Effect of daily discrimination on naturalistic sleep health features in young adults

Abstract

Objective.

Racial inequities in sleep health are well documented and may be partially attributable to discrimination experiences. However, the effects of acute discrimination experiences on same-night sleep health are understudied. We quantified naturalistic discrimination experiences captured using ecological momentary assessment (EMA) and examined whether reporting discrimination on a given day predicted sleep health that night.

Methods.

Participants completed baseline assessments and a 17-day EMA protocol, with text prompts delivered four times daily to collect discrimination experiences. Seven different daily sleep characteristics were ascertained each morning. Discrimination reasons (e.g., because of my racial identity) were reported by participants and categorized into Any, Racial, or Non-racial discrimination. Outcomes included the seven sleep diary characteristics. We fit generalized linear mixed effects models for each sleep outcome and discrimination category, controlling for key covariates.

Results.

The analytic sample included 116 self-identified Black and White individuals (48% Black, 71% assigned female at birth, average age = 24.5 years). Among Black participants, race-based discrimination was associated with a 0.5-hour reduction in total sleep time. Among White individuals, Non-racial discrimination was associated with a 0.6-hour reduction in total sleep time, an earlier sleep offset, and reduced sleep efficiency (partly attributable to more nighttime awakenings).

Conclusions.

Young adults may sleep worse on nights after experiencing discrimination, and different types of discrimination affect different sleep outcomes for Black and White individuals. Future studies may consider developing treatments that account for different sleep vulnerabilities for people experiencing discrimination on a given day.

Introduction

Poor sleep health (e.g., short sleep duration, poor sleep continuity, irregular sleep timing) is associated with many negative physical and mental health outcomes including cardiometabolic disease (Cappuccio et al., 2010a; St-Onge et al., 2016), mood disorders (Rumble et al., 2015), susceptibility to infectious illness (Prather et al., 2015), inflammation (Irwin et al., 2016), substance use (Boness et al., 2022; Graupensperger et al., 2022; Hasler et al., 2015, 2022; Logan et al., 2018), and premature mortality (Cappuccio et al., 2010b; Wallace, Buysse, et al., 2019; Wallace et al., 2022; Wallace, Lee, et al., 2019). Moreover, inequities in sleep health exist for a variety of communities and individuals. For example, racial inequities in sleep health are well documented, and studies have shown that Black individuals (and individuals of other minoritized racial and ethnic identities) have shorter sleep duration, poorer sleep efficiency, and greater daytime sleepiness compared to White individuals (Carnethon et al., 2016; Jackson, Powell-Wiley, et al., 2020; Johnson et al., 2019; Laposky et al., 2016; Petrov & Lichstein, 2016; Yip et al., 2021). While less studied, a similar pattern has been found for LGBTQIA+ individuals. Men with minoritized sexual orientation identities experience worse sleep than heterosexual individuals (Gibbs & Fusco, 2023); adolescents with minoritized gender identities are more likely to report inadequate sleep, shorter sleep duration, and poorer sleep quality than heterosexual cisgender adolescents (Levenson et al., 2021); and transgender adults who experience discrimination are more likely to have sleep problems that include poor sleep quality, short sleep duration, and use of alcohol or sleep medications to fall asleep (Eom et al., 2022).

Although the determinants of inequities in sleep health are not well understood, they involve a multi-level etiology that includes social, environmental, behavioral, and biological factors (Ahn et al., 2021; Billings et al., 2021; Grandner & Fernandez, 2021). Interpersonal discrimination (e.g., being treated with less respect than others) is one potential driver of inequities in sleep health (Grandner et al., 2016; Jackson, Powell-Wiley, et al., 2020; Jean-Louis et al., 2022; Johnson et al., 2019; Petrov & Lichstein, 2016). These discrimination experiences may be attributable to any number of reasons (e.g., age, disability, sexual orientation, physical appearance, religion) and subsequently contribute to poor sleep health by impacting sleep through multiple possible mechanisms including physiological arousal, psychological distress, and/or hyperactivation of the sympathetic nervous system (Slopen et al., 2016).

