Sleep Quality among Homeless-Experienced Older Adults: Exploratory Results from the HOPE HOME Study

Abstract

Background

Sleep is essential to health and affected by environmental and clinical factors. There is limited longitudinal research examining sleep quality in homeless older adults.

Objective

To examine the factors associated with poor sleep quality in a cohort of older adults in Oakland, California recruited while homeless using venue-based sampling and followed regardless of housing status.

Design

Longitudinal cohort study.

Participants

244 homeless-experienced adults aged ≥ 50 from the Health Outcomes in People Experiencing Homelessness in Older Middle Age (HOPE HOME) cohort.

Main Measures

We assessed sleep quality using the Pittsburgh Sleep Quality Index (PSQI). We captured variables via biannual questionnaires and clinical assessments.

Key Results

Our sample was predominantly men (71.3%), Black (82.8%), and had a median age of 58.0 years old (IQR 54.0, 61.0). Two-thirds of participants (67.2%) reported poor sleep during one or more study visits; sleep duration was the worst rated subdomain. In a multivariable model, having moderate-to-severe depressive symptoms (AOR 2.03, 95% CI 1.40–2.95), trouble remembering (AOR 1.56, 95% CI 1.11–2.19), fair or poor physical health (AOR 1.49, 95% CI 1.07–2.08), two or more chronic health conditions (AOR 1.76, 95% CI 1.18–2.62), any ADL impairment (AOR 1.85, 95% CI 1.36–2.52), and being lonely (AOR 1.55, 95% CI 1.13–2.12) were associated with increased odds of poor sleep quality. Having at least one confidant was associated with decreased odds of poor sleep (AOR 0.56, 95% CI 0.37–0.85). Current housing status was not significantly associated with poor sleep quality.

Conclusions

Homeless-experienced older adults have a high prevalence of poor sleep. We found that participants’ physical and mental health was related to poor sleep quality. Poor sleep continued when participants re-entered housing. Access to physical and mental healthcare, caregiving support, and programs that promote community may improve homeless-experienced older adults sleep quality, and therefore, their overall health.

Supplementary Information

The online version contains supplementary material available at 10.1007/s11606-023-08429-4.

INTRODUCTION

Homelessness is dynamic with individuals moving in and out. Some experience short, isolated episodes while others experience longer ones. Nearly a third of single adults experiencing homelessness are chronically homeless.¹ The homeless population in the United States is aging. Approximately a half of single adults experiencing homelessness are aged 50 and older.² Homeless-experienced (currently or recently homeless) adults experience accelerated aging. Homeless adults in their 50’s and 60’s have a prevalence of geriatric conditions similar to housed adults in their 70’s and 80’s.^{3, 4} Therefore, homeless-experienced adults are considered “older” at 50.

Sleep is an essential component of health; chronically poor sleep is associated with morbidity and increased all-cause mortality.^5–7 Several demographic (e.g., age), social (e.g., loneliness), clinical (e.g., depression, substance use), and environmental (e.g., noise) factors impact sleep. Certain clinical factors, including depression, have bidirectional associations with sleep.⁸ People experiencing homelessness face environmental barriers to good quality sleep. Those experiencing unsheltered homelessness sleep on hard surfaces and are exposed to inclement weather, noise, and light.⁹ People in congregate shelters sleep on cots or mattresses on the floor in crowded spaces with poor temperature control.

Previous studies of sleep in homeless adults used cross-sectional methods to examine aspects of sleep (i.e., duration, insomnia) in younger and mostly sheltered samples.^9–11 These found that homeless adults have shorter sleep duration and a higher prevalence of poor sleep than the general population.^{9, 10} Substance use, mental health symptoms (i.e., feeling depressed, anxious), and safety concerns were associated with poor sleep quality in younger homeless women.¹¹

Older age is associated with poor sleep quality stemming from multifactorial physiologic and psychosocial risk factors of aging (e.g., chronic diseases, social isolation).^12–18 Chronic physical and mental health conditions, substance use disorders, and limited social support are risk factors for poor sleep; they are more prevalent in homeless populations, compared to non-homeless.^19–21

