The NHANES is a continuous program to assess the health and nutritional status of the civilian population in the United States.; it is conducted by the National Center for Health Statistics (NCHS), which belongs to the U.S. Centers for Disease Control and Prevention (CDC). Details about the NHANES study design, study protocol, and data collection have been described elsewhere (19). NHANES is approved by NCHS Ethics Review Board and written informed consent was obtained from all participants. The data in the present study were publicly released and were used in compliance with its data usage guidelines.

The present study population was limited to seven cycles of continuous NHANES (2005–2018). After excluding individuals less than 18 years old, during pregnancy, and without definite responses about sleep disorder, the remaining participants were included in the final analysis.

The questionnaire about sleep disorder (SLQ) has been added to NHANES since 2005, so we extracted sleep disorder data from 2005. In all the seven cycles of NHANES (2005–2018), the specific question, SLQ050, “Have you ever told a doctor or other health professionals that you have trouble sleeping?” was asked, and those participants responding “Yes” were considered to have self-reported sleep disturbance; participants responding “No” were considered not to have self-reported sleep disturbance; participants who chose “Refused” or “Don’t know” or did not answer this question were excluded in our analysis.

All-cause and disease-specific mortality were determined by linking to the NDI (20), in which the public-use Linked Mortality Files (LMF) are available for 1999–2018 NHANES. Death from cardiovascular disease (CVD) was defined as codes I00–I09, I11, I13, I20–I51, and I60–I69; death from cancer was defined as codes C00-C97; death from chronic lower respiratory disease (LRD) was defined as codes J40-J47, according to the 10th revision of the International Statistical Classification of Diseases, Injuries, and Causes of Death (ICD-10) guidelines.

The follow-up time has been calculated using months from the date of the interview to the date of death or the end of the mortality period (31 December 2019) for each participant of the NHANES.

Demographics, basic physical examination (namely body mass index, BMI, calculated as weight in kilograms divided by the square of height in meters), behavioral factors, and chronic conditions collected when enrolled into the NHANES (baseline), along with an interview regarding sleep problems, were analyzed as covariates based on previous literature to reduce the effect of potential confounding (21–23).

Demographic covariates included age (years), sex (male, female), race (Mexican American, non-Hispanic white, non-Hispanic black, and others), education (less than 9th grade, 9–11th grade, high school, college or AA degree, college graduate or above), marital status (married, widowed, divorced, separated, never married, and living with partner), and family income to poverty ratio (PIR, a ratio of family income to poverty threshold, continuous from 0 to 4.99, values are 5 if the ratio is 5 or over).

Behavioral covariates included cigarette smoking (defined as never, former, and now smoker, based on responses to the following two questions: “Have you smoked at least 100 cigarettes in your entire life?” and “Do you now still smoke cigarettes?”); alcohol use, defined as never (had <12 drinks in lifetime), former (had ≥12 drinks in lifetime but did not drink last year), mild (≤1 drink per day for women or ≤2 drinks per day for men), moderate (≤2 drinks per day for women or ≤3 drinks per day for men), and heavy (>2 drinks per day for women or >3 drinks per day for men); caffeine consumption (evaluated as the mean consumption from 2 typical days and recorded in mg/day) (24); and overall diet quality (assessed by the Healthy Eating Index, HEI-2015) (25).

Chronic comorbidity conditions, namely hypertension, diabetes, and coronary heart disease (CHD) were obtained through questionnaires, based on the response of participants to the question of whether or not they were ever told by a doctor or other health professionals that they had a specific disease.

Statistical analyses were performed according to NHANES recommended guidelines (26). Since NHANES used a complex, multistage probability sampling design to select participants, and generated a representative sample of the US civilian non-institutionalized resident population, this survey design was considered and survey sample weights were used in most of our analyses unless otherwise noted. Data are presented as survey-weighted mean (95% confidence intervals, CIs) for continuous variables, or number (survey-weighted percentage, %) for categorical variables, respectively. Survey-weighted linear regression (svyglm) and survey-weighted Chi-square test (svytable) were used to detect statistical differences between the means and proportions between the two groups.

