Perceived Discrimination and Adolescent Sleep in a Community Sample

LITERATURE REVIEW

Public health concern is growing over sleep quality and duration, also known as sleep health, in the United States. Nearly 25 percent of the U.S. population report insomnia complaints and 10 percent meet clinical criteria for insomnia (see Irwin 2015). Due to the importance of sleep for numerous health and behavioral outcomes, interest in studying sleep quality and duration is considerable (Kingsbury, Buxton, and Emmons 2013; Gregory and Sadeh 2012; Irwin 2015). Moreover, increasing evidence suggests that sleep characteristics differ across populations and are linked to social and environmental conditions.

Sleep patterns appear to vary by sociodemographic factors that include age, socioeconomic conditions, race and ethnicity, and education (Hale, Emanuele, and James 2015). Adolescents, for example, as a group require more sleep on average (approximately nine hours) but are less likely to actually get adequate sleep than adults (Becker, Langberg, and Byars 2015). People living in economically disadvantaged and segregated neighborhoods also have poorer sleep quality, in part due to the excess noise and crowding found frequently in such environments (Hale, Emanuele, and James 2015; Massey 2004). Race differences are also documented, particularly among African Americans relative to whites; African Americans are a population at higher risk for living in or near economically disadvantaged neighborhoods and at higher risk for disruptive chronic stressors (Williams 2012). Consequently, this population commonly shows signs of harmful sleep patterns by sleeping both shorter and longer durations than the recommended average and accompanied by poorer sleep quality that is less restorative (Kingsbury, Buxton, and Emmons 2013; Mezick et al. 2008; Profant, Ancoli-Israel, and Dimsdale 2002).

Yet much of the literature has used subjective sleep reports rather than objective measures of sleep, and that subjective reports have relatively low reliability relative to more objective measures is well documented (Short et al. 2012, 2013). Respondents asked retrospectively about their sleep duration, for example, are more likely to overestimate duration and underestimate number of awakenings during the night; adolescents in particular may tend to only report more salient, recent information (Wolfson et al. 2003). Daily diaries are another self-report approach to measuring sleep that has the advantage of capturing day-to-day variation usually over an extended period to characterize sleep. Use of temporally proximal data on sleep habits across a range of days is more strongly correlated with laboratory-based gold standard measures of sleep quality than one-shot, long-term subjective self-reports (Wolfson et al. 2003). Among self-report measures, sleep diaries appear to provide the most reliable sleep measures, particularly among adolescents (Short et al. 2013). Sleep diaries, however, tend to overestimate total objective sleep time and underestimate awakening frequency during the night because individuals may not always be aware of waking during the night or other factors that may make sleep more or less restful (Short et al. 2012; Wolfson et al. 2003).

Although self-reported measures of sleep can provide meaningful information and in some cases be correlated with objective sleep measures, they are not as reliable in accurately measuring total sleep duration, waking after sleep onset, or activity during sleep. In nonlaboratory settings, the current state-of-the-art method for objectively measuring sleep dimensions is actigraphy (Short et al. 2012). Actigraphy uses accelerometers placed on the wrist to document sleeping and waking states (Marino et al. 2013). Though actigraphy cannot measure specific sleep architecture or sleep staging, it has been validated to accurately measure distinctions between sleeping and waking along with total sleep time (TST) in both sleep disordered and general populations (Ancoli-Israel et al. 2003). Actigraphs are particularly useful for nonlaboratory studies because they can be worn for extended periods for tracking sleep patterns, which cannot be feasibly measured in sleep labs. Despite the convenience of actigraphy for measuring sleep in the field, use of it is not yet as common in studies of adolescence. We broadly characterize adolescents’ sleep using a rich set of actigraphy measures taken nightly over a two-week period, emphasizing a key exclusionary social experience, discrimination, and relationships to different features of sleep.

DATA AND METHODS

Data in this study come from the Community Connections and Family Wellness Study, which was designed to assess the intergenerational health and well-being of parent-child dyads and was originally conceived of as a two-part pilot study comprising a mail survey screening for an in-lab component to develop and assess recruitment protocols for a larger social neuroscience project (Falk et al. 2013). Sampling and data collection methods included three stages ranging over two years. Figure 1 outlines the stages (S1–S3) of data collection for this multistage sample. In stage 1, in collaboration with a local school system in a mid-sized Midwestern city, we contacted a subset of 2,181 (1,000 white, 1,181 African American or biracial) middle- and high school students ages eleven through fourteen in the school district during the fall of 2014. Two survey packets, one for the parent or guardian and another for the student, were mailed to each family in summer 2014 with a $2 incentive for each. A total of 817 parent-guardian and child dyads completed the mail surveys in stage 1.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

CCFW Study Design and Discrimination Measures

The stage 2 data collection took place from December 2014 to June 2015 with a subset of 141 parent-child pairs from stage 1 who expressed interest in participating in future studies. These participants were invited to visit a laboratory space on a local college campus where anthropometrics, biospecimans, experimental, and additional survey data were collected. Participants in this phase of the study were restricted to families of white, African American, and biracial youth with one African American parent. A $20 incentive was given to each participant along with meal vouchers for participants who were interested in visiting campus (based on responses to an incentive question during stage 1). The overall response rate for participation in the stage 2 portion of the study was 32.4 percent.

