Loneliness and Sleep in the American Indian Blackfeet Community

Abstract

Background:

Sleep is largely understudied in American Indians (AIs), even though sleep is implicated in the chronic diseases which disproportionately affect AI communities.

Objective:

To investigate relationships between daily self-reported loneliness and sleep as measured with actigraphy.

Methods:

In a sample of 98 Blackfeet adults living on the Blackfeet reservation in Montana, we used Ecological Momentary Assessment and actigraphy over a week-long period to investigate relationships between loneliness and sleep. Loneliness was measured daily using the Short Loneliness Scale and actigraphy was used to measure total sleep time, sleep onset latency (SOL), wake after sleep onset (WASO), and sleep efficiency (SE).

Results:

Using a series of generalized linear mixed-effects models controlling for demographic characteristics, anxiety and depressive symptoms, and adverse childhood experiences, we found that those who were lonelier had higher WASO and SOL, and lower SE relative to those who were less lonely. Within-subject effects indicate that participants who were lonelier for a given day relative to their own weekly average had higher WASO that night relative to their own weekly average.

Conclusions:

Our findings provide initial preliminary evidence suggesting that loneliness may be a psychosocial factor which contributes to poor sleep in AI communities.

The American Indian (AI) population suffers from persistent disparities in health, with disproportionately high incidence of diabetes, cardiovascular disease, and depression [1–7]. Known risk factors for these chronic diseases include smoking, physical inactivity, obesity, and poor general health status, all of which are more prevalent in AIs compared to non-hispanic Whites [8]. In addition to these traditional risk factors, a growing body of work in other racial and ethnic groups indicates that poor sleep quality is a risk factor for the chronic diseases which disproportionately affect AIs as a population [9–13]. However, the potential causes and consequences of poor sleep are largely understudied in AI communities.

The Blackfeet tribal reservation is located approximately 40 miles South of the Canadian border in Browning, Montana, and the Blackfeet tribe consists of approximately 18,000 enrolled members. In 2017, the Blackfeet tribal community completed a comprehensive health assessment identifying the main health concerns for the community. The report noted that diabetes and cardiovascular disease were among the top health issues demanding attention [14]. However, to date, sleep on the Blackfeet reservation has not been studied, and it remains possible that sleep may be a pathway which contributes to incidence of these chronic diseases.

Although the literature on AI sleep is limited, initial data indicate that AIs have worse sleep compared to other racial and ethnic groups. For example, AIs are less likely to have healthy sleep duration [15–16]. However, sleep research in exclusively AI adult populations has been limited methodologically to self-reported measures of sleep. Furthermore, while sleep research in other racial and ethnic groups indicates that sleep is affected and shaped by psychosocial factors including stress, depressive symptoms, family environments, trauma exposure and social integration [17–21], these relationships have not been adequately investigated in AIs.

One psychosocial factor that may be particularly relevant for sleep in AIs is loneliness. In our previous work, we found that sense of belonging was related to sleep in a sample of AI college students [24]. While sense of belonging and loneliness are related in that they are both subjective states related to social experiences, they are distinct constructs. Sense of belonging reflects the internal relationship between an individual and a group to which they belong. An individual who reports a high sense of belonging with reference to a group or community feels that they are integrated member of this group or community. When an individual has a low sense of belonging, they are more likely to have negative emotional experiences, including loneliness. While loneliness may be informed by one’s sense of belonging, it is more broadly defined as the distressing feeling one experiences when they perceive their social connections and relationships to be inadequate. As such, loneliness reflects one’s overall satisfaction with all social connections and relationships. This may include one’s sense of belonging to a group or community, but also reflects their overall satisfaction with all social relationships.

