This cohort study aims to identify the prevalence and factors associated with mortality in a cohort of homeless adults 50 years and older.
Key Points
Question
What are the factors associated with mortality among adults 50 years and older experiencing homelessness?
Findings
In this cohort study of 450 homeless adults 50 years and older, 26% died over a median follow-up of 55 months, with a median age at death of 64.6 years. Factors associated with death included a first episode of homelessness in late life and homelessness or institutionalization at any follow-up compared with being housed.
Meaning
Factors associated with mortality among homeless older adults included late-life homelessness and ongoing homelessness, pointing to the urgent need for policy approaches to prevent and end homelessness among older adults in the US.
Abstract
Importance
The population of homeless older adults is growing and experiences premature mortality. Little is known about factors associated with mortality among homeless older adults.
Objective
To identify the prevalence and factors associated with mortality in a cohort of homeless adults 50 years and older.
Design, Setting, and Participants
In this prospective cohort study (Health Outcomes in People Experiencing Homelessness in Older Middle Age [HOPE HOME]), 450 adults 50 years and older who were homeless at baseline were recruited via venue-based sampling in Oakland, California. Enrollment occurred in 2 phases, from July 2013 to June 2014 and from August 2017 to July 2018, and participants were interviewed at 6-month intervals.
Exposures
Baseline and time-varying characteristics, including sociodemographic factors, social support, housing status, incarceration history, chronic medical conditions, substance use, and mental health problems.
Main Outcomes and Measures
Mortality through December 31, 2021, based on state and local vital records information from contacts and death certificates. All-cause mortality rates were compared with those in the general population from 2014 to 2019 using age-specific standardized mortality ratios with 95% CIs.
Results
Of the 450 included participants, median (IQR) age at baseline was 58.1 (54.5-61.6) years, 107 (24%) were women, and 360 (80%) were Black. Over a median (IQR) follow-up of 55 (38-93) months, 117 (26%) participants died. Median (IQR) age at death was 64.6 (60.3-67.5) years. In multivariable analyses, characteristics associated with mortality included a first episode of homelessness at 50 years and older (adjusted hazard ratio [aHR], 1.62; 95% CI, 1.13-2.32), homelessness (aHR, 1.82; 95% CI, 1.23-2.68) or institutionalization (aHR, 6.36; 95% CI, 3.42-11.82) at any follow-up compared with being housed, fair or poor self-rated health (aHR, 1.64; 95% CI, 1.13-2.40), and diabetes (aHR, 1.55; 95% CI, 1.06-2.26). Demographic characteristics, substance use problems, and mental health problems were not independently associated. All-cause standardized mortality was 3.5 times higher (95% CI, 2.5-4.4) compared with adults in Oakland. The most common causes of death were heart disease (n = 17 [14.5%]), cancer (n = 17 [14.5%]), and drug overdose (n = 14 [12.0%]).
Conclusions and Relevance
The cohort study found that premature mortality was common among homeless older adults and associated factors included late-life homelessness and ongoing homelessness. There is an urgent need for policy approaches to prevent and end homelessness among older adults in the US.
Introduction
The median age of the homeless population is increasing. More than one-third of single homeless adults are 50 years and older compared with 11% in 1990.1,2 Homeless people experience accelerated aging, including premature onset of chronic medical conditions, functional and cognitive impairments, and mortality.3,4,5,6,7 A growing literature among other historically marginalized populations (eg, persons with severe mental illness, racial and ethnic minority groups) shows that adverse life-course exposures (eg, discrimination, adverse childhood experiences) are associated with premature aging and mortality.8,9 People experiencing homelessness have these negative experiences, along with a high prevalence of comorbidities and behaviors associated with premature mortality.4,10 Experiences while homeless (eg, exposures, stress, limited health care access) may contribute directly to premature mortality.4,11 Little is known about how life-course and time-varying exposures and co-occurring conditions affect mortality risk among homeless older adults.
