
Sleep duration (min)a	−14.32 (−55.67, 27.04)
Sleep regularity (min)b	17.63 (−18.45, 53.71)

Abbreviation: WASO, wakefulness after sleep onset.

^aThe day-by-day association between indoor environmental factors and sleep measures (except sleep regularity) was assessed using data of participants’ person-days

^bThe association between average levels of environmental factors across the days of the study and sleep regularity was assessed

^cCI not including 0 suggested that the estimate was statistically significant.

When assessing dose-response pattern of the association using the quartiles of indoor environmental factors (Fig. 1, Supplementary Table S2), only individuals in the highest quartile (i.e., the fourth quartile) of indoor noise had a lower sleep efficiency (β = −2.73%, 95% CI: −4.91, −0.54) compared with those in the lowest quartile (i.e., the first quartile). We also found that being in the second quartile of CO₂ was associated with higher sleep efficiency (β = 2.44%, 95% CI: 0.65, 4.23) than in the lowest quartile, whereas other quartiles were not associated. A similar trend was also observed in the association between quartiles of noise and WASO. Individuals in higher quartiles of pressure had a longer sleep duration with a gradual increase in the effect size. Indoor environment with higher quartiles of pressure was also linked to more irregular sleep, especially among those in the highest quartile (β = 57.45 minutes, 95% CI: 19.72, 95.18). We found that participants in the highest quartile of pressure had larger variability of pressure compared with those in the second and third quartiles (Supplementary Table S3).

Figure 1. Association between indoor environmental factors and sleep measures by quartiles (as referenced to quartile 1 for each measure).

Note: Linear mixed models were used to assess the day-by-day association between indoor environmental factors and sleep measures (except sleep regularity) using data of participants’ person-days. Linear models were used to assess the association between average levels of indoor environmental factors across the days of the study and sleep regularity. CI not including 0 suggested that the estimate was statistically significant (p-value <0.05). Positive estimates indicate that environmental factors were associated with the increased sleep measures, whereas negative estimates indicate that environmental factors were associated with the decreased sleep measures.

In the sensitivity analysis, the association between average indoor environmental factors and sleep measures was generally consistent with the day-by-day association, with larger uncertainty around estimates (Supplementary Table S4). Using noon-to-noon averages of daily environmental factors yielded consistent results compared with the midnight-to-midnight measures in the main analysis (Supplementary Table S5). Multiple-imputation analysis also generally yielded consistent results with the complete case analysis but with less precision (Tables S6-S8).

Discussion

In this study, we found that day-by-day changes in the indoor environment were associated with lower sleep efficiency, more WASO, longer sleep duration, and irregular sleep among African American adults in the Atlanta Metro area. Specifically, higher indoor noise was associated with lower sleep efficiency, and higher pressure was associated with longer sleep duration but more irregular sleep. CO₂ was not associated with sleep efficiency in a dose-response pattern, but the participants within the second quartile of CO₂ had better efficiency compared with those in the lowest quartile. Joint exposure to higher indoor environmental factors was linked to poorer sleep efficiency with an imprecise estimate. The study suggests that the indoor environment could be an important contributor to sleep health, and maintaining a comfortable indoor environment may potentially ameliorate the prevalent inadequate sleep among the African American population.

Our findings are consistent with existing evidence that noise is a strong predictor of poor sleep quality.^33,35-38 Basner et al showed that higher indoor noise exposure was associated with decreased sleep efficiency in a linear dose-response pattern.²⁵ They also found that compared with the lowest quintile of noise exposure, exposure from the third to fifth quintile was associated with decreased sleep efficiency. In our study, however, the linear association between noise and sleep efficiency was mainly due to the significantly lower efficiency in the highest quartile compared with the lowest. The difference between studies could be due to the lower mean (SD) noise levels that our participants were exposed to (41.6 [3.3] dBA) compared with theirs (49.3 [7.8] dBA), although the difference could also be due to the differences in the environmental measurement device. The noise levels in other quartiles may not be intense enough to interrupt the sleep efficiency in our population. Basner et al have also found that higher noise exposure was associated with more WASO. We found the same trend as Basner et al but with imprecise estimates. Given the smaller sample size we had, we expect to gain more precise estimates of the association between noise and WASO with a larger sample size.

We did not find a linear dose-response pattern between CO₂ and sleep efficiency, but we observed that having a smaller increase in CO₂ (second quartile vs. lowest quartile) was unexpectedly associated with higher sleep efficiency. In contrast, Basner et al. found that higher CO₂ concentration was linked to lower sleep efficiency in a dose-response pattern. In our study, the average CO₂ concentration (mean [SD]) was 727.04 (73.02) ppm, whereas in Basner et al., the number was 1194.0 (523.5) ppm. The different distributions of CO₂ may explain the discrepancy of our findings in contrast to Basner et al, which may reflect different geographic settings (Atlanta vs. Philadelphia), different devices for measuring CO₂, or both. Some studies show that supplying a low dose of CO₂ to patients with sleep apnea and insomnia for therapeutic purposes improves sleep quality;^45,46 however, these findings were based either on relatively high concentrations of CO₂ delivered through a mask⁴⁵ or via an unconventional and uncontrolled CO₂ “spray” apparatus,⁴⁴ and those results likely have limited applicability to the more subtle effects of varying CO₂ concentrations in indoor air that we have observed in our study. The unexpected association in the second quartile may also be due to characteristics of this group that could confound the association. However, given the limited sample size, we may not be able to sufficiently control the potential confounding. Current evidence consistently suggests that high CO₂ levels can compromise sleep health by interfering with the respiratory system and preventing normal breathing during sleep.⁴⁷ High CO₂ levels have also been found to increase saliva cortisol levels after waking, which reflects increased stress response and activity of the sympathetic nervous system.⁴⁸ Therefore, controlling indoor CO₂ levels, such as through air filtration and ventilation, may be necessary to stabilize breathing and other relevant physiologic functions, thereby improving sleep health.