Loneliness has been linked to poor sleep in other racial and ethnic groups [25–27]. It is theorized that for humans, being embedded in a social network and feeling connected to others gives rise to a sense of security, while feeling lonely or isolated provides a sense of vulnerability and increased vigilance [28]. Based on this model it is postulated that loneliness may impair sleep, since the state of sleep is not conducive to be highly vigilant. In line with this theory, feelings of loneliness have been linked to reduced subjective sleep quality, daytime dysfunction and sleep fragmentation [29], and have been found to prospectively predict short sleep duration and sleep disturbances [26]. In some prior work, loneliness has been associated with sleep disturbances, but not sleep duration [293–2]. Given the known relationships between loneliness and sleep in other racial and ethnic groups, it is possible that loneliness may negatively impact sleep in Blackfeet adults. As such, sleep could be a critical pathway through which a psychosocial factor (i.e., loneliness) exacerbates risk for the most common chronic diseases on the Blackfeet reservation. Examining this relationship in an AI community may be particularly important given that previous work indicates incidence of loneliness differs across racial groups [33] and given that race and ethnicity appear to moderate the relationship between loneliness and health-relevant outcomes [34–36].

The present study was conducted in the Blackfeet Community and utilized community based participatory research (CBPR) methods_, which places emphasis on the importance of equitable involvement of community members and researchers in all stages of the research process [37–39]. The large majority of work focused on the relationship between social experiences and sleep in other racial groups uses the construct of loneliness to capture individuals’ satisfaction with their social relationships [29–32]. Thus, in order to be able to place the findings from this research in the context of this body of work, we chose to measure loneliness over other belongingness measures as a first step to understanding the link between social environments and sleep in AIs. To assess the connection between sleep and loneliness, we used ecological momentary assessment to measure daily reported loneliness over a week-long period in a sample of 100 adults living on the Blackfeet reservation. In addition to reporting on their daily loneliness, participants concurrently wore a wrist-accelerometer to obtain objective, actigraphy-derived measures of sleep, including total sleep time (TST), wake after sleep onset (WASO), sleep onset latency (SOL), and sleep efficiency (SE). We hypothesized that greater loneliness would be associated with poorer sleep, including lower SE, and higher WASO and SOL. Based on previous work which found that loneliness is related to sleep disturbances but not sleep duration [30], we did not predict that loneliness would be related to TST in this sample. To our knowledge, the findings from this work are the first to provide insight into the relationships between loneliness and objectively measured sleep in an AI community.

Methods

This study was approved by the Blackfeet Nation Institutional Review Board and the Montana State University Institutional Review Board. Following CBPR methodology, research, questionnaires, and measurements, were reviewed and approved by a community advisory board consisting of 4 members of the Blackfeet Community. Exclusionary criteria for participation included any chronic health or sleep disorder diagnoses. Participants were asked as part of eligibility screening if they had any chronic health conditions or sleep disorders. If community members reported any chronic health conditions or sleep disorders, they were not invited to participate in the study. While data collection occurred on the Blackfeet Community College Campus, we placed advertisements on various Blackfeet social media sites, community centers, and on in the local newspaper, with the intent of recruiting a socioeconomically diverse and representative sample. We obtained a sample of 100 AI adults ranging from 20 to 78 years of age (M=42.30, SD=15.08). Participants visited an office located in a building on the Blackfeet Community College campus. They met with the project coordinator and provided written informed consent, after which they completed a series of questionnaires including demographic measures and a measure of childhood trauma.

The project coordinator installed the Illumivu Ecological Momentary Assessment application (www.lifedatacorp.com) on participants’ mobile devices. This mobile application prompted participants to complete surveys at the beginning of each day and at the end of each day. During this visit, the project coordinator showed participations how to complete each of these surveys on their mobile devices. The project coordinator also gave each participant a wrist accelerometer validated for sleep assessment (Actigraph, GT9X link, Penascola, FL), which was worn on participants’ non-dominant wrist. Participants were asked to wear the device continuously for 7 days and 7 nights, after which they met with the project coordinator to return the device and download the data from the Illumivu EMA app.

Measures

Loneliness.