In a cohort of homeless-experienced adults aged 50 and older where participants transitioned in and out of homelessness, we used longitudinal exploratory analyses to examine the prevalence of poor sleep quality and its associations with demographic, social, clinical, and environmental factors. We hypothesized that older homeless adults have a higher prevalence of poor sleep compared to older adults in the general community and to younger homeless adults. We hypothesized that clinical risk factors (e.g., functional impairment, depressive symptoms) and environmental risk factors (e.g., homeless versus housed, sheltered versus unsheltered) will be associated with poor sleep quality.

METHODS

Participants and Setting

Health Outcomes in People Experiencing Homelessness in Older Middle agE (HOPE HOME) is a longitudinal study of physical and mental health, life course events, and functional status among adults aged 50 and older in Oakland, California who were homeless at enrollment.²² From July 2013 to June 2014, we used venue-based sampling to recruit 350 participants (Wave 1) from five overnight homeless shelters, five low-cost meal programs, one recycling center, and places where unsheltered homeless individuals stayed.²² We constructed our sampling frame to approximate the source population and randomly selected potential participants at each recruitment site.^{22, 23} Between August 2017 and July 2018, we recruited an additional 100 participants (Wave 2) to replace those who died or were lost to follow-up using the same recruitment strategy.

Individuals were eligible if they were 50 years of age or older (53 or older in the Wave 2), homeless at baseline according to the Homeless Emergency Assistance and Rapid Transition to Housing (HEARTH) Act,²⁴ English-speaking, and able to provide informed consent using a teach-back method.²⁵ Trained study staff conducted in-depth, structured interviews at baseline and every six months thereafter, with brief monthly check-ins in between. Participants remained in the study regardless of housing status at follow-up. We added detailed sleep questions in 2019. We analyzed data from interviews we conducted every six months between 2019 and February 2022.

We compensated participants $25 for the screening and baseline interview, $5 for each monthly check-in, and $20 for each six-month follow-up interview. The institutional review board of the University of California, San Francisco approved the study. The HOPE HOME Community Advisory Board informed the study throughout the research process.

Measures

We used demographic information from cohort enrollment. Consistent with GEE methodology, we used data regarding the dependent variable (sleep) and independent variables from each follow-up interview in which they completed the sleep questions.

For the exploratory analysis of factors associated with poor sleep, we conceptualized independent variables as demographic/social, clinical, or environmental factors hypothesized to be associated with overall sleep quality.

Dependent Variable

Sleep quality

To assess self-reported sleep quality over the past month, we administered a modified 10-item Pittsburgh Sleep Quality Index (PSQI) composed of seven subdomains scored on a scale of 0 (“no difficulty”) to 3 (“severe difficulty”).²⁶ We summed subdomain scores to yield a global PSQI score, with higher scores indicating poorer sleep quality. We dichotomized sleep quality as “good” (global PSQI ≤ 5) or “poor” (global PSQI ≥ 6), using a validated cutoff.²⁶ For descriptive purposes, we included scores on each of the sleep subdomains: subjective quality, latency, duration, efficiency [ratio of total sleep time to time in “bed”], sleep disturbance, use of sleep medication, and daytime dysfunction [daytime sleepiness; the inability to stay awake and alert during the major waking periods of the day] (Table 3).

Table 3.