The association between barometric pressure and sleep has been rarely explored. One study suggests a U-shape relationship between barometric pressure and sleep onset, in which both lower and higher pressure, compared with the average levels, induce sleep tendency.⁴⁹ Our results are consistent with the evidence showing that higher pressure levels, particularly the highest quartile, were associated with longer sleep time. Given a normal pressure level with small variations in our population, we cannot detect an association of low pressure with sleep duration and quality. However, evidence also shows that overly increased or decreased barometric pressure can induce a series of symptoms, like headache^15,16 and pain,^17,18 that may interfere with sleep. The longer sleep duration may be a compensation for lower sleep quality resulting from adverse physical symptoms, as we also found that higher pressure was linked to higher WASO, although the result was not statistically significant. Interestingly, we observed that higher pressure was associated with more irregular sleep, especially among participants in the highest quartile of mean pressure across the study period, compared with those in the lowest. Sleep regularity was calculated by the variation of sleep time across time. Participants in the highest quartile of pressure had a greater variability of pressure than the second and third quartiles. Given that the highest quartile of pressure was the main driver of the association between pressure and sleep time, the higher variability could then yield a larger fluctuation in sleep time, thereby increasing sleep irregularity across days. This may imply that unstable weather, which contributes to variations in indoor pressure, subsequently affects individuals’ sleep regularity and may potentially be detrimental to health.

Although evidence generally suggests that high temperature and humidity are associated with poor sleep quality, ^34,39-41 we did not find evidence for this association. This is probably because our participants are from the Atlanta Metro area, in which air conditioning is typical. Therefore, they generally live in a comfortable range of indoor temperature and humidity, which may not affect their sleep quality. However, we found that the joint increase in all environmental factors was linked to lower sleep efficiency. Noise was the only factor significantly associated with lower sleep efficiency, whereas the associations of other environmental factors were largely null with wide CIs. Thus, the observed link between the joint increase in environmental factors and sleep efficiency may be mainly attributed to noise exposure, and other environmental factors introduced large uncertainty and diluted the estimate.

The study has several strengths. This is one of the few studies that systematically assessed the association between day-by-day indoor environmental factors and objectively measured sleep. We extend the current evidence to an African American population, which has a high prevalence of inadequate sleep. All environmental factors and sleep measures were measured objectively, which reduces the risk of measurement error, especially the dependent error as sleep and environmental factors were measured using different devices.

However, some limitations should be considered when interpreting the results. We used the daily average levels of environmental factors when assessing the association with sleep measures on the same day. Therefore, the cross-sectional design prevents causal inference. However, sleep would not cause indoor environmental changes, and the reverse causation is unlikely. Our sample size was small, which could have underpowered the study’s ability to detect the association and yield imprecise estimates. We did not comprehensively evaluate how air quality affected sleep in this study, as air pollutants, like PM_2.5, NO₂, and CO, were not collected. Our participants were all recruited from the Atlanta Metro area, which limits the generalizability. The present study was unable to account for sleep disorders due to limited statistical power given the sample size. It is plausible that indoor environmental factors may differentially influence sleep health among individuals with clinical sleep disorders compared with those without, particularly given the high prevalence of sleep disorders among African American adults. Future studies with larger, well-characterized samples should consider sleep disorder status as an effect modifier. Our measurements encompassed several different seasons, which could represent unmeasured confounding. However, we assumed that such major variations in climate would be reflected to some extent in our within-home measurements. Finally, the study is not free from the risk of confounding bias given the observational design, although we adjusted for multiple covariates to minimize the potential bias.

Conclusion

Among an African American population in Atlanta, GA, day-by-day changes in indoor environmental factors were associated with sleep efficiency, duration, and irregularity. The study emphasizes the importance of maintaining a comfortable indoor environment for optimal sleep health, which could be beneficial to long-term health outcomes. Actively monitoring the indoor environmental profile can help identify individuals at risk of poor sleep health. Air filtration and ventilation are potentially accessible and nontherapeutic approaches to improve sleep health by maintaining a comfortable indoor environment. Studies with more indoor environmental factors, such as air pollutants and light exposure, a larger sample size, and a longitudinal design, are warranted to establish a more robust linkage between the indoor environment and sleep and inform targeted household intervention strategies on improving sleep health. More diverse populations from different geographical regions should also be explored to achieve greater generalizability. The indoor environment and sleep are modifiable; thus, targeting these factors may help to improve population health.

What was known

Sleep health is influenced by multiple environmental factors (e.g., carbon dioxide, noise, and temperature). African American adults have a disproportionately higher rate of poor sleep health compared with their White counterparts. Environmental factors and housing conditions may contribute to poor sleep health among African American adults.

What this study adds

The study examined day-to-day variation in multiple indoor environmental exposures in relation to multiple dimensions of sleep health using objective measurements among African American adults. Findings suggest that indoor noise and barometric pressure were prospectively associated with sleep efficiency and sleep duration and irregularity, respectively, highlighting the indoor environment as a potentially modifiable determinant of sleep health. These results underscore the importance of studying within-person environmental variability and its implications for sleep health equity in populations disproportionately affected by adverse environmental conditions.

Data sharing

The data used in this study are available upon request.