At the end of each day, before going to bed, participants completed the short loneliness scale on their mobile devices [40]. Using a 4-point likert scale, respondents indicate the frequency with which they felt isolated, lacking in companionship, and left out. The scale has been used previously in large population studies as an acute measure of loneliness [40]. The average Cronbach’s α across the week of monitoring was 0.88 for the short loneliness scale.

Sleep Measures.

Sleep measures were obtained using 7 nights of data from wrist-actigraphy. Wrist accelerometers are widely used in research studies and have been validated against polysomnography [41–43]. All participants had at least 5 nights of valid data, as prior work indicates that a period of 5 days is ideal to provide an adequate reflection of an individual’s sleep-wake cycles [44–45]. Nighttime sleep periods were determined using self-reported sleep onset and offset from sleep diaries in combination with visual inspection of activity levels. Data were collected using the zero-crossing mode in one-minute epochs and sleep parameters were estimated using a medium detection threshold and the Cole-Kripke algorithm [46]. Nightly sleep duration was the sum of the number of sleep minutes during the night-time sleep period. WASO was calculated as the number of minutes spent awake within the total nighttime sleeping period. SE was calculated as the number of sleep minutes divided by the total duration of time between sleep onset and offset, multiplied by 100.

Symptoms of Depression and Anxiety.

We used the Hospital Anxiety and Depression Scale (HADS) as a measure of current symptoms of depression and anxiety during the initial lab visit [47]. The HADS has 14 items, 7 items comprise a depression subscale, and 7 items comprise an anxiety subscale. Individuals respond to each item using a four-point response category (0–3), with possible scores ranging from 0–21 for depression and 0–21 for anxiety. Cronbach’s αwas 0.85 for the anxiety subscale and 0.88 for the depression subscale.

Early life Trauma.

The Risky Family questionnaire was used to assess childhood exposure to physical, mental and emotional abuse or neglect [48]. Using a 5-point Likert scale (1=not at all and 5=very often), participants indicated the frequency with which certain events or situations occurred during childhood (ages 5–15). Example items from this measure include, “How often would you say there was quarreling, arguing or shouting between your parents?” and “How often would you say that a parent or other adult in the household behaved violently toward a family member or visitor in your home.” Responses to each item are summed to create a total score, with higher numbers reflecting more early life trauma. The measure demonstrated good internal reliability (Cronbach α=70)

Annual Income.

Participants reported their annual household income on a scale from 1 (US$20,000 and below) to 6 (US$110,000 and above) [49–50].

Analytic Technique

Hypotheses were tested using a series of generalized linear mixed-effects models estimated using the lme4 package in R [51]. All models included a random effect for subject and fixed effects for other predictors. The primary independent variable was loneliness, which was measured at Level-1 and between and within-person effects calculated separately. Within-person effects for loneliness were estimated by calculating the deviation of each person’s observed loneliness for a given day from that person’s average loneliness across the week. Between-person effects were estimated by calculating the deviation of each person’s average loneliness across the week from the grand mean for loneliness. Level-2 covariates included age, gender, income, anxiety symptoms, depressive symptoms, and adverse childhood experiences, all of which were grand-mean centered. To account for possible linear changes in sleep across the week, measurement time point was included as a Level-1 covariate. Level-1 dependent variables included TST, summed scores for the WASO, SOL, and SE. Intra-class correlation coefficients (ICCs) were estimated to quantify the within and between-person variances in the dependent variables. Four generalized linear mixed-effects models were estimated to test associations among within-person daily loneliness, between-person loneliness, and each sleep outcome.

Results

Two participants were missing data on either the sleep outcomes or loneliness and were removed from analyses. All remaining participants had complete data. Fifty five percent of our sample had a high school education or less. Table 1 displays the means, standard deviations, and bivariate correlations for all study variables averaged across the week. The average TST across the week was 361 minutes (approximately 6 hours) a night. Average SOL across the week was 19 minutes and the average SE was 81%. Greater average depressive symptoms and loneliness scores across the week were significantly correlated with higher WASO. Greater average depressive symptoms across the week were significantly correlated with lower average sleep efficiency.

