Individual and neighborhood socioeconomic status and long-term individual trajectories of sleep duration among Black and White adults: the Southern Community Cohort Study

Abstract

Study Objectives

Sleep duration can change over the life course; however, previous studies rarely investigated the association between socioeconomic status (SES) and individual sleep trajectories over time. We examined the association between baseline socioeconomic characteristics and long-term sleep trajectories among Black and White adults.

Methods

This study used data from the Southern Community Cohort Study (N = 45 035). Diverse trajectories of sleep duration were constructed using self-reported sleep duration at baseline and after ~10 years of follow-up. The associations between baseline socioeconomic characteristics and sleep trajectories were examined using multinomial logistic regression.

Results

Both Black and White participants experienced similar long-term individual sleep trajectories for baseline educational attainment and employment status albeit the associations appeared stronger among White participants. Lower education and unemployment were associated with higher odds of various suboptimal sleep trajectories suggesting worsening long-term sleep patterns among both racial groups. However, there were some racial differences in the experience of long-term sleep trajectories for household income and neighborhood SES. Household income was notably more important among White than Black individuals; lower household income was associated with higher odds of more suboptimal long-term sleep trajectories for White than Black individuals. Also, neighborhood SES was slightly more important among White than Black individuals; lower neighborhood SES was associated with higher odds of a few suboptimal long-term sleep trajectories for both racial groups.

Conclusions

Lower socioeconomic characteristics were associated with various suboptimal long-term sleep trajectories among Black and White participants. Substantial improvements in socio-economic characteristics may contribute to improved sleep patterns.

Graphical Abstract

Graphical Abstract.

Statement of Significance.

Sleep duration can be dynamic over time but existing research on the association between socioeconomic characteristics and long-term sleep trajectories. This study examined the link between baseline socioeconomic characteristics and individual long-term sleep trajectories. We show that lower socioeconomic characteristics were linked to diverse unhealthy long-term sleep trajectories among Black and White participants.

Introduction

Sleep is universal and performs several essential functions in human populations [1]. Sleep deficiencies may have deleterious short- and long-term consequences on human well-being [2, 3]. Specifically, both short and long sleep may have been linked to cognitive decline, cardiovascular disease, metabolic dysfunction, cancer, and premature death among numerous others [4–6]. Sleep is a dynamic function that may change over the life course. Recent studies have shown that long-term patterns in sleep duration may play a unique role in human health. For instance, various studies have linked decreases and increases in sleep duration to cardiovascular and non-cardiovascular deaths [7, 8], higher risks of hypertension [9], type 2 diabetes [10], as well as impaired cognitive function [11], and subsequent cancer risk [12]. However, studies focusing on long-term individual trajectories of sleep are still limited, and it remains unclear how individual and contextual factors may shape sleep trajectories.

A wide range of studies has shown significant associations between individual-level and neighborhood-level socioeconomic status (SES) and one-time measures of sleep duration [13–16]. Other studies have also examined sleep trends over time based on cross-sectional national surveys and have found reduced sleep duration [17] or increased short and long sleep duration in the United States over time [18], particularly among race/ethnic minorities. Conversely, there is a dearth of evidence regarding whether individual- and neighborhood-level socioeconomic factors play a role in long-term individual sleep trajectories among longitudinal cohorts. Meanwhile, several previous studies have shown that neighborhood socioeconomic characteristics may play a significant role in sleep duration [19–22]. However, studies have not examined the association of neighborhood SES with long-term sleep trajectories. To address this gap, we examined the association between baseline individual- and ­neighborhood-SES and sleep duration patterns over ten years in the Southern Community Cohort Study (SCCS), a socio-demographically unique cohort of middle-to-older aged Black and White Americans in the southeastern United States. Various existing studies have established significant sleep duration disparities between Black and White individuals in the United States over time [17, 18]. Besides, our initial interaction analysis showed significant results for some predictors such as education and household income by race. Therefore, the current study stratified the analysis by Black and White individuals to examine them separately. The study is aimed at providing a better understanding of the various long-term sleep trajectories and their associations with multiple ­individual- and neighborhood-level socioeconomic characteristics among Black and White individuals.