Sleep Quality in Homeless-Experienced Older Adults at their First Sleep Interview

Sleep Subdomaina	Total (N = 244) Mean (SD)	Good sleep quality (N = 80) Mean (SD)	Poor sleep qualityb (N = 164) Mean (SD)	p-value
Subjective sleep quality	1.0 (0.8)	0.5 (0.5)	1.2 (0.8)	< 0.001
Sleep latency	1.1 (1.0)	0.4 (0.5)	1.4 (1.1)	< 0.001
Sleep duration	1.6 (1.3)	0.7 (0.9)	2.1 (1.1)	< 0.001
Sleep efficiency	1.1 (1.2)	0.4 (0.7)	1.5 (1.2)	< 0.001
Sleep disturbance	1.0 (0.8)	0.4 (0.5)	1.3 (0.7)	< 0.001
Use of sleep medication	0.8 (1.3)	0.1 (0.5)	1.1 (1.4)	< 0.001
Daytime dysfunction	0.8 (1.0)	0.3 (0.5)	1.0 (1.1)	< 0.001

^aEach subdomain is scored 0–3 (higher = worse sleep quality)

^bPoor sleep defined as Pittsburgh Sleep Quality Index score ≥ 6

Independent Variables

Demographics

We collected demographic information including age, sex, and race/ethnicity.

Social

Social Support

We asked participants how many close friends or relatives they had in whom they could confide (0 vs ≥ 1).^{27, 28}

Loneliness

We used the short UCLA Loneliness Scale, which asked participants how often they felt three different components of subjective loneliness on a scale of 1 (“Hardly Ever”) to 3 (“Often”).²⁹ We considered someone with a score ≥ 6 to be lonely (range: 3–9).³⁰

Clinical

Health Status

We assessed self-reported health,³¹ categorizing responses as fair or poor versus good, very good or excellent. To assess chronic health conditions, we used National Health Interview Survey (NHIS) questions to assess if a healthcare provider had diagnosed participants with hypertension, lung disease (COPD or asthma), coronary artery disease, congestive heart failure, arthritis, or liver disease (including hepatitis) or cirrhosis.³² We assessed urinary incontinence during the previous six months by asking participants how often they leaked urine.³³ To assess functional status, we asked participants if they had difficulty performing any of the five activities of daily living (ADLs).³⁴ We measured if participants had difficulty performing any of the six instrumental activities of daily living (IADLs) using the Brief Instrumental Functioning Scale (BIFS), which has been validated in homeless populations.³⁵ We dichotomized both variables as zero versus ≥ 1 difficulties. To evaluate cognitive impairment, we used the Modified Mini-Mental State Examination (3MS).³⁶ We defined cognitive impairment as a score below the 7th percentile (i.e., 1.5 standard deviations below a reference cohort mean) or inability to complete the assessment. To evaluate traumatic brain injury (TBI) over the lifetime, we asked participants whether they had ever experienced a head injury resulting in loss of consciousness or hospitalization.³⁷

Mental Health

We adapted mental health questions from the Addiction Severity Index (ASI) to assess whether participants had experienced severe anxiety in the prior 30 days or had hallucinations or trouble remembering in the past six months.³⁸ We defined moderate-to-severe depressive symptoms as scores ≥ 22 on the Center for Epidemiologic Studies Depression Scale (CES-D).^{39, 40} Consistent with previous literature, we considered a score of ≥ 4 on the Primary Care Post-Traumatic Stress Disorder (PC-PTSD) Screen to indicate PTSD.⁴¹

Substance Use

We used questions from the California Tobacco Survey to classify participants as current smokers.⁴² We administered the Alcohol Use Disorders Identification Test (AUDIT) to assess severity of alcohol use symptoms in the previous six months and used a validated score of ≥ 8 to indicate problematic usage.^{43, 44} We used the World Health Organization’s Alcohol, Smoking and Substance Involvement Screening Test (WHO-ASSIST) to assess participants’ use of cannabis, cocaine, amphetamines, and heroin/non-prescribed opioids in the past six months.⁴⁵ We classified scores of ≥ 4 for any illicit drug as problematic use.

Environmental

Residential History

For descriptive purposes, we used a follow-back residential calendar to describe participants’ self-reported living environments over the previous six months.⁴⁶

Duration of Adult Homelessness

We calculated the total number of years participants had spent homeless since age 18.