Table 1.

Means, standard deviations, and correlations

Variable	M	SD	1	2	3	4	5	6	7	8	9	10
1. Age	42.28	14.96
2. Gender	1.55	0.50	.11
3. Income	1.76	1.10	−.03	−.15
4. ACES	6.25	3.23	.10	.07	.21*
5. Anxiety Symp.	9.68	3.52	.12	.04	.02	.23*
6. Depression Symp	9.67	4.50	.01	.14	.11	.26**	.59**
7. Av. Total Sleep	361.20	78.51	−.03	−.06	.21*	.01	−.00	−.06
8. Av. Loneliness	5.82	1.83	−.07	.05	.02	−.19	.09	.25*	.08
9. Av. WASO	51.51	25.63	−.12	.03	.06	.03	.12	.27**	−.02	.51**
10. Av. Sleep Latency	19.25	7.25	−.03	−.03	.10	.12	−.04	.11	.10	.19	.48**
11. Av. Sleep	81.29	11.52	.5	−.02	.01	−.00	−.17	−.25*	.19	-	-	-
Efficiency										.35**	.66**	.64**

Note. M and SD are used to represent mean and standard deviation, respectively.

^*indicates p < .05.

^**indicates p < .01. ACES = Adverse Childhood Experiences. Symp. = Symptoms. Av. = Average.

ICCs were calculated from unconditional models for each of the four sleep outcomes to determine the degree to which subject accounted for variation in the variables of interest. Figure 1 displays spaghetti plots to visualize within and between-person variation in sleep trajectories over time. Overall, 98% of the variance in total sleep time was attributed to between-person effects and 2% of the variance was attributed to within-person effects. Additionally, 92% of the variance in WASO scores was attributed to between-person effects and 8% of the variance was attributed to within-person effects. For sleep onset latency, 91% of the variance was attributed to between-person effects and 9% was attributed to within-person effects and 96% of the variance for sleep efficiency was attributed to between-person effects and 4% was attributed to within-person effects.

Figure 1.

Spaghetti plot depicting between and within-person variability in sleep outcomes across the week. Figure 1a variability in total sleep time across the week. Figure 1b variability in WASO scores across the week. Figure 1c variability in sleep onset latency scores. Figure 1d variability in sleep efficiency scores.

A series of generalized linear mixed models were used to estimate within and between associations between loneliness, TST, WASO, SOL, and SE after accounting for demographic characteristics, anxiety and depressive symptoms, adverse childhood experiences. Table 2 displays the model estimates. There was a linear effect of time, with SOL and SE becoming significantly lower across the week. Additionally, those with higher income had greater TST across the week relative to those with lower income. Between-subject effects indicate that those who were lonelier had higher WASO, higher SOL, and lower SE relative to those who were less lonely. Within-subject effects indicate that participants who were lonelier for a given day relative to their own weekly average had higher WASO that night relative to their own weekly average. We did not find evidence of a significant between or within-person effects of loneliness for TST, or significant within-person associations for SOL or SE.

Table 2.

Estimates from General Linear Mixed Models Predicting Total Sleep Time, WASO, Sleep Onset Latency, and Sleep Efficiency.