Methods

Study population

The SCCS was established in 2002 to investigate population health disparities among people ages 40–79 in the Southeastern United States. The majority of the participants (86 per cent) were recruited from community health centers that served uninsured or underinsured patients in 12 Southeastern states (Alabama, Arkansas, Florida, Georgia, Kentucky, Louisiana, Mississippi, North Carolina, South Carolina, Tennessee, Virginia, and West Virginia). The rest were enrolled by a questionnaire sent to a random sample of the general population in these states. Details of the survey design and study population have been previously published [23]. The study recruited over 75 248 participants who were predominantly African Americans (nearly 70 per cent). At the baseline, participants reported their residential address, demographic characteristics, health behaviors including sleep, as well as health status. From 2008 to 2013, participants reported residential address and sleep duration again in a follow-up survey. The Institutional Review Boards at Vanderbilt University and Meharry Medical College provided ethical approval for the study. Of the 75 248 Black and White participants at baseline, we excluded participants with missing information on sleep duration at baseline or follow-up, and the final sample for the analysis comprised 45 035 participants (White = 16 276, Black = 28 759).

Individual trajectories of sleep duration

The participants were asked to report hours of sleep during 24 h on weekdays and weekends separately, and the mean sleep duration at each time point was calculated based on the weighted average of weekdays and weekends sleep duration (5*weekday duration + 2*weekend duration/7). The mean sleep duration was then categorized as short (<7 h), normal (7–<9 h), and long sleep (≥9 h) for both surveys. We generated nine mutually exclusive groups for long-term sleep trajectories between the two surveys: short-short (SS), short-normal (SN), short-long (SL), normal-short (NS), ­normal-normal (NN), normal-long (NL), long-short (LS), ­long-­normal (LN), and long-long (LL). The individual sleep trajectories were put under five groups – stable healthy, stable unhealthy, unstable unhealthy, improving, and deteriorating – indicating the pattern or direction of the trajectories. Of these, normal-normal (NN) was used as the reference group.

Socioeconomic characteristics

Individual-level SES factors included educational attainment (<High school, High school, Vocational/some college, and college or more), employment status (Employed, Unemployed), and household income (<$15 000, $15 000–$24 999, $25 000–$49 999, $50 000+). Baseline addresses of participants were geocoded into geographical coordinates and linked to the 2000 United States census tract-level data. Drawing on the method developed by Messer et al [24]., we constructed a neighborhood socioeconomic deprivation index using 14 census tract-level variables (percent of total less than high school, percent of total unemployed, percent of households with income below poverty, percent of households with an income <$30 000, percent of households on public assistance, percent of households with no car, percent of unemployed men, percent of renter-occupied housing units, percent of housing units vacant, median value of all owner-occupied housing units, percent of female-headed household with dependent children, percent of non-Hispanic Black participants, percent of residents 65 years and above, percent of persons in the same residence since 5 years before the census). We carried out principal component analysis (PCA) on this census tract-level variables and calculated the deprivation index based on the PCA score of the first component. The score was then divided into quintiles and coded as Q1 (Lowest SES), Q2, Q3, Q4, and Q5 (Highest SES).

Statistical analysis

We performed descriptive analyses on baseline background characteristics stratified by sleep duration trajectories and race. Mean, standard deviation, and percentage were calculated, and the Pearson’s Chi-square test was performed to examine the statistical significance of group differences for continuous and categorical variables, respectively. Although multilevel models are often used to account for clustering within neighborhoods, we chose a single-level multinomial logistic regression model because 86 per cent of the tracts contain 10 or fewer people including a considerable percentage (about 39 per cent) containing only one participant. Thus, while a single-level model is suitable for estimating contextual effects in our data [25], multilevel models are more efficient when the number of participants per tract is larger. Thus, the data structure suggests that there is no strong motivation for a more complex multilevel model compared to a single-level model. In the multinomial logistic regression analysis, we focused on examining the association between baseline individual- and neighborhood-level socioeconomic characteristics and individual sleep trajectories, comparing each sleep category to the reference (normal-normal or NN) group. The analysis was stratified by race following the finding of statistically significant interaction analysis results between some predictors and race. Adjusted odds ratios and 95% confidence intervals were calculated from models that included education, household income, employment status, neighborhood SES, age group, and sex. All analyses were conducted using the R statistical software (version 4.0.3) [26].

Results

Table 1 presents descriptive statistics of baseline study characteristics by the sleep duration trajectories, for Black and White participants. For both Black and White participants, female participants formed the majority of the study sample. When compared to Black participants, White participants were on average older and more likely to have a college degree and a household income of $50 000 or more. White participants were also more likely to be married and live in neighborhoods with higher SES than Black participants. Among White participants, when compared to those in the normal-normal (NN) trajectory, participants in all other trajectories were less likely to have a college-or-higher education, be employed, report an income of $50 000 or higher, be married, or live in neighborhoods within the highest quintile of SES. In contrast, among Black participants, differences ­in ­individual- and neighborhood-level characteristics across sleep trajectories were less prominent but remain significant.

Table 1.