Univariate and multivariate Cox regression models were applied to evaluate the association between sleep disorders and mortality. For multivariate regression models, collinearity diagnostics were performed to ensure variance inflation factors (VIF) of all independent variables less than 5; then multiple stepwise regression was performed to determine the inclusion of each covariate. The crude and adjusted hazard ratios (HRs) and 95% CIs were calculated for the risk of sleep disturbance associated with all-cause or disease-specific death.

According to the recommendation of the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement (27), we simultaneously showed the results unadjusted and with different covariate adjustment strategies in three models. Minimally adjusted models (Model 1) were adjusted for age, sex, and race. Partially adjusted models (Model 2) were additionally adjusted for education, marital status, PIR, BMI, smoking, alcohol use, caffeine consumption, and HEI-2015. Fully adjusted models (Model 3) were additionally adjusted for common chronic comorbidities, namely hypertension, diabetes, and CHD.

The stratified analyses were performed to demonstrate the effect of self-reported sleep disturbance in different subgroups using the fully adjusted model except for the specific stratification variable and interactions of self-reported sleep disturbance with stratification variables, which were inspected by the likelihood ratio test.

Data were analyzed with the use of the statistical packages R (The R Foundation; http://www.r-project.org; version 4.2.0) and EmpowerStats (www.empowerstats.net, X&Y solutions, Inc. Boston, Massachusetts; version 4.1). A two-sided p < 0.05 was considered significantly different.

The total number of participants in 7 cycles of NHANES (2005–2018) was 70,190. After excluding 28,047 individuals with ages less than 18 years old, another 737 pregnant individuals, another 121 individuals ineligible for mortality follow-up (insufficient identifying data), and another 28 responding to the question, “Have you ever told a doctor or other health professional that you have trouble sleeping?”, with “Refused” or “Don’t know” or no response, 41,257 participants were finally included in our study (Figure 1).

The enrolled 41,257 participants represented 228.6 million US adults. The survey-weighted mean age was 46.6 years old (95% CI: 46.2–47.1). The cohort comprised 20,897 women (survey-weighted percentage, 51.1%); 16,879 non-Hispanic White (66.6%), 9,063 non-Hispanic Black (11.4%), 6,595 Mexican American (8.6%), and 8,702 participants of other ethnicities (13.4%) (Table 1).

The self-reported sleep disturbance was estimated to affect 27.0% (95% CI: 26.1%–27.8%) of the whole population. Table 1 also shows the demographics, behavioral variables, and chronic comorbidities in all participants stratified by those with or without the self-reported sleep disturbance. Subjects reporting sleep disturbance were likely to be older (50.5 vs. 45.2), female (58.6% vs. 48.4%), white (74.4% vs. 63.7%), smoker (22.6% vs. 18.1% for current smoker and 28.1% vs. 21.5% for former smoker), with higher BMI (30.1 vs. 28.4), diabetes (13.0% vs. 7.3%), hypertension (42.5% vs. 25.4%), and CHD (5.2% vs. 2.8%) at the time of their participation in NHANES, than those not complaining about sleep.

To depict the detailed manifestations of self-reported sleep disturbance, general sleep symptoms were compared according to whether participants had self-reported sleep disturbance. As shown in Table 2, it was more common for individuals with self-reported sleep disturbance to have shorter sleep duration, to have trouble falling asleep, to wake up during the night, to wake up too early in the morning, to snore or stop breathing during sleep, to take pills to help sleep, to feel like they are not getting enough sleep, to feel unrested and overly sleepy during the day.