	Total Sleep Time			WASO			Sleep Onset Latency			Sleep Efficiency
	B	SE	95% CI	B	SE	95% CI	B	SE	95% CI	B	SE	95% CI
(Intercept)	303.03	8.15	287.05 – 319.00	51.93	2.32	47.38 – 56.47	19.90	0.74	18.44 – 21.35	82.00	1.13	79.78 – 84.21
Time	0.14	0.19	−0.23 – 0.52	−0.11	0.14	−0.39 – 0.17	−0.14	0.04	−0.22 – −0.05	−0.18	0.04	−0.27 – −0.10
Age	−0.20	0.56	−1.30 – 0.91	−0.12	0.16	−0.43 – 0.18	0.01	0.05	−0.09 – 0.11	0.01	0.08	−0.14 – 0.16
Gender	−3.63	16.97	−36.89 – 29.64	−0.42	4.71	−9.65 – 8.80	−0.81	1.51	−3.77 – 2.15	0.52	2.33	−4.06 – 5.09
Income	15.46	7.79	0.20 – 30.72	−0.19	2.16	−4.42 – 4.05	0.21	0.69	−1.14 – 1.57	0.53	1.07	−1.57 – 2.62
Adverse Child. Experiences	0.07	2.78	−5.37 – 5.51	0.82	0.77	−0.69 – 2.33	0.36	0.25	−0.13 – 0.84	−0.06	0.38	−0.81 – 0.69
Anxiety Symptoms	1.67	2.91	−4.04 – 7.37	−0.16	0.81	−1.74 – 1.42	−0.32	0.26	−0.83 – 0.19	−0.17	0.40	−0.96 – 0.61
Depressive Symptoms	−2.24	2.40	−6.95 – 2.47	0.81	0.67	−0.50 – 2.11	0.21	0.21	−0.21 – 0.63	−0.38	0.33	−1.03 – 0.27
Loneliness (Between)	2.45	4.79	−6.94 – 11.85	7.10	1.33	4.50 – 9.71	0.90	0.43	0.07 – 1.74	−2.05	0.66	−3.34 – −0.75
Loneliness (Within)	−0.21	0.51	−1.22 – 0.80	1.71	0.39	0.96 – 2.47	0.02	0.12	−0.20 – 0.25	0.03	0.12	−0.20 – 0.26
Random Effects
σ2	95.49			53.57			4.87			4.94
τ00	6442.71SUBJECTID			488.62SUBJECTID			50.32SUBJECTID			121.40SUBJECTID
Observations	668			668			668			668
Marginal R2/Conditional R2	0.052 / 0.986			0.279 / 0.929			0.084 / 0.919			0.148 / 0.967

Notes: Child. = Childhood. All covariates were grand-mean centered. Bolded values are significant at p < .05.

Discussion

To our knowledge, this is the first study to utilize actigraphy to measure sleep-wake cycles in an exclusively AI adult population. Furthermore, while the relationship between loneliness and sleep has been studied previously in other racial and ethnic groups [25–32], we believe this is the first investigation to focus on the relationship between loneliness and sleep in AIs. Specifically, our sample was comprised of adults living in Browning, Montana on the Blackfeet reservation. This is important since sleep is implicated in the most common chronic diseases on the reservation, including cardiovascular disease and diabetes. Our work indicates that on average, Blackfeet adults in this sample are sleeping for 6 hours nightly. This is in line with previous survey data indicating that AIs are less likely than other racial and ethnic groups to have healthy sleep (> 7 hours) [15]. The low average sleep duration we observed here calls for more research to understand effective approaches to lengthen sleep duration in this community.

Our research design allowed for the examination of between- and within-subject differences in the relationship between daily loneliness and sleep. Between-subjects analyses indicated that individuals who were lonelier had higher WASO, higher SOL, and lower SE relative to those who were less lonely. In a similar manner, within-subject analyses revealed that on days where participants reported greater loneliness compared to their weekly average of loneliness, they had greater WASO relative to their weekly WASO average. We did not find evidence of between-person associations among loneliness and TST and our within-subjects analyses did not provide evidence to support statistically significant effects of loneliness on TST, SOL or SE. These findings are consistent with previous work in other racial and ethnic groups investigating relationships between loneliness and sleep. Specifically, across several investigations, loneliness was associated with sleep disturbances, but was not related to sleep duration [29–32]. It is posited that loneliness feels unsafe, which places lonely individuals in a vigilant state [28]. Feeling unsafe is thought to disrupt the sleep of lonely individuals by increasing sleep disturbances [31]. As such, while lonely individuals may not have lower sleep duration, the quality of their sleep may suffer due to a higher number of microawakenings. Future work is needed to understand why in AI adults, loneliness appears to have implications for WASO for a given person but not for the other objective measures of sleep collected in this research. The within-person effects we report for loneliness on sleep raise the possibility that interventions designed to lower loneliness in a given day may have positive impact on Blackfeet adults’ sleep in the evening, regardless of whether they have consistently poor sleep outcomes.