Baseline characteristics by individual trajectories of sleep duration among Black and White participants in the SCCS

Characteristics	Individual trajectories of sleep duration									P
	Stable healthy	Stable unhealthy		Unstable unhealthy		Improving		Deteriorating
	NN	SS	LL	SL	LS	SN	LN	NS	NL
WHITE
N (%)	5938 (36.5)	3012 (18.5)	751 (4.6)	679 (4.2)	187 (1.1)	2400 (14.7)	724 (4.4)	1282 (7.9)	1303 (8.0)
Age, mean (SD)	56 (9.1)	53 (8.5)	56 (9.1)	53 (8.9)	52 (8.6)	55 (8.8)	55 (8.9)	54 (9.2)	56 (9.2)	<.001
Female, %	60.6	67.6	62.1	71.0	65.2	65.4	66.0	68.3	65.0	<.001
College/higher, %	34.4	17.3	21.4	8.5	9.1	20.5	18.0	20.6	23.0	<.001
Employed, %	53.0	42.1	25.4	26.4	29.5	47.7	30.3	45.3	39.3	<.001
Household income, 50 000+, %	35.3	17.1	16.6	6.3	6.0	21.3	16.4	21.0	19.0	<.001
Married, %	61.9	51.1	50.1	40.0	46.5	51.6	52.5	56.1	50.5	<.001
Neighborhood SES, %
Q1 (Lowest SES)	9.9	13.9	13.4	16.4	16.9	11.4	15.2	14.6	14.3
Q2	19.9	23.0	25.0	26.6	19.7	24.1	24.8	22.9	22.0
Q3	21.7	22.7	22.0	23.7	32.2	24.6	20.8	23.1	22.3	<.001
Q4	24.9	25.8	24.0	23.4	17.5	24.5	23.7	25.5	23.4
Q5 (Highest SES)	23.6	14.6	15.7	9.9	13.7	15.4	15.5	14.0	17.9
BLACK
N (%)	6951 (24.2)	6095 (21.2)	1377 (4.8)	1519 (5.3)	979 (3.4)	4547 (15.8)	1975 (6.9)	3205 (11.1)	2111 (7.3)
Age, Mean (SD)	53 (8.7)	51 (7.9)	52 (8.8)	52 (8.3)	51 (8.4)	52 (8.2)	52 (8.5)	52 (8.6)	53 (9.0)	<.001
Female, %	65.9	68.4	66.0	63.8	65.6	66.4	63.2	67.1	63.3	<.001
College/higher, %	15.7	15.6	9.0	8.4	8.2	13.9	9.0	11.7	10.6	<.001
Employed, %	47.3	46.6	26.4	34.9	30.3	46.5	36.9	43.5	35.2	<.001
Household income, 50 000+, %	9.1	9.9	4.6	4.0	4.1	8.6	5.4	6.9	4.6	<.001
Married, %	35.9	31.3	28.0	27.7	31.6	33.2	31.3	31.1	32.0	<.001
Neighborhood SES, %
Q1 (Lowest SES)	55.7	54.7	62.5	61.6	61.6	56.8	61.4	57.1	60.1
Q2	20.2	19.5	18.8	19.4	19.6	20.5	19.3	19.8	20.4
Q3	10.2	10.5	9.4	8.5	10.1	9.6	9.7	9.6	9.4	<.001
Q4	8.5	9.8	6.2	6.9	5.6	7.9	5.9	7.9	7.2
Q5 (Highest SES)	5.4	5.4	3.2	3.6	3.1	5.2	3.7	5.6	2.9

SES: Socioeconomic Status; SD: Standard deviation; P-value: Kruskal–Wallis rank sum and Pearson’s Chi-squared test; NN, Normal-normal; SS, Short-Short; LL, Long-Long; SL, Short-Long; LS, Long-Short; SN, Short-Normal; LN, Long-Normal; NS, Normal-Short; LS, Long-Short.