Approximately a year later, a subset of forty parent-youth dyads from the in-lab participant pool were recruited for a third stage (stage 3) in-home data collection between February and May 2016. At the initial home visit, after parental consent and child assent, the parent and child independently completed baseline survey measures of health, daily experiences, diet, sleep behavior, and anthropometric measurements. Participants were then trained to use ActiGraph Black/wActiSleep BT and Red/wGTX-BT wristbands and the daily online e-diary procedure (Eufemia et al. 2012). Both the parent and youth were sent an email nightly with a link to their e-diary, which was formatted for use on either a computer or smart device. Both the parent and youth were asked to wear the actigraph wristband all day and night for fourteen consecutive days to measure sleep activity. At the end of the period, parents and youth returned the actigraph wristbands and, to facilitate recruitment and retention of participants, were compensated with cash (up to $200 for the family based on adherence to the study protocol). High rates of protocol adherence (approximately 90 percent) were obtained with both e-diary and actigraph protocols.

Of the 113 families contacted for participation, fifty-eight could not be reached or did not return messages, and thirteen refused, had moved, or were otherwise ineligible. Based on the race of the child, twenty-one of the families were white, four were African American, and fifteen were biracial (one biological African American parent). However, because of missing sleep and biological data, only thirty-five families were used in these analyses. The distribution across race groups in the current analytic sample was twenty-one white, four African American, and ten biracial adolescents. Of these, sixteen were male and nineteen were female. For this study, African American and biracial youth were combined into one category because of the small sample sizes.

MEASURES

Sleep measures are derived from the mathematical decomposition of accelerometer data recorded continuously on the actigraph wristbands. The devices were configured to collect data at 60hz with the idle sleep mode enabled. The actigraphy measures were calculated using the ActiLife 6 software package (ActiGraph 2012). The actigraph data were segmented into sixty-second epochs and wear time validation was assessed using Troiano’s algorithm (Troiano et al. 2008). Sleep analyses then utilized Sadeh’s algorithms and were manually adjusted using bed and wake time information collected in the nightly e-diary (Sadeh 2011; Sadeh, Sharkey, and Carskadon 1994).

We present a diverse number of sleep measures, including multiple measures of both sleep duration and quality, in this study. Descriptions are presented in table 1 for reference. Example data are presented in figure 2 for illustrative purposes. The lightly shaded sleep period denotes the time in bed and the darker shaded period denotes the time asleep. The top chart in figure 2 shows a highly efficient night of sleep with only short interruptions as indicated by physical movement. The second chart shows much less efficient and poorer sleep with longer periods of disrupted wakefulness and delayed sleep onset. In addition to the actigraphy data, the bottom row of table 1 describes a sleep factor score that was constructed after conducting exploratory factor analysis of the following sleep quality items: sleep efficiency, waking after sleep onset (WASO), average awakening length, movement index, and sleep fragmentation. The scale was then validated using a confirmatory factor analysis to provide an overall sleep quality measure. All sleep measures in the statistical analysis were standardized over the entire sleep distribution to facilitate effect size comparisons.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Actigraph Sleep Data, Good (Top) and Poor (Bottom) Nights of Sleep

RESULTS

Descriptive statistics for the sleep variables are presented in table 2 for the total sample, by race and by gender. In general, participants fell asleep within two minutes and with a sleep efficiency rating of approximately 80 percent, which is less than normal healthy sleep efficiency (85 percent). The youth in this sample spent considerable time in bed, nearly (521/60 = ) 8.7 hrs and had a TST of nearly seven hours (413/60), well below the recommended sleep time for this age range of approximately nine hours. The wake after sleep onset averaged around ninety minutes and awakenings per night, of approximately two minutes each, totaled about twenty-three. Differences between groups were minimal other than an indication that biracial and African American youth had more fragmented or restless sleep than white youth, which falls in line with prior literature that finds African Americans have lower average sleep time and more disrupted sleep quality (Hale and Do 2007; Krueger and Friedman 2009).

Table 3 presents the descriptive demographic and health characteristics along with averages of both daily and retrospective discrimination levels reported in the sample. The youth in this sample were relatively disadvantaged, about half coming from homes making less than $45,000 per year. Overall, retrospective EDiS average across items was low with the 1-value category reflecting “less than once a year” and the two-value category reflecting “a few times a year.” However, youth reported 0.6 events per day, on average, suggesting that the retrospective EDiS may underreport the amount of discriminatory experiences that young people perceive when the questions are posed more closely in time to those experiences. It is also important that youths’ average retrospective reports of discrimination using the EDiS showed similar average trends to our initial stage 1 sample of youth from which this sample was drawn upon, who reported discrimination (see figure 3). Daily discrimination report counts were higher for biracial or African American youth, but they had fewer somatic complaints. Females reported both more somatic complaints and poorer self-rated health. Both biracial and female youth were more likely to come from disadvantaged families. The biracial youth had lower CRP values, as did females, but biracial and African American youth had higher average Hba1c and EBV antibody levels.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Kernel Density Plot Everyday Discrimination Distribution