In studies specifically examining racial and ethnic identities, the effects of discrimination on sleep health are mixed and vary across subjective or objective measurements of sleep. For instance, among cross-sectional and non-EMA longitudinal studies examining retrospective subjective sleep, discrimination is associated with more complaints of sleep difficulties and insomnia (Lewis et al., 2013; Slopen et al., 2016), shorter sleep duration (Hill et al., 2021; Huynh & Gillen-O’Neel, 2016; Johnson et al., 2021; Slopen & Williams, 2014), and poorer sleep quality (Gordon et al., 2020; Hill et al., 2021; Johnson et al., 2021; Zeiders, 2017). Further, racism-related vigilance, the preparation and/or anticipation for racial discrimination, has also been shown to be associated with poor sleep quality (Gordon et al., 2020). Additionally, research has shown that discrimination mediates the association between racial or ethnic identities and certain sleep measures including insomnia (Cheng et al., 2020), a latent variable for sleep problems (assessed from reported latency, duration, efficiency, and quality) (Fuller-Rowell et al., 2017), and sleep architecture (Cheng et al., 2020; Fuller-Rowell et al., 2017; Tomfohr et al., 2012). Among studies assessing the effect of discrimination on objectively measured sleep characteristics, results are mixed (Beatty et al., 2011; Lewis et al., 2013; Slopen et al., 2016; Tomfohr et al., 2012; Yip et al., 2020). Some studies report that discrimination is associated with shorter sleep duration, greater wake after sleep onset, a smaller proportion of REM sleep, a smaller proportion of slow wave sleep, a larger proportion of light sleep, lower sleep efficiency, and lower sleep onset latency, while others report null associations between discrimination and these characteristics. Importantly, self-report of sleep health may be underestimating the magnitude of racial inequities relative to objective measurement (Jackson et al., 2018; Jackson, Ward, et al., 2020).

One plausible reason for the abovementioned mixed findings is that many studies have primarily relied on cross-sectional observational studies that examine between-person effects (e.g., people who experience more discrimination on average also report worse average sleep health) and rely on global retrospective measures of discrimination, rather than examining acutely occurring experiences of discrimination within-person. Examination of acute instances of discrimination (e.g., on a given day) is needed to reduce recall bias and understand the within-person effect that discrimination experiences during the day may have on a person’s sleep health that night. Using a daily diary or an ecological momentary assessment (EMA) study design may partially address some limitations in cross-sectional or non-EMA longitudinal designs by assessing more acute and naturalistic effects of discrimination on sleep health within-person. One 9-day daily diary study of 124 African American college students found that on days when participants experienced more discrimination, subsequent sleep problems (trouble falling asleep, feeling poorly rested) increased; this association was stronger for individuals with higher levels of internalized racism (Fuller-Rowell et al., 2021). However, this study remains limited in its sample (African American students attending predominantly White universities), and responses to questions related to both discrimination and sleep may still be subject to 24-hour recall bias as both measures were assessed via evening diaries.

Notably, previous research is limited in other ways as well. Sleep is a complex construct with multiple interrelated dimensions that differentially relate to various health outcomes. Thus, it is important to consider sleep health from a multidimensional perspective which ideally includes measures of sleep regularity, satisfaction, alertness, timing, and duration (Buysse, 2014). However, studies have primarily only focused on the effect of discrimination on sleep duration, sleep quality, or complaints of sleep difficulties, with relatively few studies examining its effect on sleep continuity (i.e., sleep efficiency, sleep onset latency, and wake after sleep onset) or sleep timing (i.e., sleep midpoint, onset, and offset). Additionally, most studies focus on non-specific experiences of discrimination or on only one attribution of discrimination (e.g., an individual’s racial identity) as opposed to including experiences that may occur for numerous attributions (e.g., age-, sexual orientation-, gender identity-, weight-based discrimination, etc.). Additionally, samples typically include middle aged participants (Hill et al., 2021; Johnson et al., 2021; Lewis et al., 2013) or adolescents (Huynh & Gillen-O’Neel, 2016; Zeiders, 2017). Only two studies (Gordon et al., 2020; Slopen & Williams, 2014) included young adult participants – a population in which a number of adverse outcomes including obesity, mental health issues, and pain have been shown to be associated with poor sleep, and for which early interventions could help prevent and/or manage longer-term consequences of poor sleep (Bruce et al., 2017).