The multivariable analysis was stratified by Black and White participants after observing significant results for race (p < 0.001), education (p < 0.001), and household income by race (p < 0.001) in an interaction analysis (Results not shown). Associations between individual-level and neighborhood-level SES and long-term individual trajectories of sleep duration among White and Black participants are presented in Table 2. The results show that long-term sleep patterns were associated with baseline educational attainment, household income, employment status, and neighborhood SES among both races. In addition, some notable differences existed in some of the associations between Black and White participants. Among White participants, lower educational attainment was associated with higher odds of four out of six suboptimal sleep trajectories (SS, SL, LS, NS), and was also associated with increased odds of two optimal improving trajectories (SN, LN). Of the four suboptimal sleep trajectories, the unstable unhealthy group (SL and LS) showed the strongest association with lower education: white participants who did not complete high school were 2.52 and 2.28 times more likely to experience SL and LS trajectories, respectively, than their counterparts with a college education. Lower household income was consistently associated with higher odds of all six suboptimal sleep trajectories (SS, LL, SL, LS, NS, NL) and two optimal improving trajectories (SN, LN). The unstable unhealthy group (SL and LS) again showed the strongest association whereby white participants with the lowest household income had 5.92 and 6.08 odds of experiencing SL and LS trajectories, respectively, compared with those with $50 000+ household income. Also, when compared to White participants who were employed at baseline, those who were unemployed had higher odds of experiencing five of six suboptimal sleep trajectories (SS, LL, SL, LS, NL) but also one of two improving trajectories (LN): LL under stable unhealthy (OR (95% CI)), 2.67 (2.19, 3.25)), and LN under improving (1.97 (1.58, 2.33)) trajectories. Lastly, lower neighborhood SES was associated with higher odds of only three of the six suboptimal sleep trajectories (SS, SL, NS) and one improving trajectory (SN) for White individuals: NS under deteriorating (1.67 (1.31, 2.13)) and SN under improving (1.27 (1.08, 1.50)) trajectories.

Table 2.

Associations (OR (95% CI)) between baseline socioeconomic characteristics and individual trajectories of sleep duration in Black and White participants in the SCCS