It is important to highlight the low within-person variability for all sleep outcomes in this sample of AI Blackfeet adults. The amount of variance for all of the objective sleep outcomes that was attributed to within-person effects was less than 9%. This finding suggests that in addition to having relatively poor sleep overall compared to other racial and ethnic groups [15], the sleep patterns of Blackfeet adults in this sample appear to be highly consistent across the period of one week. A systematic review found that high intraindividual variability (IIV) in sleep was associated with poor health outcomes including physical health conditions, higher body mass index, and depression symptomology [52]. Based on this, it is possible that the low sleep IIV observed in this sample could be health protective, however it is not clear whether the benefits of low IIV would outweigh the well-documented negative implications of consistently low sleep duration [53]. More work is needed to understand the environmental and psychosocial factors which may contribute to low levels of sleep IIV we observed in this investigation and to better understand the independent contributions of sleep IIV and total sleep time to health outcomes in this population.

These findings provide initial evidence that social environments may be important predictors of health in AI tribal communities. Similar to relationships observed in other racial and ethnic groups, satisfaction with one’s social relationships appears to inform important sleep outcomes. Loneliness in tribal communities may be higher than what is observed in other populations, in part because of geographic isolation which characterizes many tribal reservations. In the future, loneliness data should be collected across a wide range of tribal communities in order to better understand the prevalence of loneliness in these communities and to understand the biobehavioral and psychosocial pathways through which the experience of loneliness might affect sleep. Based on the findings reported here, it is possible that psychosocial interventions which work to reduce loneliness may positively affect sleep, and consequently improve health outcomes for AI communities which have persistent disparities in health linked to sleep.

The following three limitations are acknowledged. First, all participants were AI adults residing on the Blackfeet reservation, and thus the findings cannot be generalized to AI communities broadly. Furthermore, the sample of Blackfeet adults in this sample do not necessarily represent the larger Blackfeet community. It is also possible that the relationship between loneliness and sleep is different across tribal communities, as well as for AI adults who do not reside on tribal reservations. Second, while we had a daily measure of loneliness, these data do not provide a nuanced understanding of the social experiences and environments for participants which may have contributed to fluctuations in daily reported loneliness. This will be an important consideration in future investigations. Finally, we excluded community participants who reported any chronic health conditions or sleep disorders from the study. As such, our sample includes relatively healthy community members, and the investigated relationships may be different in a sample which includes individuals with chronic health conditions or sleep disorders.

In summary, the findings reported in the present study significantly advance the understanding of the associations between loneliness and objective sleep in an AI population with known health disparities. Utilizing actigraphy and ecological momentary assessment, there were significant inter-individual relationships between loneliness and WASO, SOL, and SE and intra-individual relationships between loneliness and WASO in AI Blackfeet adults. These data support previous findings that have utilized self-report measures to demonstrate that AIs have poor sleep compared to other racial and ethnic groups [15]. Based on findings in other racial and ethnic groups, which highlight the role of sleep in the chronic diseases that disproportionately impact AIs [10–14], it is possible that sleep is a key physiological process that contributes to disease risk in AIs. Future work is needed to understand how sleep might prospectively predict biological and physiological risk for the most common chronic diseases in the Blackfeet community, as well as other AI populations living on and off tribal lands.