Examining sleep disturbance among sheltered and unsheltered transition age youth (TAY) experiencing homelessness

Abstract

Background:

The estimated 3.5-million transition age youth (TAY) who experience homelessness in the United States annually are routinely exposed to inadequate sleep environments and other psychosocial risk factors for deficient sleep. While staying in a shelter versus being unsheltered may facilitate sleep, research suggests that perceived safety wherever one sleeps may be just as important. In this study, which is the first known study to investigate sleep disturbances among TAY experiencing homelessness, we examine associations of sleep disturbances with sheltered status and perceived safety of usual sleep environment.

Methods:

We surveyed TAY (aged 18-25) experiencing homelessness in Los Angeles, CA about their sleep, psychosocial health, and living situations. Participants (n=103; 60% sheltered) self-reported sleep disturbances using the Patient-Reported Outcomes Measurement Information System (PROMIS) Sleep Disturbance short form, while individual items assessed sheltered status and perceived safety where they usually slept. Regression analyses examined associations of sheltered status and perceived sleep environment safety with sleep disturbance, adjusting for age, sex, race, self-rated health, depression symptoms, serious mental illness, high-risk drinking, and severe food insecurity.

Results:

Twenty-six percent of participants reported moderate-severe sleep disturbances. Sleep disturbance was not associated with sheltered status, but was positively associated with feeling unsafe in one’s sleep environment, depression symptoms, severe food insecurity, and decreased age.

Conclusions:

Our findings suggest that sleep disturbances among TAY experiencing homelessness are associated more closely with how safe one feels rather than one’s sheltered status. This highlights the importance of providing safe places to live for sheltered and unsheltered TAY.

Introduction

The estimated 1 in 10 young adults aged 18-25 who experience homelessness in the United States annually¹—often referred to as transition age youth (TAY)—constitute a vulnerable population that is routinely exposed to inadequate and unsafe sleep environments.² While chronic sleep deficiency (referring to inadequate or non-restorative sleep)³ at this age is a known threat to health, safety, and achievement,⁴ TAY experiencing homelessness are also at increased risk for psychosocial health issues, including depression, substance use, and food insecurity,^5–7 each of which are risk factors for deficient sleep.^8–10 The consequences associated with chronic sleep loss among homeless TAY may exacerbate existing health issues and present obstacles to attaining housing stability. While there is a growing body of literature documenting sleep deficiencies among chronically homeless adults,^11–16 no published studies have examined deficient sleep among TAY experiencing homelessness.

When examining sleep deficiencies among TAY experiencing homelessness, it is likely important to differentiate between those who are sheltered versus those who are unsheltered. The former not only includes TAY living in shelters, but can also include those living in hotel rooms provided by the homeless service system, and by some federal definitions, those who are temporarily staying in other peoples’ homes (i.e., “couch surfing”).^17,18 Unsheltered homelessness refers to sleeping in a place not meant for human habitation, such as outdoors, in a tent, vehicle, or abandoned building.¹⁷

While providing protection from the elements and a bed, couch, or floor to sleep on may improve the built sleep environment as compared to living unsheltered, there is research that suggests detrimental aspects of sheltered contexts for sleep. In two studies that qualitatively investigated sleep among homeless adult shelter utilizers, participants described shelters as often crowded, volatile, and lacking privacy, where disruptions, safety concerns, and physical discomfort due to poor bedding and temperature control could make it more difficult to sleep than in certain unsheltered contexts.^15,16 In another study examining relationships between child abuse, mental health, and sleeping arrangements among TAY experiencing homelessness, sleeping in an unsheltered location was associated with less sexual victimization compared to couch surfing with a family member.² This suggests that how safe TAY feel in the places that they sleep may be an important consideration in addition to whether they are sheltered.

This study sought to address a gap in the literature by investigating sleep disturbance (a measure of deficient sleep) among a sample of sheltered and unsheltered TAY experiencing homelessness. Our aim was to investigate associations of unsheltered status and perceived safety of one’s sleep environment with sleep disturbance. We hypothesized that both unsheltered status and perceived lack of sleep environment safety would be positively associated with sleep disturbance, after adjusting for demographic characteristics and psychosocial risk factors for sleep disturbance among TAY.