While prior research demonstrates the between-person associations between discrimination and sleep health, further examination of how acute experiences of discrimination affect subsequent sleep health is needed. Understanding the dynamic unfolding of which components of sleep health are affected by naturalistic experiences of discrimination can help identify proximal treatment targets to reduce inequities in sleep health. The current project examines discrimination experiences attributed to 12 individual attributes (e.g., race, education/income, height, sexual orientation, religion) and seven sleep health characteristics (e.g., total sleep time, sleep midpoint, sleep efficiency) across a 17-day assessment window to extend this literature. Based on scientific literature and clinical expertise, we hypothesized that experiencing discrimination on a given day would predict reduced sleep efficiency, shorter sleep duration (total sleep time), and later sleep midpoint that night. Additional features (i.e., sleep onset, sleep offset, sleep onset latency, and minutes awake after sleep onset) were considered secondary to further probe the nuanced components of sleep midpoint and sleep efficiency. We examined racial discrimination separately from discrimination attributed to other characteristics within our sample of participants identifying as Black or African American given the frequency of these experiences and existing literature showing the deleterious effects of racial discrimination.

Method

Participants

The present study is part of a larger ongoing alcohol administration research project focused on examining alcohol response in Black and White young adults. Details of participant recruitment are reported elsewhere (Hunter et al., 2022). Briefly, participants were recruited from an urban community in the greater Pittsburgh, Pennsylvania area of the United States and completed a phone screen to determine initial eligibility. Participants were required to report drinking alcohol at least weekly, consume at least 4/5+ drinks (female/male participants) on at least three occasions in the past 3 months (^~1/month), and be 21-30 years of age. Given the alcohol administration protocol component of the study, we excluded individuals based on a number of criteria contraindicated with alcohol use (e.g., currently receiving treatment for substance use disorder, diagnosis of psychiatric disorder, currently being pregnant or breastfeeding) (Hunter et al., 2022). Individuals not identifying as White/European American or Black/African American race or those who identified as Hispanic were excluded to allow for sufficient power to focus on the primary aims of the study. During enrollment, participants were matched across racial identity on self-reported past 30-day drinking behavior (quantity and frequency) from the screener to reduce differences in acute alcohol tolerance that could affect alcohol response.

One hundred and sixty-four participants were enrolled in the study and completed the baseline survey. Based on the focus of our current study, we included only those participants who recorded at least one response to the discrimination survey during the EMA period, and who completed at least one sleep diary assessment. One hundred and sixteen (70.7%) participants met these criteria and were included in our analytic sample. Differences in demographic characteristics between participants included in versus excluded from our analytic sample were negligible (Cohen’s d or Cramer’s V between 0.06-0.278; Supplementary Table 1).

Procedures

Study procedures were approved by the University of Pittsburgh Institutional Review Board and are detailed elsewhere (Hunter et al., 2022). Briefly, all participants completed an interview session to ascertain DSM-5 diagnoses (American Psychiatric Association, 2013), complete a series of questionnaires, and confirm eligibility. Participants then completed a within-subjects alcohol administration protocol (two beverage sessions with either moderate dose of alcohol or placebo). Participants were compensated with $50.00 for interview completion, $100.00 for alcohol session completion, and $100.00 for placebo session completion.

Following the second beverage session, participants were trained in the 17-day EMA protocol that started the following Friday and spanned 3 weekends. Participants received four prompts daily (15 minutes after typical wake time, 15 minutes before typical bedtime, early afternoon, and early evening) via text message with a direct link to a password-protected web-based questionnaire. On average, participants completed the questionnaire within 3.1 minutes from the time the prompt was received. Participants were also asked to self-initiate a survey (pre- and post-drink) immediately prior to consuming alcohol and following completion of each alcoholic beverage.

The present study utilizes data from the four daily prompts to assess acute discrimination experiences and daily marijuana use. A combination of the daily prompts and the self-initiated pre- and post-drink surveys were used to determine quantity of daily alcohol consumption. Participants could earn up to $10.00/day if they completed three or more of the daily prompts. Participants completing over 85% of prompts across the 17-day period earned a $55.00 bonus payment to incentivize high response rates. The overall compliance rate of the daily prompts was 82.4% across the EMA protocol.

Measures

Sleep Outcomes.

Sleep health was measured using a daily morning sleep diary based on a modified version of the Pittsburgh Sleep Diary (Monk et al., 1994). Primary outcomes and secondary outcomes (to further probe the effect of discrimination on sleep timing and continuity) included the following measures.

Total Sleep Time (Primary).

Computed as the time in hours between sleep onset time and sleep offset time minus minutes awake after sleep onset.

Sleep Midpoint (Primary).

Computed as the midpoint of sleep onset and sleep offset.

Sleep Onset Time (Secondary).

The fall-asleep time. Calculated as the reported time trying to fall asleep (i.e., lights out time) plus the reported time to fall asleep (i.e., sleep onset latency).

