Associations between Aircraft Noise Exposure and Self-Reported Sleep Duration and Quality in the United States-Based Prospective Nurses' Health Study Cohort

Abstract

Background: Sleep disruption is linked with chronic disease, and aircraft noise can disrupt sleep. However, there are few investigations of aircraft noise and sleep in large cohorts.

Objectives: We examined associations between aircraft noise and self-reported sleep duration and quality in the Nurses' Health Study, a large prospective cohort.

Methods: Aircraft nighttime equivalent sound levels (Lnight) and day-night average sound levels (DNL) were modeled around 90 U.S. airports from 1995 to 2015 in 5-y intervals using the Aviation Environmental Design Tool and linked to geocoded participant residential addresses. Lnight exposure was dichotomized at the lowest modeled level of 45 A-weighted decibels [dB(A)] and at multiple cut points for DNL. Multiple categories of both metrics were compared with $<45$ dB(A). Self-reported short sleep duration ($<7$ h/24-h day) was ascertained in 2000, 2002, 2008, 2012, and 2014, and poor sleep quality (frequent trouble falling/staying asleep) was ascertained in 2000. We analyzed repeated sleep duration measures using generalized estimating equations and sleep quality by conditional logistic regression. We adjusted for participant-level demographics, behaviors, comorbidities, and environmental exposures (greenness and light at night) and examined effect modification.

Results: In 35,226 female nurses averaging 66.1 years of age at baseline, prevalence of short sleep duration and poor sleep quality were 29.6% and 13.1%, respectively. In multivariable models, exposure to Lnight $\ge 45$ dB(A) was associated with 23% [95% confidence interval (CI): 7%, 40%] greater odds of short sleep duration but was not associated with poor sleep quality (9% lower odds; 95% CI: $-30\%$, 19%). Increasing categories of Lnight and DNL $\ge 45$ dB(A) suggested an exposure-response relationship for short sleep duration. We observed higher magnitude associations among participants living in the West, near major cargo airports, and near water-adjacent airports and among those reporting no hearing loss.

Discussion: Aircraft noise was associated with short sleep duration in female nurses, modified by individual and airport characteristics. https://doi.org/10.1289/EHP10959.

Figure 1.

Map of 90 study airports across the United States symbolized by U.S. Census region and quartiles of Nurses’ Health Study (NHS) participants.

Figure 2.

Odds ratio (OR) point estimates and 95% confidence intervals (CIs) investigating exposure–response relationship ($p_{\text{Trend}}<0.01$) between categorical aircraft nighttime sound level (Lnight) exposure [$<45$ dB(A) (reference), 45–49 dB(A), and $\ge 50$ dB(A)] and short sleep duration ($<7$ h/24-h day), using GEEs from repeated measures in survey years 2000 (study baseline), 2002, 2008, 2012, and 2014 in the Nurses’ Health Study (NHS). OR and CI estimates can be found in Table S6. Models adjusted for age (age, ${\text{age}}^2$) were sequentially further adjusted for other demographics, behaviors, comorbidities, and environmental factors. Other demographics: U.S. region of residence, race, living alone, spouse’s education. Behaviors: smoking status, alcohol consumption. Comorbidities: diabetes, hypertension. Environmental: greenness (NDVI), LAN. Models of short sleep duration used GEEs to estimate odds from repeated measures in survey years 2000 (study baseline), 2002, 2008, 2012, and 2014. Note: dB(A), A-weighted decibel; GEE, generalized estimating equation; LAN, light at night; $N_{\text{cases}}$, number of cases; NDVI, Normalized Difference Vegetation Index; $N_{\text{obs}}$, number of observations.

Figure 3.

Odds ratio (OR) point estimates and 95% confidence intervals (CIs) investigating exposure–response relationship ($p_{\text{Trend}}=0.03$) between categorical aircraft day–night average sound level (DNL) exposure [$<45$ (reference), 45–54, 55–64, and $\ge 65$ dB(A)] and short sleep duration ($<7$ h/24-h day), using GEEs from repeated measures in survey years 2000 (study baseline), 2002, 2008, 2012, and 2014 in the Nurses’ Health Study (NHS). OR and CI estimates can be found in Table S6. Models adjusted for age (age, ${\text{age}}^2$) were sequentially further adjusted for other demographics, behaviors, comorbidities, and environmental factors. Other demographics: U.S. region of residence, race, living alone, spouse’s education. Behaviors: smoking status, alcohol consumption. Comorbidities: diabetes, hypertension. Environmental: greenness (NDVI), LAN. Models of short sleep duration used GEEs to estimate odds from repeated measures in survey years 2000 (study baseline), 2002, 2008, 2012, and 2014. Note: dB(A), A-weighted decibel; GEE, generalized estimating equation; LAN, light at night; $N_{\text{cases}}$, number of cases; NDVI, Normalized Difference Vegetation Index; $N_{\text{obs}}$, number of observations.

Figure 4.

Odds ratio (OR) point estimates and 95% confidence intervals (CIs) investigating exposure–response relationship ($p_{\text{Trend}}=0.37$) between categorical aircraft nighttime equivalent sound level (Lnight) exposure [$<45$ (reference), 45–49 dB(A), and $\ge 50$ dB(A)] and poor sleep quality (trouble falling/staying asleep $\ge$ “a good bit of the time”) using conditional logistic regression at study baseline (2000) in the Nurses’ Health Study (NHS). OR and CI estimates can be found in Table S6. Models were adjusted for age (age, ${\text{age}}^2$) other demographics, behaviors, comorbidities, and environmental factors. Other demographics: U.S. region of residence (removed from the region-specific models), race, living alone, spouse’s education. Behaviors: smoking status, alcohol consumption. Comorbidities: diabetes, hypertension. Environmental: greenness (NDVI), LAN. Conditional logistic regression models of sleep quality were used to estimate odds only for the baseline study year. Note: dB(A), A-weighted decibel; LAN, light at night; $N_{\text{cases}}$, number of cases; NDVI, Normalized Difference Vegetation Index; $N_{\text{obs}}$, number of observations.

Figure 5.

Odds ratio (OR) point estimates and 95% confidence intervals (CIs) investigating exposure–response relationship ($p_{\text{Trend}}=0.37$) between categorical aircraft day–night sound level (DNL) exposure [$<45$ (reference), 45–54, 55–64, and $\ge 65$ dB(A)] and poor sleep quality (trouble falling/staying asleep $\ge$ “a good bit of the time”) using conditional logistic regression at study baseline (2000) in the Nurses’ Health Study (NHS). OR and CI estimates can be found in Table S6. Models were adjusted for age (age, ${\text{age}}^2$) other demographics, behaviors, comorbidities, and environmental factors. Other demographics: U.S. region of residence (removed from the region-specific models), race, living alone, spouse’s education. Behaviors: smoking status, alcohol consumption. Comorbidities: diabetes, hypertension. Environmental: greenness (NDVI), LAN. Conditional logistic regression models of sleep quality were used to estimate odds only for the baseline study year. Note: dB(A), A-weighted decibel; LAN, light at night; $N_{\text{cases}}$, number of cases; NDVI, Normalized Difference Vegetation Index; $N_{\text{obs}}$, number of observations.