Methods

Sample and procedures

Participants were 103 TAY experiencing homelessness in Los Angeles, CA. Participants were recruited from shelter programs and drop-in centers via study flyers and information sessions to participate in a larger study of HIV risk behavior.¹⁹ Eligibility criteria included being 18-25 years old, spending the prior night somewhere that meets federal definitions of TAY homelessness,¹⁸ and being able to respond to English-language surveys without assistance.

After enrolling and providing informed consent, participants used an iPad (Apple, USA) to complete a self-administered questionnaire that lasted about 1 hour. Participants were compensated $20 for completing the questionnaire and $5 for the eligibility screener. The [blinded] Institutional Review Board approved all study protocols.

Measures

Our primary sleep outcome was self-reported sleep disturbance, measured using the Patient-Reported Outcomes Measurement Information System (PROMIS) Sleep Disturbance 4-item short form, a reliable and widely utilized measure of prior 7-day sleep disturbance, regardless of a specific sleep disorder.²⁰ Short form items asked respondents to rank their sleep based on quality, feeling refreshed upon waking, problems, and difficulty falling asleep using 5-point Likert scales, with higher scores corresponding to greater sleep disturbance. Responses were uploaded to the HealthMeasures Scoring Service,²¹ a free web-based platform that uses response pattern scoring to calculate normalized T-scores for PROMIS measures. The PROMIS sleep calibration sample was a mixed clinical and community-based adult sample, which generally has a greater burden of chronic illness than the general population. PROMIS cutpoint guidelines were used to classify sleep disturbance T-scores based on standard deviations (SDs) above the calibration sample mean (severe: ≥70 or 2 SDs; moderate: ≥60 or 1 SD; mild: ≥55 or 0.5 SD; or within normal limits: <55).²²

Sheltered status was determined based on participants’ response to the question, “Currently, where do you stay most of the time?” coded as “1” for unsheltered (e.g., outdoors, in a tent, vehicle, or abandoned building) and “0” for sheltered contexts (i.e., shelter programs, hotels/motels, or “couch surfing”). Perceived sleep environment safety was assessed using a single item that asked, “How safe/secure do you feel in the place you sleep most of the time currently?” on a 5-point Likert scale coded from “very much” (1) to “not at all” (5).

Demographic questions assessed sex assigned at birth, age, race, and Hispanic/Latinx ethnicity. Race was coded as “1” for Black/African American and “0” for white, bi-/multi-racial, another race, or not identified, because of low insufficient cell sizes in other groups. “Not identified” includes participants who identified as Hispanic/Latinx ethnicity but not with any race.

Other psychosocial measures assessed for general self-rated health, depression symptoms, serious mental illness (SMI) diagnoses other than major depressive disorder, high-risk drinking, and food insecurity. Self-rated health was assessed with a 5-point Likert scale ranging from poor (1) to excellent (5).²³ The Patient Health Questionnaire (PHQ-9) assessed depression symptoms in the prior 2 weeks, with higher scores indicative of more severe symptomatology.²⁴ Non-depression SMI diagnosis was coded as “1” if participants indicated that they had ever been diagnosed with bipolar disorder, schizophrenia/schizoaffective disorder, or a personality disorder. Major depression diagnoses were excluded from this indicator because we used the PHQ-9 as a more proximate measure of depression. Recent frequency of high-risk drinking was assessed using questions adapted from the 2017 Youth Risk Behavior Survey (YRBS) that asks how many of the prior 30 days respondents consumed “five alcoholic drinks,” if male, or “four,” if female, “in a row, that is within a couple of hours.”²⁵ The Household Food Insecurity Access Scale (HFIAS), which is recommended for use with homeless populations,²⁶ was used as an indicator of being severely food insecure in the prior 30 days, as opposed to being food secure or mildly–moderately food insecure.²⁷