Sleep Offset Time (Secondary).

The participanťs reported wake-up time.

Sleep Efficiency (Primary).

Computed as total sleep time / time between sleep onset and the time out of bed * 100. The analytical value for sleep efficiency was log transformed as log(100 - sleep efficiency +1) to follow a relatively more normal distribution.

Sleep Onset Latency (SOL; Secondary).

The minutes to fall asleep. The analytic value for SOL was square root transformed to follow a relatively more normal distribution.

Minutes Awake after Sleep Onset (WASO; Secondary).

The total amount of nighttime awakenings in minutes.

We reported intraclass correlation coefficients (ICCs) of the sleep outcomes in Supplementary Table 2. The ICCs summarize the extent to which the within-person repeated measures are correlated with one another and range from 0.248 to 0.723 for Black participants and from 0.208 to 0.612 for White participants across all seven outcomes.

Discrimination.

Our main predictors of interest were based on daily EMA measures of discrimination. At each EMA prompt, participants were asked to respond on a 5-point Likert scale (not at all, slightly, somewhat, moderately, extremely) to six questions – five from a modified version of the Everyday Discrimination Scale (Short Version) [EDS] (Williams et al., 1997), one asking whether “you were discriminated against” to ensure capture of experiences beyond the specific types of discrimination queried in the EDS. In a non-EMA setting, the EDS (short version) provides good internal consistency (Cronbach’s alpha = 0.77) (Sternthal et al., 2011). The EDS has also been used to evaluate reliability and validity of similar scales in diverse populations (Gonzales et al., 2016; Krieger et al., 2005; Taylor et al., 2004) and has been employed in EMA studies of health outcomes such as substance use and mood (Livingston et al., 2017; Newberger et al., 2022). “Any discrimination” on a given day was indicated if a participant responded with something other than “not at all” to at least one question during the day. “Racial discrimination” was indicated if a participant responded with something other than “not at all” to at least one question and reported that the “main reason for the experience” was related to “race” or “shade of skin color.” “Non-racial discrimination” was indicated if a participant responded with something other than “not at all” to at least one question and reported that the main reason for the experience was related to something other than “race” or “shade of skin color” (i.e., gender identity, sexual orientation, religion, age, physical disability, height, weight, other aspect of physical appearance, education or income level, ancestry or national origins). Thus, in analytic models, we coded each discrimination measure (Any, Racial, or Non-racial) as 0 (no discrimination experienced on a given day) versus 1 (discrimination experienced on a given day).

Covariates.

Covariates selected a priori included time-varying EMA substance use, a COVID and a weekend indicator, and baseline demographic measures.

Alcohol Use.

Number of alcoholic beverages consumed on a given day was computed by utilizing data across the 4 daily EMA prompts as well as self-initiated surveys completed after every alcoholic beverage was consumed. Quantity of alcohol consumed was included as a covariate given known effects on sleep (Koob & Colrain, 2020).

Marijuana Use.

At each of the 4 EMA prompts, participants indicated whether they had used marijuana since the last assessment. We computed whether or not marijuana was used on a given day and included this variable as a covariate given cannabis’s known effects on sleep.

Baseline Demographic Measures.

Self-identified Black or White racial identity, queried at the initial visit, was also a key predictor of interest. Time-invariant covariates measured at the initial visit included participant age, past year household income, and assigned sex at birth (male or female).

Non-EMA Time-variant Measures.

Time-variant covariates included a COVID-19 indicator denoting the onset of COVID-19 lockdowns beginning in March 2020 and an indicator denoting a weekend day (Saturday or Sunday) given their known impact on sleep.

Analytic Plan

In preliminary analyses, we summarized the sample by racial identity and assessed distributions of key covariates (age, household income, assigned sex at birth, alcohol use, and marijuana use), discrimination indicators, and averaged daily measures of sleep. We further examined frequencies and percentages of the various reasons for discrimination.