Housing Status

We considered participants homeless at the time of their first sleep interview (starting in 2019) if they met federal HEARTH criteria for homelessness.²⁴ We categorized housing status in the 14 days prior to each sleep interview into three categories: homeless, housed/not homeless, and living in a skilled nursing facility (SNF). We defined homeless as staying outdoors, in a place not meant for human habitation, an emergency shelter, or staying in housing where participants needed to leave within two weeks without another place to go.²⁴ Since we hypothesized that living inside a non-congregate building would be associated with better sleep quality, we classified those staying long-term (> 2 weeks) in transitional housing or private hotels rooms in the housed/not homeless group.

Shelter status

We asked participants whether they experienced unsheltered homelessness (i.e., sleeping in a place not meant for human habitation, such as cars, parks, sidewalks, abandoned buildings) in the previous six months and calculated the number of nights they reported spending unsheltered. We asked participants whether they stayed any nights at a homeless shelter in the previous six months and calculated the number of nights they did so.

Victimization

We asked participants if they had experienced 1) physical or sexual abuse or 2) robbery or theft in the prior six months and created two corresponding variables.⁴⁷

Analysis

Our primary dependent variable was poor sleep quality (PSQI score ≥6). We assessed univariable associations between sleep quality at any complete sleep interview and a priori independent variables from baseline (for time constant) or the same completed sleep interview (for time varying) using Chi-squared tests and t-tests. For multivariable analyses, we clustered participants by ID to account for repeated measures with robust Huber–White cluster-adjusted standard errors. We report odds ratios resulting from a GEE analysis with a binominal distribution and a logit link and an unstructured covariance structure. We performed all analyses with SAS version 9.4.⁴⁸ We evaluated the association between selected independent variables and poor sleep quality using the GENMOD procedure in SAS. We included both fixed and time-varying covariates in the adjusted analysis if they were significant at the 0.2 level in the bivariate analysis or selected a priori based on the literature. Then, we reduced the model using backward elimination retaining variables with p values <.05 in the final model. Missing data were minimal (i.e., <5%) for all outcomes and independent variables with the following exceptions: one person was missing self-rated health and four people were missing loneliness scores. We omitted person-visits with missing data using listwise deletion for each analysis, depending on the included variables. Our primary model examined the relationship between select variables and the odds of poor sleep quality. We completed an additional analysis to examine the relationship between types of homelessness (sheltered vs. unsheltered) and odds of poor sleep quality.

RESULTS

Sample Characteristics

This multivariable analysis includes 947 distinct observations from 244 participants who completed at least one sleep interview (range: 1–7). (Figs. 1 and 2) Our participants were predominantly men (71.3%), Black (82.8%), and had a median age of 58.0 years old (IQR 54.0, 61.0) (Table 1). Approximately half of the sample described their health as fair or poor. Several (43.0%) reported difficulty performing at least one Activity of Daily Living (ADL). Forty percent met the HEARTH criteria for homelessness at the time of their first sleep interview. In the six months prior to their first sleep interview, 38.5% spent at least one night unsheltered (M_nights = 50.0, SD 72.5) and 17.2% spent at least one night in a shelter (M_nights = 12.8, SD 38.6) (Table 1). The median duration of follow-up for the cohort was 6.8 years (IQR: 4.0, 7.9).

Figure 1.

Recruitment of HOPE HOME Sample. Two-hundred and forty-four participants completed at least one sleep interview.

Figure 2.

Frequency of sleep interviews among HOPE HOME participants (n = 244).

Table 1.

Characteristics of Homeless-Experienced Older Adults According to Sleep Quality (n = 244)