Few prior studies have examined factors associated with mortality in homeless persons; those that have are retrospective and rely on medical records.6,12 These studies include only individuals who engage with care, and they lack standardized assessments of life-course or time-varying factors, including housing status. A recent cross-sectional study examined trends in mortality among homeless persons in San Francisco before vs during the COVID-19 pandemic but did not examine factors associated with mortality.13 Understanding how time-varying factors affect mortality risk in homeless older adults is necessary to identify high-risk individuals, inform interventions to prevent premature mortality, and target resources.
In this prospective cohort study of adults 50 years and older who were homeless at study entry, we examined the prevalence, associated factors, and causes of mortality. We also compared mortality rates with those of the general population.
Methods
Design Overview
We conducted a prospective cohort study of 450 homeless adults 50 years and older recruited via venue-based sampling in Oakland, California (Health Outcomes in People Experiencing Homelessness in Older Middle Age [HOPE HOME]). We interviewed participants at baseline and every 6 months. The institutional review board of the University of California, San Francisco approved the study, and participants provided written informed consent. We followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines for cohort studies.14
Sample and Recruitment
We sampled homeless persons using a purposive venue-based approach to approximate the source population.4,15 Sampling locations included low-cost meal programs and overnight shelters, a recycling center, and places where unsheltered people stayed. We set recruitment goals for each location based on estimates of the number of unique persons who visited a site or were unsheltered annually. We invited individuals who met a brief eligibility screen to participate in an enrollment interview.
Enrollment occurred in 2 phases: (1) 350 participants enrolled from July 2013 through June 2014 and (2) 100 participants from August 2017 through June 2018. Study staff conducted interviews at nonprofit community-based organizations in Oakland. A community advisory board guided all study protocols. Eligibility criteria included being 50 years and older for the initial cohort and 53 years and older for the second; being English speaking; current homelessness based on the Homeless Emergency Assistance and Rapid Transition to Housing Act, defined as lacking a regular nighttime residence (including staying in emergency shelter or a place not meant for human habitation) or staying temporarily in an institution, losing housing within 14 days, or fleeing interpersonal violence, all without another place to stay16; and ability to provide informed consent. We excluded non-English speakers because during the recruitment periods, few non-English speakers were homeless and 50 years and older in Oakland.
We conducted follow-up interviews every 6 months for a maximum of 96 months for the first cohort and 48 months for the second. Participants remained in the study regardless of housing status. We interviewed participants at a community-based field site or, if the participant preferred, where they were staying (eg, encampment, home). To maximize follow-up, we collected detailed contact information at baseline and conducted monthly check-ins by telephone or in person. Participants gave permission for us to contact emergency contacts if they missed 2 check-ins. Participants received $25 incentives for enrollment interviews, $20 for 6-month follow-ups, and $5 for monthly check-ins.
Measures
Mortality
The primary outcome was mortality, assessed through December 31, 2021. To limit missed outcomes, we used multiple methods to assess mortality. We queried local vital statistics offices quarterly for participants who missed check-ins who we could not otherwise confirm were alive; we contacted local coroner’s offices to review unidentified deaths using photographs. We received annual California state death records and matched participants using names and birth dates. We queried emergency contacts and searched social media and online obituaries. We requested death records for suspected deaths. Vital records departments in some states require that a relative or legal surrogate request death certificates; for individuals who died in these states (n = 3), we used date of death provided by emergency contacts or other sources. We used information from death certificates to classify causes of death using International Statistical Classification of Diseases and Related Health Problems, Tenth Revision underlying cause of death codes.17 We defined drug overdoses as drug-poisoning deaths that were unintentional or of undetermined intent.5
Baseline Measures
Measures assessed at baseline included sociodemographics, housing, substance use, mental health problems, and incarceration. Sociodemographics included self-reported age, gender, race and ethnicity (Black, Latino, White, multiracial/other [including Asian]), marital/partner status, highest level of education, and usual occupation.18 Participants reported their age when they first experienced adult homelessness, defined as experiencing homelessness for 1 or more nights at 18 years or older.