For our primary aims, we fit generalized linear mixed effects models for all continuous sleep outcomes except WASO. For WASO, we fit a zero-inflated negative binomial mixed effects model to account for the zero-inflated nature of its distribution. Since Black participants predominantly reported racial discrimination, and the percentage of White participants reporting racial discrimination was small, discrimination indicators that include race-specific types of discrimination (Any discrimination and Racial discrimination) were analyzed for Black participants only. Therefore, for each sleep outcome, our primary models examined the effect of: (1) Any discrimination in Black participants only, (2) Racial discrimination in Black participants only, and (3) Non-racial discrimination among all participants. In secondary models, we also examined the effect of Non-racial discrimination among Black and White participants separately. For the model including Non-racial discrimination in all participants, we considered racial identity and a racial identity by discrimination interaction term; however, if the interaction was not statistically significant at alpha=0.05, we excluded it from the final model. All models also included baseline demographic measures, non-EMA time-variant measures, EMA marijuana use, and EMA alcohol use. We included a random effect to account for within-subject correlations and specified an auto-regressive 1 correlation matrix to account for observed auto-correlation. We did not estimate a random slope coefficient due to the observational nature of the study. We report detailed model specifications in Supplementary Materials.

In sensitivity analyses, we added to each model a sum of the number of days in which the specific type of discrimination was reported over the 17-day EMA period (i.e., in an analysis of Any discrimination as the main predictor, we additionally included the person-specific sum of EMA days where Any discrimination was reported). This was used to determine whether within-person associations of discrimination and sleep remained present above and beyond the effect of more chronic between-person levels of discrimination.

Reports of statistical significance were based on a p-value threshold of 0.05. To control for multiple comparisons across our seven sleep outcomes, we also examined whether discrimination measures met a Bonferroni-adjusted p-value threshold (p < 0.05/7 outcomes = 0.007). Missing outcome data were assumed to be missing completely at random or missing as a result of observed covariates in the model; this type of missingness is accommodated using mixed effects models. Analyses were performed in R; mixed effects models with continuous outcomes used the lme function from the nlme package (J. Pinheiro et al., 2022; J. C. Pinheiro & Bates, 2000), and the zero-inflated model used the mixed_model function from the GLMMadaptive package (Rizopoulos, 2022).

Results

Participant Characteristics

Overall, 60% of the 116 participants included in the analytic sample reported data for at least one discrimination measure and one sleep measure on all 17 days of the EMA protocol, including 63% of White and 57% of Black participants; another 30% reported these data on 13-16 days, and 10% reported data on 12 days or fewer. Thus, among all recorded observations (i.e., person-days) in the analytic sample, no discrimination data were missing, 4.3% of person-days were missing sleep efficiency (4.7% Black, 3.8% White), and 3.5% of person-days were missing all other sleep characteristics (3.7% Black, 3.2% White).

Descriptive statistics of study participants are presented in Table 1 by racial identity. The analytic sample was 48% Black and 71% assigned female at birth with an average age of 24.5 years, 35% fell into the highest income category ($100k-$149k), and 49% reported marijuana use during the EMA period. Black participants reported slightly lower average total sleep time [mean(SD) = 7.56(1.13) hours for Black versus 7.69(0.83) hours for White participants], later sleep midpoint [Black = 4.46(1.12); White = 4.23(0.97)], and lower sleep efficiency [Black = 88.1(6.25); White = 91.2(4.29)] over the EMA period compared to White participants (Table 1). Effect sizes comparing Black versus White participants were negligible or small on demographic and sleep characteristics except sleep efficiency (Cohen’s d = −0.584).

Table 1.

Descriptive characteristics of study participants^a

	Overall (N=116)	Black (N=56)	White (N=60)	Effect Sizeb
Age, mean(SD)	24.5 (3.02)	25.0 (3.06)	24.1 (2.94)	0.305
Race, n(%)
Black	56 (48.3%)	56 (100%)	0 (0%)
White	60 (51.7%)	0 (0%)	60 (100%)
Household income, n(%)				0.253
under $10,000	21 (18.1%)	10 (17.9%)	11 (18.3%)
$10,000-$19,999	4 (3.4%)	0 (0%)	4 (6.7%)
$20,000-$34,999	19 (16.4%)	13 (23.2%)	6 (10.0%)
$35,000-$49,999	14 (12.1%)	7 (12.5%)	7 (11.7%)
$50,000-$74,999	14 (12.1%)	7 (12.5%)	7 (11.7%)
$75,000-$99,999	3 (2.6%)	1 (1.8%)	2 (3.3%)
$100,000-149,999	41 (35.3%)	18 (32.1%)	23 (38.3%)
Assigned sex, n(%)				0.06
Female	82 (70.7%)	38 (67.9%)	44 (73.3%)
Male	34 (29.3%)	18 (32.1%)	16 (26.7%)
Any discrimination, n(%)				0.272
No	31 (26.7%)	8 (14.3%)	23 (38.3%)
Yes	85 (73.3%)	48 (85.7%)	37 (61.7%)
Racial discrimination, n(%)				0.58
No	73 (62.9%)	19 (33.9%)	54 (90.0%)
Yes	43 (37.1%)	37 (66.1%)	6 (10.0%)
Non-racial discrimination, n(%)				0.115
No	44 (37.9%)	18 (32.1%)	26 (43.3%)
Yes	72 (62.1%)	38 (67.9%)	34 (56.7%)
Average total sleep time, mean(SD)	7.63 (0.982)	7.56 (1.13)	7.69 (0.829)	−0.136
Average sleep midpoint, mean(SD)	4.34 (1.05)	4.46 (1.12)	4.23 (0.972)	0.219
Average sleep efficiency, mean(SD)	89.7 (5.53)	88.1 (6.25)	91.2 (4.29)	−0.584
Average daily # drinks, mean(SD)	2.04 (1.85)	2.04 (1.84)	2.05 (1.87)	−0.004
Used marijuana, n(%)				0.051
No	59 (50.9%)	27 (48.2%)	32 (53.3%)	0.051
Yes	57 (49.1%)	29 (51.8%)	28 (46.7%)