	Individual trajectories of sleep duration
	Stable healthy	Stable unhealthy		Unstable unhealthy		Improving		Deteriorating
	NN	SS	LL	SL	LS	SN	LN	NS	NL
WHITE
Education
<High school	Ref	1.93 (1.63, 2.29)	0.85 (0.63, 1.14)	2.52 (1.78, 3.57)	2.28 (1.21, 4.30)	1.84 (1.54, 2.21)	1.64 (1.22, 2.20)	1.72 (1.36, 2.16)	1.07 (0.85, 1.34)
High school	Ref	1.63 (1.41, 1.89)	0.97 (0.76, 1.25)	2.01 (1.45, 2.80)	2.19 (1.21, 3.94)	1.52 (1.30, 1.77)	1.39 (1.07, 1.80)	1.45 (1.19, 1.77)	1.08 (0.89, 1.31)
Vocational/some college	Ref	1.39 (1.21, 1.60)	1.12 (0.89, 1.42)	2.05 (1.48, 2.84)	1.73 (0.95, 3.14)	1.35 (1.17, 1.57)	1.14 (0.88, 1.48)	1.30 (1.08, 1.57)	1.09 (0.90, 1.30)
College/higher	Ref	Ref	Ref	Ref	Ref	Ref	Ref	Ref	Ref
Household income
<15 000	Ref	2.33 (1.99, 2.73)	2.37 (1.79, 3.12)	5.92 (4.05, 8.67)	6.08 (2.96, 12.5)	1.91 (1.62, 2.26)	2.23 (1.67, 2.96)	1.64 (1.33, 2.02)	2.53 (2.04, 3.13)
15 000–24 999	Ref	1.76 (1.49, 2.08)	1.96 (1.46, 2.63)	4.00 (2.69, 5.95)	3.15 (1.46, 6.81)	1.54 (1.30, 1.83)	1.70 (1.25, 2.31)	1.34 (1.07, 1.67)	1.94 (1.55, 2.43)
25 000–49 999	Ref	1.37 (1.18, 1.60)	1.41 (1.07, 1.87)	2.18 (1.46, 3.28)	3.47 (1.68, 7.16)	1.21 (1.04, 1.42)	1.37 (1.03, 1.82)	1.32 (1.08, 1.61)	1.49 (1.21, 1.83)
50 000+	Ref	Ref	Ref	Ref	Ref	Ref	Ref	Ref	Ref
Employment
Employed	Ref	Ref	Ref	Ref	Ref	Ref	Ref	Ref	Ref
Unemployed	Ref	1.17 (1.06, 1.31)	2.67 (2.19, 3.25)	1.79 (1.47, 2.20)	1.79 (1.23, 2.58)	0.91 (0.82, 1.02)	1.92 (1.58, 2.33)	1.07 (0.93, 1.24)	1.32 (1.15, 1.53)
Neighborhood SES
Q1 (Lowest SES)	Ref	1.18 (0.99, 1.42)	1.18 (0.87, 1.61)	1.46 (1.04, 2.06)	1.02 (0.58, 1.80)	1.10 (0.91, 1.35)	1.23 (0.90, 1.66)	1.67 (1.31, 2.13)	1.18 (0.93, 1.49)
Q2	Ref	1.10 (0.94, 1.29)	1.22 (0.93, 1.59)	1.38 (1.00, 1.89)	0.69 (0.40, 1.19)	1.27 (1.08, 1.50)	1.12 (0.86, 1.47)	1.37 (1.10, 1.71)	0.98 (0.80, 1.20)
Q3	Ref	1.07 (0.92, 1.25)	1.04 (0.79, 1.36)	1.28 (0.93, 1.76)	1.10 (0.67, 1.82)	1.25 (1.06, 1.47)	0.92 (0.70, 1.21)	1.32 (1.07, 1.64)	0.97 (0.80, 1.19)
Q4	Ref	1.21 (1.04, 1.40)	1.09 (0.84, 1.41)	1.36 (0.99, 1.86)	0.71 (0.41, 1.21)	1.21 (1.03, 1.41)	1.05 (0.81, 1.37)	1.42 (1.16, 1.75)	0.97 (0.80, 1.18)
Q5 (Highest SES)	Ref	Ref	Ref	Ref	Ref	Ref	Ref	Ref	Ref
BLACK
Education
<High school	Ref	0.98 (0.86, 1.12)	1.33 (1.04, 1.70)	1.63 (1.28, 2.06)	1.68 (1.25, 2.24)	1.14 (0.98, 1.31)	1.58 (1.28, 1.96)	1.31(1.11, 1.55)	1.18(0.97, 1.44)
High school	Ref	0.95(0.84, 1.07)	1.32(1.04, 1.66)	1.35 (1.07, 1.69)	1.39 (1.05, 1.84)	1.07(0.94, 1.22)	1.53(1.25, 1.87)	1.27(1.08, 1.48)	1.17 (0.97, 1.40)
Vocational/some college	Ref	1.11(0.99, 1.25)	1.23 (0.97, 1.55)	1.30(1.04, 1.64)	1.46 (1.10, 1.93)	1.15 (1.01, 1.30)	1.38 (1.13, 1.69)	1.24 (1.06, 1.44)	1.10 (0.91, 1.32)
College/higher	Ref	Ref	Ref	Ref	Ref	Ref	Ref	Ref	Ref
Household income
<15 000	Ref	0.88 (0.76, 1.02)	1.29 (0.94, 1.76)	1.81 (1.32, 2.47)	1.48 (1.02, 2.15)	0.97 (0.83, 1.15)	1.27 (0.98, 1.64)	1.20 (0.98, 1.45)	1.56 (1.20, 2.01)
15 000–24 999	Ref	0.83 (0.71, 0.96)	1.12 (0.81, 1.54)	1.58 (1.15, 2.16)	1.09 (0.74, 1.59)	0.96 (0.81, 1.13)	1.13 (0.87, 1.46)	1.07 (0.88, 1.30)	1.45 (1.12, 1.88)
25 000–49 999	Ref	0.87 (0.75, 1.00)	0.99 (0.71, 1.37)	1.13 (0.82, 1.57)	1.12 (0.76, 1.64)	0.94 (0.79, 1.11)	1.04 (0.80, 1.35)	1.01 (0.83, 1.23)	1.25 (0.96, 1.62)
50 000+	Ref	Ref	Ref	Ref	Ref	Ref	Ref	Ref	Ref
Employment
Employed	Ref	Ref	Ref	Ref	Ref	Ref	Ref	Ref	Ref
Unemployed	Ref	1.13 (1.04, 1.22)	2.32 (2.01, 2.67)	1.46 (1.29, 1.660	1.84 (1.57, 2.16)	1.06 (0.97, 1.15)	1.39 (1.24, 1.56)	1.11 (1.01, 1.22)	1.49 (1.33, 1.66)
Neighborhood SES
Q1 (Lowest SES)	Ref	0.99 (0.84, 1.17)	1.42 (1.01, 1.99)	1.12 (0.83, 1.51)	1.29 (0.88, 1.89)	1.02 (0.85, 1.22)	1.28 (0.98, 1.68)	0.86 (0.70, 1.04)	1.65 (1.23, 2.21)
Q2	Ref	0.98 (0.82, 1.16)	1.30 (0.91, 1.860	1.09 (0.80, 1.50)	1.24 (0.83, 1.85)	1.03 (0.85, 1.25)	1.19 (0.89, 1.59)	0.86 (0.70, 1.06)	1.64 (1.21, 2.23)
Q3	Ref	1.00 (0.83, 1.21)	1.35 (0.93, 1.98)	0.98 (0.69, 1.39)	1.28 (0.83, 1.96)	0.94 (0.77, 1.16)	1.19 (0.88, 1.63)	0.83 (0.66, 1.04)	1.56 (1.12, 2.16)
Q4	Ref	1.15 (0.95, 1.39)	1.15 (0.77, 1.71)	1.06 (0.75, 1.52)	0.94 (0.59, 1.49)	0.97 (0.78, 1.20)	0.91 (0.65, 1.27)	0.85 (0.67, 1.08)	1.51 (1.08, 2.12)
Q5 (Highest SES)	Ref	Ref	Ref	Ref	Ref	Ref	Ref	Ref	Ref

Derived from multinomial logistic regression models including age, sex, education, household income, employment status, and neighborhood SES.