Analyses

Descriptive statistics were used to characterize the sample, using chi-square, Fisher’s-exact, and T-tests to identify significant differences by sheltered status. Three separate linear regression models examined correlates of sleep disturbance. All 3 models adjusted for age in years, female sex assigned at birth, Black/African American race, self-rated health score, PHQ-9 score, non-depression SMI diagnosis, recent frequency of high-risk drinking, and severe food insecurity. Model 1 included unsheltered status as a predictor of sleep disturbance, Model 2 included perceived safety of participants’ usual sleep environment, and Model 3 included both unsheltered status and perceived safety in order to assess their relative associations when controlling for the other.

Results

Sixty-two participants (60%) were sheltered. As noted in Table 1, our sample was 31% female, 56% Black/African American, 26% Hispanic/Latinx, and 22.1 years old (SD=2.02) on average. The mean self-rated health score was 3.62 (SD=1.21). Twenty-three percent of the sample had PHQ-9 scores indicative of clinically significant depression and 40% had ever been diagnosed with non-depression SMI. Fifty-four percent of participants were severely food insecure. None of these psychosocial characteristics demonstrated statistically significant differences by sheltered status.

Table 1.

Participant characteristics by sheltered status most of the time at enrollment.

	Sheltered (n=62) M (SD) or n (%)	Unsheltered (n=41) M (SD) or n (%)	Full Sample (n=103)a M (SD) or n (%)
Age	21.81 (2.16)	22.44 (1.73)	22.06 (2.02)
Female sex assigned at birth	23 (37.1%)	9 (22.0%)	32 (31.1%)
Race
Black or African American	37 (59.7%)	21 (51.2%)	58 (56.3%)
White	7 (11.3%)	4 (9.8%)	11 (10.7%)
Bi/multi-racial	4 (6.5%)	6 (14.6%)	10 (9.7%)
Another race	7 (11.3%)	1 (2.4%)	8 (7.8%)
Not identified	7 (11.3%)	9 (22.0%)	16 (15.5%)
Hispanic/Latinx ethnicity	13 (21.0%)	14 (34.1%)	27 (26.2%)
General self-rated healthb	3.60 (1.18)	3.66 (1.26)	3.62 (1.21)
Clinically significant depressionc	16 (25.8%)	8 (19.5%)	24 (23.3%)
Serious mental illness diagnosisd	23 (37.1%)	18 (43.9%)	41 (39.8%)
High-risk drinking days (past month)e	1.56 (2.25)	1.10 (2.10)	1.38 (2.19)
Severe food insecurityf	30 (48.4%)	26 (63.4%)	56 (54.4%)
Sleep Disturbance Severityg
Within normal limits	35 (56.5%)	18 (43.9%)	53 (51.5%)
Mild	14 (22.6%)	9 (22.0%)	23 (22.3%)
Moderate	12 (19.4%)	11 (26.8%)	23 (22.3%)
Severe	1 (1.6%)	3 (7.3%)	4 (3.9%)
Safety of usual sleep environmenth	2.39 (1.23)	3.32 (1.17)	2.76 (1.29)***

^a*p<0.05; **p<0.01; ***p<0.001

^bPoor (1) – excellent (5)

^cPatient Health Questionnaire (PHQ-9) scores range from 0-27, with scores >14 indicating moderately severe or worse depression that warrants treatment. Mean PHQ-9 scores also did not vary significantly by sheltered status.

^dEver diagnosed with a non-depression serious mental illness, including bipolar disorder, schizophrenia/schizoaffective disorder, or a personality disorder.

^eRange: 0-8 days. N=62 (60.2%) of participants reported 0 drinking days.

^fBased on Household Food Insecurity Access Scale (HFIAS) category.

^gBased on Patient-Reported Outcome Measurement Information System (PROMIS) guidelines for symptom severity cut points. Mean normalized PROMIS Sleep Disturbance t-scores also did not vary significantly by sheltered status.

^hHow safe/secure do you feel in the place you sleep most of the time currently?” [“Very much” (1) to “not at all” (5).]