^aNo data were missing among demographic characteristics nor among person-level summaries of EMA data reported in this table.

^bEffect size compares Black versus White participants using Cohen’s d for continuous variables and Cramer’s V for categorical variables.

Discrimination Measures

Overall, 73% of participants reported at least one ‘Any discrimination’ event during the 17-day EMA period. Among Black participants, 66% reported at least one ‘Racial discrimination’ event (i.e., experiencing either race-based or skin color-based discrimination), representing a medium effect size difference (Cramer’s V = 0.58) compared to White participants experiencing Racial discrimination (10%). 68% of Black and 57% of White participants reported at least one ‘Non-racial discrimination’ event (Table 1, Figure 1). Nearly one-third of Black participants also experienced education/income, age, and physical appearance-based discrimination (Figure 1). About 20% of White participants experienced education/income, age, physical appearance, or gender-based discrimination during the EMA period. Finally, a higher proportion of Black participants experienced each discrimination reason compared to White participants, except in the case of discrimination related to gender identity or physical disability where the proportion of White participants exceeded that of Black participants by 2 percentage points (Figure 1).

Figure 1.

Percentage of participants (top) reporting at least one discrimination experience during the EMA period, overall and by racial identity. The bottom portion of the figure identifies the specific discrimination experiences that constitute Racial discrimination, Non-racial discrimination, and Any discrimination.

Sleep Duration (Total Sleep Time)

For Black participants, reporting Racial discrimination during the day was associated with an estimated 0.51 (95% CI: −0.85, −0.17) fewer hours of TST that night (Figure 2, Supplementary Table 2). For all participants, we observed a statistically significant interaction between racial identity and Non-racial discrimination; this was driven by the effect of Non-racial discrimination in White participants, indicating a 0.64 (95% CI: −0.99, −0.29) hour decrease in TST on days when they experienced Non-racial discrimination. In contrast, Non-racial discrimination was associated with a 0.09 (95% CI: −0.26, 0.44) hour increase in TST for Black participants. Experiencing Any discrimination was not significantly associated with TST.

Figure 2.

Effect size estimates and 95% confidence intervals for the effect of discrimination (Any, Racial or Non-racial) within each sample (Black only or White only) on total sleep time, controlling for age, assigned sex, COVID-19, total alcoholic drinks, income, marijuana use, and weekend day.

Sleep Timing (Sleep Midpoint, Onset, Offset)

For sleep offset in all participants, there was no statistically significant interaction between racial identity and Non-racial discrimination. However, among White participants, experiencing Non-racial discrimination was associated with a 0.38 hour (95% CI: −0.67, −0.08) earlier sleep offset; while significant at alpha=0.05, this relationship did not remain significant after adjusting for multiple comparisons (Figure 3C, Supplementary Table 2). Comparatively, Non-racial discrimination was associated with a 0.17 hour (95% CI: −0.44, 0.10) earlier offset in Black participants (Figure 3C, Supplementary Table 2). We observed no statistically significant associations between sleep midpoint, sleep onset, or sleep offset and Any or Racial discrimination. Additionally, neither sleep midpoint nor sleep onset was associated with Non-racial discrimination (Figure 3, Supplementary Table 2).

Figure 3.