Abbreviations: SCCS: Southern Community Cohort Study; SES: Socioeconomic Status; NN: Normal-Normal; SS: Short-Short; LL: Long-Long; SL: Short-Long; LS: Long-Short; SN: Short-Normal; LN: Long-Normal; NS: Normal-Short; LS: Long-Short.

Among Black participants, also, lower education was associated with increased odds of four of six suboptimal sleep trajectories (LL, SL, LS, NS), but associated with increased odds of one optimal improving trajectory (LN), with the notable trajectories being SL (1.63 (1.28, 2.06)), and LS (1.68 (1.25, 2.24)) under the unstable unhealthy group of trajectories. For Black participants, however, lower household income was associated with higher odds of only three of six suboptimal sleep trajectories (SL, LS, NL) but had no association with any optimal improving trajectories (SN, LN): SL (1.81 (1.32, 2.47)), and NL (1.56 (1.20, 2.01)) under the unstable unhealthy and deteriorating group of trajectories, respectively. Also, among Black participants, unemployed individuals consistently had higher odds of experiencing all six suboptimal sleep trajectories (SS, LL, SL, LS, NS, NL) and one improving trajectory (LN) compared with their employed counterparts: LL under stable unhealthy (2.32 (2.01, 2.67)) and LN under improving (1.39 (1.24, 1.56)) trajectories. Lower neighborhood SES was associated with higher odds of only two of six suboptimal sleep trajectories (LL, NL) but had no association with any improving trajectories (SN, LN) among Black individuals: LL (1.42 (1.01, 1.99)) and NL (1.65 (1.23, 2.21)) under stable unhealthy and deteriorating groups of trajectories, respectively.

For both Black and White participants, the associations of educational attainment and employment status with long-term sleep trajectories were similar, although educational attainment showed relatively smaller effect estimates or weaker statistical significance among Black than White individuals. Also, for both racial groups, unemployment status showed a stronger association with several suboptimal trajectories that involved long sleep duration at baseline or follow-up (LL, LN, NL) than with trajectories involving short sleep alone (SS, SN, NS). However, there appeared to be some notable differences in the associations between household income, neighborhood SES, and sleep trajectories among Black and White participants.

Discussion

In this study, we examined the association of baseline ­individual- and neighborhood-level factors with individual trajectories of sleep duration over ~10 years in a large prospective cohort of ­middle-to-older aged Black and White population in the Southeastern United States. The study found that lower SES at the individual- and neighborhood-levels were associated with increased odds of various suboptimal long-term sleep trajectories among both Black and White participants. There were, however, some nominal racial differences, which generally appeared stronger among White than Black individuals.

Generally, lower educational attainment appeared to be associated with higher odds of experiencing multiple suboptimal long-term individual sleep trajectories among Black and White participants. Several cross-sectional studies have documented an association between short and long sleep duration patterns and lower educational attainment [14, 27, 28]. In a multi-year trend analysis in the American Time Use Survey, Basner and Dinges [29] reported that a decrease in sleep duration was associated with increasing educational attainment. Meanwhile, there has been limited longitudinal research on the link between educational attainment and long-term sleep patterns. For instance, using the Alameda County Health and Ways of Living Study, Stamatakis et al [30]. have linked lower educational attainment to increased short sleep over time among all racial groups. However, unlike the current study, the study did not examine various long-term sleep trajectories.

We found that lower educational attainment was associated with various suboptimal long-term sleep trajectories indicating unstable, unhealthy, and deteriorating long-term sleep patterns for both races. Overall, the observed patterns of the associations between educational attainment and long-term sleep trajectories were similar for both racial groups, even though the effect estimates appeared stronger among White than Black individuals. In a similar vein, both some White and Black individuals with lower education appeared to have experienced improving long-term sleep trajectories. Conversely, some previous studies also reported racial differences in the association between education and sleep duration. For instance, in a cross-sectional analysis, Luo et al [31]. observed opposite educational gradients in short sleep duration among Black and White populations, with higher educational attainment being associated with reduced short sleep among White individuals while associated with increased short sleep duration among Black individuals. Similarly, in a previous study, education has been found to show different associations with sleep among Black and White individuals [32]. These findings suggested that while lower education was a predictor of long-term short sleep trajectories among White individuals, it may be more closely associated with long long-term sleep trajectories among Black individuals, which is an interesting finding that may potentially warrant further investigation. Given the lack of studies focusing on long-term sleep trajectories, our study serves as a foundation for further research on the prospective link between education and long-term sleep trajectories. Future studies can, therefore, take advantage of prospective cohorts with repeated sleep measurements to better understand the relationship between educational attainment and long-term individual sleep trajectories among diverse subpopulations.