DISCUSSION

This study examines the links between everyday discrimination, daily variations in discrimination exposure, and objective daily measures of sleep using actigraphy among a diverse sample of adolescents. It contributes to the existing literature documenting the harmful consequences of discrimination for health outcomes by demonstrating the associations between discriminatory and exclusionary experiences for objective daily measures of sleep quality in adolescence. To our knowledge, this is the first study to concurrently measure both retrospective reports of Everyday Discrimination (EDiS) and daily diary reports of discrimination with concurrent objective sleep measures.

The findings support our first hypothesis that average sleep would be linked to discrimination. In fact, average discrimination, operationalized using the retrospective EDiS scale was associated with poorer sleep outcomes across all but two included measures. In general, youth who report more discrimination had shorter sleep duration and poorer sleep quality. Specifically, on indicators of sleep duration, youth who reported discrimination took longer to fall asleep (latency), had less efficient sleep, and spent more time in bed but less time asleep. In terms of sleep quality, youth who reported discrimination had more awakenings after sleep onset, longer duration awake during a sleep disruption, moved more while asleep and had more fragmented sleep. This study falls in line with research using gold standard laboratory techniques finding evidence among adults that discrimination exposure is linked to less time in restorative slow wave sleep (stage 4) (Thomas et al. 2006).

Hypothesis 2 posited that day-to-day variation in perceived discrimination would be adversely associated with sleep variability around the person-average. Thus hypothesis 2 is a within-person hypothesis, suggesting discrimination as a potential source of individual variability in sleep quality. This hypothesis was consistently contradicted by our results, which suggest that contemporaneous exposure is linked to improved sleep efficiency, longer total sleep time, less time awake after sleep onset, and decreased movement and sleep fragmentation. Given that youth more likely to report discrimination having happened over the course of the day are generally more likely to report accumulated everyday discrimination, these results suggest a partial recovery in sleep quality following the negative experience. It is important, however, that despite daily indications of improved sleep on the day of the event, the average sleep duration and quality remain lower in youth who report discriminatory events. In general, around three discriminatory experiences in one day would be required to make up for the average decrease in sleep quality over days across those sleep features given the effect sizes estimated.

In her study measuring discrimination exposure and minority youth psychological well-being, Yip posits that for minority youth who experience discrimination, sleep may in fact be a health coping mechanism that lessens the deleterious impact of discrimination in the short term (2014). Her study finds that youth who experience discrimination and have better daily sleep quality experience higher self-esteem and lower depressive symptoms. It does not, however, contemporaneously measure both daily sleep and discrimination exposure contemporaneously, pointing to the need for more research with larger and more diverse samples using designs similar to those in this study.

An important unmeasured component in this study that may shed light on our surprising results and merits examination is the role of active coping style in the face of social stressors. Sleep is an essential component for healthy adolescent development, yet variations in how coping styles impact sleep or how sleep can be used as a coping tool have yet to be systematically examined. Escape to sleep describes how individuals who have disengaged coping styles may use sleep to regulate exposure to adverse stressful emotions or social conditions. Conversely, individuals who engage in more emotionally focused coping may interpret sleep as a loss of mastery and thus are more prone to heightened arousal and sleep disruption (Sadeh, Keinan, and Daon 2004). Although this study does not include specific measures of coping styles, integrating behavioral and emotional coping among diverse populations is an important direction to take future research in this area.

Despite the novelty of these findings, this study is limited in several important ways. First, although actigraphy provides validated objective measures of sleep quality and duration, validation studies indicate that actigraphs can suffer from low specificity or accuracy when detecting wakefulness, which may affect a number of sleep indices (Sadeh 2011). However, it has been suggested that aggregate data over at least four to five nights can compensate for this issue. This study includes fourteen days of data. Second, the sample size is both small and is based on convenience sample in a single community. Obviously, more powerful samples, and samples constructed using state-of-the-art sampling methodologies are important for better characterizing sleep variability and enhancing generalizability to broader populations. Finally, though racial and ethnic heterogeneity was also low, the racial diversity in this sample is mostly biracial youth with an African American parent, which is novel. More diversity, however, is needed for understanding how the experience of discrimination shapes sleep and health over the early life course.

Overall, discrimination is consistently related to poorer sleep and poor sleep appears to be related to an important long-term health marker already by adolescence. Moreover, supplementary analyses indicated that sleep did not predict discrimination reports the following day, and measures of depressive symptoms did not predict sleep or mediate the discrimination parameters. Taken together, these results support the small but growing literature demonstrating the harmful consequences of discrimination for sleep health and the extensive literature demonstrating the association between discrimination and health risk (Slopen, Lewis, and Williams 2016; Williams 2012).