Characteristic	Total N/Median (%/IQR)	Good sleep quality (N = 80) N/Median (%/IQR)	Poor sleep quality (N = 164)a N/Median (%/IQR)	p-value
Demographics
Age, in years	58.0 (54.0, 61.0)	59.0 (54.0, 61.0)	58.0 (55.0, 61.0)	0.97
Male	174 (71.3%)	61 (76.3%)	113 (68.9%)	0.23
Black	202 (82.8%)	71 (88.8%)	131 (79.9%)	0.08
Social Support
≥ 1 confidant	188 (77.0%)	66 (82.5%)	122 (74.4%)	0.16
Loneliness
Lonelyb	83 (34.0%)	16 (20.0%)	67 (40.9%)	0.001
Health Status
Self-reported fair/poor healthc	120 (49.2%)	30 (37.5%)	90 (54.9%)	0.009
Chronic Conditionsd
Hypertension	167 (68.4%)	53 (66.3%)	114 (69.5%)	0.61
Lung diseasee	96 (39.3%)	23 (28.8%)	73 (44.5%)	0.02
Coronary artery disease	17 (7.0%)	5 (6.3%)	12 (7.3%)	0.76
Congestive heart failure	38 (15.6%)	15 (18.8%)	23 (14.0%)	0.34
Arthritis	149 (61.1%)	45 (56.3%)	104 (63.4%)	0.28
Liver disease or cirrhosis	69 (28.3%)	19 (23.8%)	50 (30.5%)	0.27
Urinary incontinencef	103 (42.2%)	25 (31.3%)	78 (47.6%)	0.01
ADL impairmentg	105 (43.0%)	17 (21.3%)	88 (53.7%)	< 0.001
IADL impairmenth	69 (28.3%)	12 (15.0%)	57 (34.8%)	0.001
Cognitive impairmenti	20 (8.2%)	6 (7.5%)	14 (8.5%)	0.64
Traumatic Brain Injuryk	11 (4.5%)	1 (1.3%)	10 (6.1%)	0.09
Mental Health
Anxiety, past 30 days	81 (33.2%)	13 (16.3%)	68 (41.5%)	< 0.001
Moderate to severe depressive symptoms, past weekj	48 (19.7%)	3 (3.8%)	45 (27.4%)	< 0.001
Hallucinations, past six months	20 (8.2%)	3 (3.8%)	17 (10.4%)	0.08
Trouble remembering, past six months	65 (26.6%)	8 (10.0%)	57 (34.8%)	< 0.001
PTSD symptomsj	47 (19%)	7 (8.8%)	40 (24.4%)	0.004
Problematic Substance Usel
Current smoker	166 (68.0%)	49 (61.3%)	117 (71.3%)	0.11
Alcohol	31 (12.7%)	7 (8.8%)	24 (14.6%)	0.20
Cannabis	114 (46.7%)	35 (43.8%)	79 (48.2%)	0.57
Cocaine	51 (20.9%)	16 (20.0%)	35 (21.3%)	0.83
Amphetamines	22 (9.0%)	3 (3.8%)	19 (11.6%)	0.04
Heroin and non-prescribed opioids	10 (4.1%)	4 (5.0%)	6 (3.7%)	0.63
Residential History, past 6 months
Unsheltered (any nights)m	94 (38.5%)	24 (30.0%)	70 (42.7%)	0.06
Homeless shelter (any nights)	42 (17.2%)	12 (15.0%)	30 (18.3%)	0.52
Duration of Adult Homelessness
Duration of adult homelessness in years, median (range)	5.0 (9.2)	3.6 (9.2)	5.0 (9.6)	0.49
Housing Status
Homeless at time of interviewm	97 (39.8%)	28 (35.0%)	69 (42.1%)	0.29
Victimization, past 6 months
Physical or Sexual Abuse	33 (13.5%)	7 (8.8%)	26 (15.9%)	0.13
Robbery/Theft	68 (27.9%)	16 (20.0%)	52 (31.7%)	0.06

P-values are based on Chi-squared tests for categorical variables and t-tests for continuous variables

^aPittsburgh Sleep Quality Index ≥ 6

^bScore of ≥ 6 on the 3-item UCLA Loneliness Scale

^cSelf-rated using Ware et al. 1-item health screen

^dSelf-report of ever receiving diagnosis from a physician

^eDefined as chronic obstructive pulmonary disease (COPD) or asthma diagnosis

^fDefined as “leaked urine, even a small amount” in the past 6 months

^gSelf-reported difficulty performing one or more ADLs

^hSelf-reported difficulty performing one or more IADLs as assessed by the Brief Instrumental Functioning Scale