Besides education, household income is another commonly used indicator of individual-level SES. Several previous studies have documented a strong association between low household income and unhealthy sleep behaviors such as short sleep [30], sleep quality problems [33–36], and short and long sleep duration [14, 28]. There has also been evidence of the increased prevalence of short sleep duration over time in lower household income quintiles (versus higher quintiles) [30]. In our analysis, lower household income emerged as a consistent predictor of higher odds of all the various suboptimal long-term sleep trajectories showing unstable, unhealthy, and deteriorating patterns among White individuals. On the other hand, the associations appeared considerably constrained among Black individuals, suggesting some notable racial differences in the association between low household income and suboptimal long-term sleep trajectories. Furthermore, our findings suggest that while some White individuals with lower household incomes may more likely improve their long-term sleep patterns, there is no such evidence for Black individuals with lower household income. The reason for the difference is unclear but this may likely stem from unobserved socio-cultural variations within both racial groups. A previous study examining Black-White differences in sleep in different housing types reported that in suboptimal housing conditions (e.g. Mobil home/trailer), Black women were less likely to report sleep problems when compared to their White counterparts [37]. This finding, together with ours, suggests that the link between low-income status and poor sleep may be stronger among White than Black populations. However, it is also noteworthy that, among both White and Black individuals, the strongest associations between household income and suboptimal long-term sleep trajectories were observed for the unstable unhealthy (SL and LS) trajectories. This group of sleep trajectories is characterized by not only unhealthy sleep duration in both baseline and follow-up but also drastic changes in long-term sleep durations. Adequate evidence also suggests that short and long sleep patterns are major risk factors for various adverse health outcomes [38]. Our findings suggest that low household income individuals may be most susceptible to the detrimental health effects associated with unstable long-term sleep trajectories, mainly among White than Black individuals in this unique study population. This directly contrasts the increased odds of short sleep duration associated with higher income levels observed among Black individuals compared with White individuals in a nationally representative cohort of middle-aged to older Black and White adults [39].

Furthermore, previous research has linked sleep concerns to employment status [34, 40–42]. In this study, unemployment status was associated with a higher likelihood of various suboptimal long-term sleep trajectories for both Black and White individuals. Our findings on all the long-term sleep trajectories further showed that the association between employment status and suboptimal sleep trajectories was stronger for sleep trajectories that involved long sleep, either at baseline or follow-up, among unemployed people. This echoes the findings of a previous study based on the American Time Use Survey, which observed that unemployed individuals slept longer over time, albeit Black individuals also slept longer than White individuals over the study period [29]. Additionally, an increase in both short and long sleep duration patterns has been linked to unemployment in other studies [28, 40]. In our study, we further found that unemployment was associated with consistent short sleep (SS) and unstable unhealthy group (LS and SL) of sleep trajectories in both Black and White individuals, suggesting a strong link between unemployment and long-term short sleep patterns among both racial groups. There can be several mechanisms through which unemployment may lead to unhealthy sleep patterns, including a lack of daily routine, stress, and anxiety, all of which may have a deleterious impact on long-term sleep patterns. However, research on racial differences in the link between employment and long-term sleep trajectories remains limited and, therefore, needs further investigation. Meanwhile, it has been reported that Black individuals spend considerably less time in bed under unemployment conditions than White individuals [41]. While inverse short sleep gradients have been found between Black and White individuals regarding occupational roles in a nationally representative study [43], our overall findings show an important role of employment in supporting similar suboptimal long-term sleep patterns among both racial groups.