The overall mean sleep disturbance t-score was 53.7 (SD=9.97). According to PROMIS cut points, over one quarter (26%) of the sample reported moderate to severe sleep disturbance, while 56% and 44% of sheltered and unsheltered participants, respectively, reported sleep disturbances that were “within normal limits.” Neither mean T-scores nor sleep disturbance severity varied significantly by unsheltered status. Unsheltered participants did, however, rate their usual sleep environments significantly more unsafe on average than sheltered participants (p<0.001).

Results of regression analyses are located in Table 2. In Model 1, being unsheltered was not significantly associated with sleep disturbance (p=0.13). Feeling more unsafe in one’s usual sleep environment (Model 2), however, was positively associated with sleep disturbance (β=1.64, p=0.01), and this association remained significant after controlling for unsheltered status in Model 3 (β=1.48, p=0.04). In all models, increased age was associated with significantly lower sleep disturbance (p=0.02), while higher depression scores and severe food insecurity were positively associated with sleep disturbance (p<0.001). No other covariates were significanlty associated with sleep disturbance.

Table 2.

OLS regression analyses examining associations of sleep disturbance with sheltered status most of the time and perceived safety of usual sleep location (n=103).

	Model 1	Model 2	Model 3
Sleep Disturbance	β (95% CI)a	β (95% CI)	β (95% CI)
Unsheltered most of the time	2.65 (−0.79, 6.09)	---	1.21 (−2.43, 4.84)
Safety of usual sleep environmentb	---	1.64 (0.37, 2.92)*	1.48 (0.10, 2.85)*
Age	−1.02 (−1.88, −0.16)*	−0.98 (−1.82, −0.15)*	−1.02 (−1.86, −0.17)*
Female sex at birth	−1.47 (−5.07, 2.13)	−2.07 (−5.55, 1.41)	−1.86 (−5.41, 1.69)
Black/African American race	0.41 (−3.01, 3.83)	−0.23 (−3.56, 3.10)	−0.05 (−3.43, 3.33)
General self-rated health	−1.33 (−2.75, 0.10)	−1.26 (−2.65, 0.14)	−1.26 (−2.66, 0.14)
Depression (PHQ9 score)	0.50 (0.25, 0.76)***	0.46 (0.21, 0.72)***	0.48 (0.22, 0.73)***
Non-depression SMI diagnosis	−2.50 (−5.91, 0.92)	−2.16 (−5.49, 1.18)	−2.25 (−5.61, 1.11)
High-risk drinkingc	−0.37 (−1.17, 0.43)	−0.47 (−1.25, 0.30)	−0.43 (−1.22, 0.36)
Severe food insecurityd	7.46 (4.00, 10.91)***	6.83 (3.39, 10.26)***	6.74 (3.29, 10.20)***
Constant	73.10***	69.79***	70.34***
F	6.67***	7.43***	6.69***
R2	0.392	0.418	0.421
N	103	103	103

^a*p<0.05; **p<0.01; ***p<0.001

^bHigher scores indicate feeling more unsafe (1=very much, 5=not at all) where one sleeps.

^cDays of consuming more than 5 drinks if male, or 4 drinks if female, within a couple of hours in the prior 30 days.

^dSeverely food insecure = 1; food secure–moderately food insecure = 0.

Discussion

In this first study to examine sleep among TAY experiencing homelessness, we unsurprisingly found evidence of elevated sleep disturbances. Our sample’s mean sleep disturbance t-score was 3.7-points higher than an adult reference population that is generally more enriched for chronic illness than the general population. However, our sample had a lower mean T-score compared with a sample of 32 homeless adults who completed the 8-item PROMIS sleep disturbance instrument and scored nearly 10 points (one standard deviation) higher than the same reference population.¹⁶ Additionally, using a method for generating “low,” “medium,” and “high” sleep disturbance scores that has been previously used with a nationally representative sample of young adults,²⁸ 56% of our sample scored with high sleep disturbances, compared to 33% of young adults in the general population.