ⁱDefined as a score below the 7^th percentile OR unable to complete the Modified Mini-Mental State Examination (3MS)

^jCenter for Epidemiologic Studies Depression Scale score ≥ 22, Primary Care PTSD Screen score ≥ 3

^kDefined as head trauma resulting in loss of consciousness or hospitalization

^lAlcohol Use Disorders Identification Test (AUDIT) score ≥ 8 for alcohol, Alcohol Smoking Substance Involvement Screening Test (ASSIST) score ≥ 4 for illicit drugs, California Tobacco Survey for smoking

^mDefined as sleeping in a place not meant for human habitation, such as cars, parks, sidewalks, abandoned buildings (on the street)

Poor Sleep Quality Prevalence

Approximately two-thirds of the sample (67.2%) met criteria for poor sleep quality at their first sleep interview (Table 1). The median PSQI score was 7.0 (range: 0.0–20.0, IQR 3.0, 10.0) (Table 2). Forty-two percent (42.2) of people with poor sleep scores met the HEARTH criteria for homelessness at the time of their first sleep interview (Table 2).

Table 2.

Median PSQI Scores Overall and by Housing Status at First Interview and Each Sleep Observation

First Sleep Interview
	Median PSQI Score (IQR)
All Participants (N = 244)	7.0 (3.0–10.0)
Housing Status At Time of First Interview
Homeless (N = 95)	7.0 (4.0–9.0)
Housed/Not Homeless (N = 147)	7.0 (3.0–10.0)
Skilled Nursing Facility (N = 2)	9.0 (7.0–11.0)
Overall
All Observations (N = 947)	7.0 (4.0–10.0)
Housing Status At Time of Sleep Observation
Homeless (N = 280)	6.0 (3.0–9.0)
Housed/Not Homeless (N = 660)	6.0 (3.0–10.0)
Skilled Nursing Facility (N = 7)	7.0 (6.0–11.0)

Among participants, the worst sleep scores occurred in the duration (M 1.6, SD 1.3), efficiency (M 1.1, SD 1.2), and latency (M 1.1, SD 1.0) subdomains (Table 3). Participants with poor sleep quality scored worse across all sleep subdomains compared with participants with good sleep quality (p < 0.001) (Table 3). On average, people with poor sleep reported mild-to-moderate problems across all sleep subdomains with worst scores in duration (M 2.1, SD 1.1), efficiency (M 1.5, SD 1.2) and latency (M 1.4, SD 1.1).

Factors Associated with Poor Sleep Quality

In a multivariable model, having moderate-to-severe depressive symptoms (AOR 2.03, 95% CI 1.40–2.95), trouble remembering (AOR 1.56, 95% CI 1.11–2.19), fair or poor physical health (AOR 1.49, 95% CI 1.07–2.08), two or more chronic health conditions (AOR 1.76, 95% CI 1.18–2.62), an ADL impairment (AOR 1.85, 95% CI 1.36–2.52), and being lonely (AOR 1.55, 95% CI 1.13–2.12) were associated with increased odds of poor sleep quality (Table 4). Having at least one confidant was associated with decreased odds of poor sleep (AOR 0.56, 95% CI 0.37–0.85). In the final multivariable model, we did not find an association between housing status (i.e., being homeless or living in a SNF, compared to being housed) or substance use (alcohol or drug use) and poor sleep quality. In additional analyses, we found that unsheltered homelessness (compared to sheltered homelessness) was not associated with poor sleep quality (Supplemental Table 1). We did not find evidence of multicollinearity.

Table 4.