During a median follow-up of 88 months (interquartile: 49–129), 4,521 deaths were documented until 31 December 2019. Among them, 1,363 deaths were due to cardiovascular disease (CVD); 1,026 deaths were due to malignant neoplasms; 228 deaths were due to chronic lower respiratory diseases (LRD), and more details about the less common causes of death were shown in Supplementary Figure S1.

The unadjusted weighted Kaplan-Meier curves for all-cause mortality and disease-specific mortality by self-reported sleep disturbance were shown in Figure 2, followed by weighted log-rank tests. Then, univariate Cox regression models were used to quantify the risk of self-reported sleep disturbance with HR of 1.47 (95% CI: 1.36–1.59) for all-cause mortality, HR of 1.32 (1.13–1.54) for CVD mortality HR of 1.30 (1.10–1.53) for cancer mortality, and HR of 2.49 (1.85–3.35) for chronic LRD mortality (Supplementary Table S1). Univariate analysis of other covariates was also summarized in Supplementary Table S1.

After adjustment for age, sex, and race, self-reported sleep disturbance was associated with elevated risks of all-cause mortality (adjusted HR: 1.26, 95% CI:1.16–1.38), CVD mortality (1.17, 1.01–1.35), and chronic LRD mortality (2.08, 1.51–2.87), but no cancer mortality (1.13, 0.95–1.36) (Table 3).

After further adjustment of other demographics and behavioral factors, namely education, marital status, PIR, BMI, smoking, alcohol use, caffeine consumption, and HEI-2015, self-reported sleep disturbance was associated with elevated risks of all-cause mortality (1.23, 1.10–1.38), CVD mortality (1.26, 1.03–1.54), chronic LRD mortality (2.04, 1.39–3.00), but no cancer mortality (1.11, 0.91–1.35, Table 3).

Finally, in the fully adjusted model of all covariates, with chronic comorbidities (diabetes, hypertension, CHD) included, self-reported sleep disturbance was still associated with elevated risks of all-cause mortality (1.17, 1.04–1.32) and chronic LRD mortality (1.88, 1.26–2.80), but not CVD mortality (1.19, 0.96–1.46) and cancer mortality (1.10, 0.90–1.35). As shown in Table 3, the associations in fully-adjusted models were largely attenuated, compared with that in non-adjusted models.

Stratified analyses were performed to demonstrate the effect of self-reported sleep disturbance in different subgroups using the fully adjusted model, which was adjusted for sex, age, race, education, marital status, PIR, BMI, smoke, alcohol use, caffeine consumption, HEI-2015 and comorbidities of hypertension, diabetes, and CHD, except for the specific stratification variable of that analysis. The self-reported sleep disturbance was a significant risk factor for all-cause mortality in populations of males (adjusted HR: 1.28, 95% CI: 1.09–1.50), less than 60 years old (1.47, 1.19–1.81), married (1.24, 1.03–1.49), with an education level of high school or below (1.20, 1.01–1.23), with low PIR (1.28, 1.09–1.51), with low BMI (1.23, 1.02–1.48), smoker (1.23, 1.07–1.41), with more caffeine consumption (1.23, 1.04–1.45), and without diabetes (1.20, 1.06–1.36) (Figure 3).

The interaction analyses were performed to test the possible interaction of self-reported sleep disturbance with the stratification variable after adjusting all other variables. The interaction with the dichotomous age of 60 years old is remarkable (p < 0.001), indicating that self-reported sleep disturbance would be a more significant risk factor for all-cause mortality in the young adults than in the old population. The interactions with other variables are not statistically significant (Figure 3).

Self-reported sleep disturbance is associated with a higher risk of all-cause mortality. Among the major leading causes of death, a higher risk of chronic LRD mortality was associated with the self-reported sleep disturbance. Future large-scale longitudinal studies with a more elaborate classification of sleep disturbances are still needed to clarify the cause-and-effect relationship. Finally, complaints about sleep disturbance should be paid attention to in public health evaluation; moreover, effective prevention and treatment of respiratory disease in individuals with sleep disturbances are recommended.