Effect size estimates and 95% confidence intervals for the effect of discrimination (Any, Racial, or Non-racial) within each sample (Black only or White only) on A) sleep onset, B) sleep midpoint, and C) sleep offset, controlling for age, assigned sex, COVID-19, total alcoholic drinks, income, marijuana use, and weekend day.

Sleep Continuity (Sleep Efficiency, SOL, WASO)

For sleep efficiency in all participants, we observed a statistically significant interaction between racial identity and Non-racial discrimination; this was driven by the effect of Non-racial discrimination in White participants, indicating a decrease in sleep efficiency (estimate for transformed sleep efficiency = 0.19; 95% CI: 0.03, 0.36) on days when they experienced Non-racial discrimination (Figure 4A, Supplementary Table 2). Moreover, for White participants who reported some nighttime awakenings, those who experienced Non-racial discrimination reported WASO 1.43 times (95% CI: 1.01, 2.02) the rate of those who did not experience Non-racial discrimination (Figure 4D, Supplementary Table 2). However, neither association between Non-racial discrimination and sleep efficiency nor WASO remained significant after adjusting for multiple comparisons. We observed no statistically significant associations between sleep efficiency, SOL, or WASO and Any or Racial discrimination. Further, Non-racial discrimination was not significantly associated with sleep onset latency nor the probability of zero WASO minutes (Figure 4, Supplementary Table 2).

Figure 4.

Effect size estimates and 95% confidence intervals for the effect of discrimination (Any, Racial, or Non-racial) within each sample (Black only or White only) on A) transformed sleep efficiency, B) transformed sleep onset latency, C) the odds for those reporting zero WASO minutes versus some minutes, and D) the rate ratio among those reporting some WASO minutes, controlling for age, assigned sex, COVID-19, total alcoholic drinks, income, marijuana use, and weekend day.

Sensitivity Analyses

For sensitivity analyses in which we added to each model the sum of the number of days in which the specific type of discrimination was reported, we observed results similar to those from the primary and secondary analyses (Figure 5, Supplementary Table 3). All statistically significant associations in the original models remained significant in the models containing the number of discrimination days, with nearly equivalent effect size estimates and 95% confidence intervals across model types (Figure 5).

Figure 5.

Comparison of effect sizes and 95% confidence intervals for statistically significant associations in the original model (circles) to those in the sensitivity model with the addition of the number of discrimination days (triangles). Y-axis denotes the sleep outcome [TST = total sleep time, SOF = sleep offset, and SE = sleep efficiency in A); and WASO = wake after sleep onset in B)], discrimination type (Racial or Non-racial), and racial identity sample (Black or White participants) specified by the model.

Discussion

In a study of 116 young adults, we examined differences in several types of naturalistic discrimination experiences of Black and White individuals and quantified within-person relationships between discrimination experiences and sleep duration, sleep timing, and sleep continuity using 17 consecutive days of EMA data collection. We found that Black individuals primarily experienced race-based discrimination and that experiences of Racial discrimination (the combination of race-and/or skin color-based discrimination) during the day was associated with a clinically-significant reduction (^~30 minutes) in total sleep time that night. White individuals experienced discrimination primarily based on education/income, age, physical appearance, and gender identity. Additionally, among White individuals, we observed that Non-racial discrimination was associated with nearly a 40-minute reduction in total sleep time, earlier sleep offset, and reduced sleep efficiency. This finding was partly attributable to greater time awake after falling asleep at night. Furthermore, sensitivity analyses showed that within-person associations of discrimination and sleep remained present above and beyond the effect of between-person levels of discrimination.

Our study includes some findings that are consistent with prior literature, but it also extends the literature in several important ways. For example, our study corroborates prior evidence specifically on the association between experiences of Racial discrimination and reduced sleep duration among Black participants, which is consistent among prior studies examining self-report sleep (Hill et al., 2021; Huynh & Gillen-O’Neel, 2016; Johnson et al., 2021; Slopen & Williams, 2014). However, the present study uniquely adds to this literature by examining specific categories of discrimination that differentially relate to sleep health among both Black and White individuals. For example, among White participants, discrimination attributed to Non-racial characteristics and identities (e.g., gender) appeared to influence sleep duration as well. Most prior studies have not specifically examined different types of discrimination (i.e., other than racial discrimination) within both Black and White individuals. The fact that we observed associations between Non-racial discrimination and several key sleep measures (sleep offset, sleep efficiency, and WASO) in White individuals but not in Black individuals may point to differences in the mechanisms through which discrimination may impact sleep. For example, experiences of Racial discrimination may be more disruptive to sleep for Black young adults than other types of discrimination experiences.