Multivariate Models of Factors Associated with Poor Sleep Quality^a

Characteristic	Unadjusted odds ratio for poor sleep quality OR (95% CI)	Adjusted odds ratio for poor sleep quality AOR (95% CI)
Demographics
Male	0.83 (0.53—1.30)
Age	1.01 (0.97 – 1.06)
Black race	0.54 (0.30—0.99)
Social Support
≥ 1 confidant	0.60 (0.41 – 0.88)	0.56 (0.37 – 0.85)
Loneliness
Lonelyb	1.87 (1.41—2.49)	1.55 (1.13 – 2.12)
Health Status
Fair/poor healthc	1.84 (1.37—2.48)	1.49 (1.07 – 2.08)
Two or more chronic conditionsd	2.09 (1.42 – 3.09)	1.76 (1.18 – 2.62)
Lung diseased,e	1.85 (1.25 – 2.76)
Urinary incontinencef	1.53 (1.11—2.10)
ADL impairmentg	2.45 (1.85—3.24)	1.85 (1.36 – 2.52)
IADL impairmenth	1.49 (1.11—2.00)
Mental Health
Anxiety, past 30 days	1.90 (1.42—2.54)
Moderate to severe depressive symptoms, past weekj	2.84 (2.07—3.91)	2.03 (1.40 – 2.95)
Hallucinations in the past six monthsj	1.68 (1.20 – 2.36)
Trouble remembering in the past six months	2.06 (1.57 – 2.71)	1.56 (1.11 – 2.19)
Positive PTSD screenj	1.99 (1.42- 2.79)
Traumatic Brain Injuryk	4.75 (1.15—19.60)
Problematic Substance Usel
Alcohol	1.82 (1.18 – 2.80)
Amphetamines	1.03 (1.00 – 1.06)
Housing Status At Time of Interview
Housed/Not Homeless	Referent
Homeless	1.42 (0.98–2.07)
Skilled Nursing Facility	1.60 (0.29–8.88)
Victimization, past 6 months
Robbery/Theft	1.70 (1.17 – 2.47)

We included independent variables with univariable P-values of ≤ 0.20 in a multivariable model and used backwards elimination to reduce the model, retaining variables with P ≤ 0.05 (bolded) in the final model

^aPoor sleep quality defined as Pittsburgh Sleep Quality Index ≥ 6

^bDefined as a score of ≥ 6 on the 3-item UCLA Loneliness Scale

^cSelf-rated using Ware et al. 1-item health screen

^dSelf-report of ever receiving diagnosis from a physician

^eDefined as chronic obstructive pulmonary disease (COPD) or asthma diagnosis

^fDefined as “leaked urine, even a small amount” in the past 6 months

^gSelf-reported difficulty performing one or more ADLs

^hSelf-reported difficulty performing one or more IADLs as assessed by the Brief Instrumental Functioning Scale

ⁱDefined as a score below the 7^th percentile OR unable to complete the Modified Mini-Mental State Examination (3MS)

^jCenter for Epidemiologic Studies Depression Scale score ≥ 22, yes to ASI question about hallucinations in prior six months, Primary Care PTSD Screen score ≥ 3

^kDefined as head trauma resulting in loss of consciousness or hospitalization

^lAlcohol Smoking Substance Involvement Screening Test (ASSIST) score ≥ 4, Alcohol Use Disorders Identification Test (AUDIT) score ≥ 8

DISCUSSION

In this exploratory study of adults with a median age of 58 who were homeless at study enrollment, we found that participants had higher prevalence of poor sleep and worse PSQI scores than in studies of housed adults 20 years older.^{49, 50} In our study, two-thirds reported poor sleep quality, compared to 44% of men and 53% of women aged 70 and older in the general population.^{49, 50} Our participants had higher prevalence of poor sleep than similarly aged adults in the general population, where 41% of men with a median age of 55 and 50% of women with a median age of 59 had poor sleep quality.^{51, 52} While aging is associated with worse outcomes in nearly every sleep subdomain, our findings support that age-related decrements in sleep quality occur at an earlier age in homeless adults.¹⁴ This may be due to co-morbid conditions prevalent in older homeless adults and associated with poor sleep, such as ADL impairments and depression or due to the difficult environmental conditions of homelessness.^{3, 53, 54}