Results do not support our first hypothesis that being unsheltered would be associated with sleep disturbance. This may be because unsheltered status itself is too broad a category. While TAY likely encounter environmental problems related to air quality, light, noise, and bedding in certain unsheltered settings that can impair sleep,^29,30 some outdoor locations may be preferable and require less vigilance than noisy, crowded, or unsafe sheltered contexts.^2,16 Other unsheltered contexts that provide bedding, additional privacy, and protection from victimization (e.g., certain RVs and other vehicles) may be particularly conducive to sleep for TAY experiencing homelessness. Future research should investigate how specific social and built environment factors in unsheltered settings differentially influence sleep.

We did find evidence for our second hypothesis, that sleep environment safety would be significantly associated with sleep disturbance, which remained significant when adjusting for unsheltered status. The importance of a safe sleep environment is consistent with the literature on socioenvironmental sleep determinants,^30,31 and suggests that efforts to improve safety in sheltered and unsheltered contexts may help to address sleep deficiencies among TAY experiencing homelessness. In places like homeless shelters, this might involve changes to staffing, sleeping arrangements, or other program policies, while unsheltered TAY might benefit from having safe designated places to camp or park their vehicles at night.

It is notable that depression symptoms, severe food insecurity, and decreased age were also significantly associated with increased sleep disturbance. The strong association we found between depression and sleep disturbance aligns with extensive research documenting tight bidirectional relationships between these constructs,^4,8,32,33 and highlights the importance of screening for deficient sleep in concert with mental health treatment in clinical services for TAY. TAY whose sleep problems co-occur with psychiatric or medical issues may benefit from cognitive behavioral therapy for insomnia, which can effectively improve sleep and health comorbidity symptoms.³⁴ Addressing sleep problems that co-occur with milder impairments, however, may be possible with brief behavioral treatment for insomnia, which could be delivered by existing clinicians with minimal training.³⁵ The relationships we found between previous trauma and disturbed sleep in both samples imply that these efforts should be implemented using a trauma-informed approach.³⁶

Our finding that severe food insecurity was associated with over a 0.5-SD increase in sleep disturbance score was also consistent with population-level research,^10,37 but raises the question of whether addressing food insecurity might result in improved sleep for this population as well. While we did not expect decreased age to be associated with higher sleep disturbance ratings, one possibility is that older TAY may have been homeless for longer, leading them to be relatively more acclimated to “sleeping rough” than their younger peers.

This study is primarily limited by its cross-sectional design and lack of objective measures for deficient sleep and related outcomes. Additionally, our dichotomous indicator of current unsheltered status “most of the time” and single-item scale for perceived sleep environment safety “most of the time, currently” were not temporally aligned with the past-7-day sleep disturbance measure, and may not have adequately captured the specific environmental factors that are most closely tied to sleep disturbance in sheltered and unsheltered contexts. Finally, we did not find high-risk drinking frequency to be associated with sleep disturbance, but this may have been because our 30-day recall period needs to be better aligned with the prior-7-day sleep disturbance measure. Prospective intensive longitudinal sampling methods, such as ecological momentary assessment, may be a better measure to tease apart associations between sleep, food insecurity, and behavioral health.

Overall, findings from this study indicate that TAY who are homeless experience elevated rates of sleep disturbance regardless of their sheltered status, which may be driven by several factors including perceived sleep environment safety, depression, and food insecurity. More research is needed to understand not only targetable risk factors for deficient sleep among this population, but also to elucidate the potential consequences that chronic sleep loss may have on long-term social, housing, and health-related outcomes. In addition to the well-documented academic and vocational consequences of chronic sleep deficiency among young people generally,^4,33 sleep-related impairment among homeless adults has been specifically described as a barrier to completing necessary tasks to exit homelessness (e.g., engaging in work, school, and services).¹⁶ The long-term health impacts of chronic sleep loss, including increased morbidity and mortality from chronic illnesses such as depression and cardiometabolic disease,^4,38 are particularly concerning for TAY, who are in a critical developmental stage that is known to influence one’s trajectory for years to come.³⁹