The neighborhood environment has been widely recognized as an important determinant of health. In our analysis, we found racial differences in long-term sleep patterns. Lower neighborhood SES was associated with increased odds of suboptimal long-term sleep trajectories – indicating unstable unhealthy or deteriorating patterns – among White individuals with evidence that some residents of lower SES neighborhoods also experienced optimal improving long-term sleep patterns. Black individuals residing in lower SES neighborhoods, on the other hand, were more likely to experience only suboptimal long-term sleep trajectories suggesting stable unhealthy or deteriorating long-term sleep patterns without any evidence of optimal improving long-term sleep patterns. It is noteworthy that the association between neighborhood SES and long-term sleep trajectories was observed after individual-level SES factors were included in the analysis, suggesting that neighborhood effects on long-term sleep trajectories may be independent of individual-level factors among both racial groups – although with diverse patterns – in this socioeconomically unique study sample. Notwithstanding, our findings indicate neighborhood SES had a lesser effect on the long-term sleep trajectories than the individual-level SES factors, for both racial groups. A growing number of studies have linked neighborhood SES to sleep duration, with these studies showing that residents of socioeconomically disadvantaged neighborhoods were more likely to report short sleep than their counterparts in higher SES neighborhoods [20, 44–47]. However, these studies have not examined racial differences in sleep duration by neighborhood socioeconomic characteristics, although racial differences in waking after sleep onset were found to be partly due to neighborhood socioeconomic characteristics [16].

Long-term sleep patterns have been associated with various adverse health outcomes in human populations. For instance, numerous studies have linked lower or unstable sleep duration, and both decrease and increase in sleep duration to a higher risk of cardiovascular diseases, all-cause mortality [7, 8, 48], and cancer development [12]. In addition, consistent short and prolonged sleep patterns were associated with higher risks of specific diseases including hypertension [9], type 2 diabetes [10], and stroke [49]. With the paucity of research on the effects of individual- and neighborhood-level SES on long-term individual sleep trajectories, our study underscored the need for further investigation to provide a better understanding of the socioeconomic contexts underpinning various suboptimal long-term sleep trajectories – stable unhealthy, unstable unhealthy, and deteriorating patterns – and the racial differences in neighborhood effects to improve overall sleep health.

Our study has some important strengths. First, we conducted prospective analyses focusing on long-term sleep patterns over ten years, which has been rarely examined in previous studies. Moreover, the study also drew on a unique study population made up of both Black and White Americans to provide race-specific analysis. Nevertheless, it is noteworthy that the current study is not without some limitations. The outcome of the study is self-reported sleep duration, which is retrospective and based on non-validated questions. In addition, we used pre-defined short, normal, and long sleep categories to classify long-term sleep trajectories. Although this method is based on widely accepted cut-off points with clear clinical and public health indications, it cannot capture more subtle changes within each group of sleep duration. Moreover, small fluctuations in sleep duration around cut-off points may be artificially magnified because they may lead to changes in sleep categories. Such fluctuations are likely non-differential, which may lead to attenuated results. As well, the SCCS did not collect information on other important aspects of sleep including quality and timing which may help to better understand sleep health. The study sample did not collect information on retirement, and we could not distinguish between individuals who were unemployed from those who were retired, which may constitute a substantial portion of our study population due to their more advanced age. We also lacked information on sleep disorders, psychiatric conditions, and related medical treatment, which may affect sleep patterns. The study used baseline socioeconomic characteristics to predict long-term sleep patterns. As such, our measure of neighborhood SES did not reflect changes in the neighborhood that may occur after baseline, which may also have an impact on sleep patterns. The study did not also account for the time lived by the individuals in the neighborhoods. Another important limitation is that the study sample is not a random sample of the population – predominantly comprised of Black, and low SES individuals – and thus may not be generalizable to the population assessed.

In conclusion, our findings suggest that multiple ­individual- and neighborhood-level socioeconomic characteristics were linked to various suboptimal long-term individual sleep trajectories among Black and White adults, albeit some of the associations may vary between racial groups. The findings indicate that individuals with lower SES at both the individual and neighborhood levels were more likely to show unhealthy or deteriorating sleep trajectories over time, which may contribute to substantial health disparities in the Southeastern United States. Some previous studies have established an increase in suboptimal sleep durations – especially among racial minorities – in the United States over time [17, 18]. Notably, two earlier analyses focusing on geographic disparities in sleep deficiencies also reported that Southern regions of the country had a higher prevalence of insufficient sleep and daytime fatigue when compared to other regions [50, 51]. Such area-level differences may be explained by differences in both population composition and neighborhood attributes across the country. Future studies should further examine racial disparities in the link between multilevel socioeconomic characteristics and long-term sleep trajectories and whether improvement in individual and neighborhood level SES can help to reduce ­sleep-related health disparities over time.

Funding

Qian Xiao received support from the National Institute on Aging (R01AG063946). Liying Luo received support from the National Institute on Aging (R03AG070812, R01AG078518), the Population Research Institute (NICHD P2CHD041025), and the Pennsylvania State University. This content is solely the responsibility of the authors and does not necessarily represent the official views of the funders.

Disclosure Statement

Financial disclosure: none.

Non-financial disclosure: none.

Data Availability

The data analyzed in the current study are available from https://ors.southerncommunitystudy.org upon request.