Fair or poor self-reported health, multiple chronic diseases, poor functional status (ADL impairments), mental and cognitive health (depressive symptoms and trouble remembering) were associated with increased odds of poor sleep quality. Poor health status and chronic diseases are risk factors for poor sleep.^{6, 13, 16, 55, 56} There is an established bidirectional relationship between poor sleep quality and depression.^{18, 57–60} Poor sleep may be a prodromal symptom of depression, people who are predisposed to poor sleep may also be predisposed to depression, or poor sleep may alter cognition and emotional regulation, thus increasing depression risk.⁵⁶ ADL impairments, a measure of functional status, could signal functional decline, which corresponds to worsening sleep, or poor sleep could lead to functional decline.^61–66

Similar to previous research, we found that social support was a protective factor for sleep quality.^{17, 67} This may be due to evolutionary reasons; individuals with social support may feel safer and sleep better.^{20, 68} Loneliness was associated with increased odds of poor sleep, supporting prior research.^{58–60, 69} Stress, associated with feeling lonely and experiencing homelessness, may mediate this relationship.^{70, 71}

We did not find an association between ongoing homelessness (as opposed to regaining housing) or substance use and poor sleep quality. Decrements in sleep quality associated with homelessness may be lasting.^{10, 72} We posit some explanations for this. Some regained housing but remained in overcrowded situations with inadequate bedding, suggesting ongoing environmental impacts on sleep. Others continued to experience anxiety related to housing stability, which may contribute to ongoing poor sleep.^{20, 73, 74} The lack of association between problematic alcohol or substance use and poor sleep contradicts findings in the general population.^75–77 This may be due to the high prevalence of problematic substance use or other causes of poor sleep.

Participants reported the most impairments in sleep duration, while the general older population report worst scores in sleep disturbances.^{49, 50} This may relate to the environmental conditions of homelessness which limit access to sleep, including noise and light pollution, hypervigilance of threats to safety, lack of convenient restrooms, exposure to cold and rain, and mandatory early awakenings in shelters. These environmental challenges appear to negatively affect sleep in both sheltered and unsheltered environments.

Our study has several limitations. We conducted our study in one city, which limits generalizability. It is possible that poor sleep quality causes or worsens the associated physical and mental health conditions, that these conditions cause poor sleep, or that the relationship is bidirectional. We measured symptoms of mental health problems (i.e., depressive symptoms) not clinical diagnoses (i.e., depression).^{53, 78} However, our study used validated scales (e.g., CES-D) for all variables.^{40, 79} Finally, this study was exploratory and reports multiple adjusted effects in the same model, which may limit interpretative validity. Some strengths of our study included our use of longitudinal data, our inclusion of measures of overall sleep quality, examining sleep in a variety of housing situations, and our focus on older adults, a group at high risk of poor sleep.

Homeless-experienced older adults have a high prevalence of poor sleep quality. Ongoing lack of quality sleep, even after regaining housing, is an underappreciated negative impact of homelessness among older adults. Poor sleep may contribute to the overall poor health status of people experiencing homelessness. Access to physical and mental healthcare, caregiving support, and programs that promote social connection may improve homeless-experienced older adults’ sleep quality and overall health. Preventing homelessness may protect against potentially long-lasting poor-quality sleep.

Supplementary Information

Below is the link to the electronic supplementary material.

Author’s Contributions

The authors would like to thank our participants who make this work possible. We gratefully acknowledge our colleagues Kenneth Perez, John Weeks, Stephen King, Celeste Enriquez, and Stacy Castellanos for their invaluable contributions to the HOPE HOME study. The authors also want to thank the staff at Lifelong Medical Care, St. Mary’s Center and the HOPE HOME Community Advisory Board for their guidance and partnership.

Funding

This study was funded by grants from the National Institute on Aging at the National Institutes of Health (R01AG041860, K24AG046372) awarded to MK. These funding sources had no role in the preparation, review, or approval of the manuscript.

Declarations:

Conflict of Interest:

All authors declare that they have no conflict of interest.