For Black and White individuals, experiences of discrimination attributed to Non-racial characteristics occur within an intersectional framework. That is, a person’s various social identities within various socioecological levels overlap to create unique and interdependent systems of discrimination or disadvantage. This brings to light one limitation of this study, which is that individuals were not able to report more than one discrimination reason for a given EMA prompt. Thus, for example, while a Black individual may have reported Racial discrimination as the reason for their experience, one must interpret this response as their primary reason; we do not know whether they may have felt there were other reasons (e.g., gender, sexual orientation, etc.) for their experience as well. Nevertheless, these results show that various types of discrimination are differentially associated with a variety of sleep health features among Black and White young adults. Early identification of these sleep vulnerabilities through, for example, culturally-tailored assessments and interventions that can address the unique needs of the individual (Zhou et al., 2022) could help prevent and manage the adverse outcomes associated with poor sleep (Bruce et al., 2017).

A second limitation of the current study is reliance on self-report of sleep characteristics. This is particularly relevant as research has shown that Black individuals may underreport sleep disturbances relative to objective sleep measures (Jackson et al., 2018; Jackson, Ward, et al., 2020). An analysis of sleep actigraphy data would be beneficial as it can be measured repeatedly over time and is conducive to this type of study. Indeed, some recent advancements in ambulatory or at-home technologies that track physiological response, measure electrical activity in the brain, etc. could be employed to do the same. As a function of acute discrimination experiences, objectively measured sleep will help to further support sleep diary findings or illuminate any discrepancies in subjective and objective sleep measures.

In future studies, it will be important to examine relationships between individual types of discrimination and sleep to understand the non-racial discrimination drivers more clearly. Relatedly, extending this work to other racial groups and incorporating intersectionality into the analytical framework is necessary. For example, experiences of sex- or gender-based discrimination may have a notable association with sleep among Black or White women, but the current study was not specifically designed to assess an intersectional framework. Furthermore, a direct analysis into the mechanistic pathways by which discrimination may impact sleep, such as examining mediation by stress or hypervigilance, is necessitated in future studies. Finally, while the present study focuses on interpersonal discrimination, future work must consider broader forms of discrimination that may occur at multiple levels of a socioecological model (Billings et al., 2021; Grandner & Fernandez, 2021) and may include organizational injustice, racial segregation, or systemic or internalized racism (Slopen et al., 2016). Indeed, the impact of acute discrimination experiences on sleep health also signifies the need for programs designed to reduce the presence of discrimination by addressing prejudice, stereotyping, and other biases within organizations. Trainings conducted as part of a systematic and planned organizational development effort (i.e., integrated training in the workplace and classroom), programs with an inclusive focus, and models that support learners through the entire diversity training process (i.e., pre-training, training, and post-training) have shown to be the most effective diversity training initiatives, though more work is needed in this area as well (Bezrukova et al., 2012; Roberson et al., 2022). Such efforts are cohesive with calls to action from the behavioral sleep medicine community to shift focus from identifying and describing inequities in sleep health to taking action to eliminate these inequities through a multi-pronged approach that includes diversity training (Hughes et al., 2023).

In conclusion, in our sample of young adults reporting regular alcohol use, individuals reported worse sleep on days when they experienced discrimination, particularly in terms of reduced total sleep time. While replication will be important, these preliminary results help to address an important gap in understanding how acute, within-person experiences of discrimination collected in the moment affect sleep health that night. Moreover, while most studies have typically focused solely on racial discrimination, we also examined, in aggregate, Non-racial discrimination stemming from education-, age-, and gender-based discrimination, with many of our findings related to experiences of Non-racial discrimination in White individuals. In the future, it will be important to develop studies that are designed and powered to examine both racial and non-racial identity-based discrimination experiences (e.g., related to age, education, socioeconomic status, sexual orientation, gender identity) at the intersection of various individual identities. Given the association between acute racial discrimination and sleep health, development of culturally tailored sleep treatments that directly include discussion of how discrimination experiences affect sleep will be beneficial.

Significance:

Daily acute discrimination experiences adversely affect same-night sleep health which contributes to racial inequities in sleep and, subsequently, to poor health outcomes. Treatments must be developed to account for different sleep vulnerabilities for individuals experiencing discrimination.