We assessed alcohol and drug use during 4 life-course periods (<18, 18-25, 26-49, and ≥50 years old).19,20 We defined a history of regular alcohol and drug use as drinking to get drunk 3 or more times weekly and using drugs (cocaine, amphetamines, or opioids) 3 or more times weekly.19
We adapted the lifetime mental health measures from the Addiction Severity Index19 to ask if participants had experienced the following problems and at what age they first occurred: serious anxiety, depression, difficulty controlling violent behavior, or hallucinations unrelated to substance use; suicide attempt; or prescription of medication for psychiatric problems.19 Participants reported if they had been hospitalized for mental health problems and whether they had served time in prison or jail.
Time-Varying Measures
We also assessed measures at each 6-month follow-up (time varying). Time-varying measures included housing, social support, health status, and incarceration. We assessed housing using a follow-back residential calendar asking participants to report where they stayed each night during the prior 6 months.21 For participants who missed an interview, we determined housing status using information from contacts and other sources. We categorized housing based on the participant’s location on the interview date as homeless (Homeless Emergency Assistance and Rapid Transition to Housing Act criteria), housed (eg, apartment, transitional housing, medical facility), or staying in an institution (skilled nursing facility or ≥3 months in jail/prison). Participants reported number of close confidants, defined as anyone in whom they could confide (0, 1-5, ≥6).22
We assessed self-rated health (fair or poor vs good, very good, or excellent). Participants reported if a health care professional had ever told them that they had hypertension, cardiac disease, congestive heart failure, stroke or transient ischemic attack, diabetes, chronic lung disease, liver disease, kidney disease, HIV/AIDS, and cancer; we assessed new diagnoses at each follow-up.23 We calculated body mass index using measured height and weight. Participants reported if they had difficulty performing activities of daily living (bathing, dressing, eating, transferring, toileting)24 and instrumental activities of daily living (taking transportation, managing medications, managing money, applying for benefits, setting up a job interview, finding a lawyer)25; we defined impairment as reporting difficulty with 1 or more tasks. We administered the Modified Mini-Mental State Examination and defined cognitive impairment as a score lower than the seventh percentile (SD, 1.5 below a reference cohort mean) or inability to complete the assessment.26 Participants reported any falls in the prior 6 months. We administered the Short Physical Performance Battery and defined reduced performance as a score of 10 or lower (range, 0-12).27
We assessed self-reported binge drinking, defined as drinking 6 or more alcoholic beverages on 1 occasion monthly or more often28 and problematic drug use, defined as a World Health Organization Alcohol, Smoking and Substance Involvement Screening Test score of 4 or higher (range, 0-39) for cocaine, amphetamines, or opioids.29 We assessed smoking (current, former, never).30
We assessed mental health problems using the Addiction Severity Index.19 Participants reported if they had a psychiatric hospitalization during the prior 6 months.
Statistical Analyses
We used descriptive statistics to analyze participant characteristics and causes of death. To examine associations of candidate factors with mortality, we used Cox regression models. Factors included variables associated with mortality in the general population and those prevalent in homeless populations that we hypothesized could be associated with mortality (Table 1). We obtained values of time-varying covariates from the visit at the beginning of each interval. We imputed time-varying measures for missed visits by carrying forward the last observation. For decedents, we carried forward measures to month of death. Of 4142 visits in the analysis, we carried forward data for 445 visits (10.7%). For surviving participants, we censored follow-up 12 months after the last interview to reduce potential misclassification and bias. We estimated the multivariable model using an iterative process using backward selection in which all variables with bivariate type III P < .20 were entered and removed one at a time. The final parsimonious model included variables with a P < .05 after adjustment. We verified that hazards were proportional for time-varying covariates by testing the interaction of visit time with each covariate and using episode-splitting techniques. We analyzed Schoenfeld residuals to assess the proportionality assumption and evaluated goodness of fit using the likelihood ratio test.
Table 1. Baseline Characteristics of the Cohort.
| Characteristic | Participants, No. (%)a | ||
|---|---|---|---|
| All (n = 450) | Died (n = 117) | Did not die (n = 333) | |
| Sociodemographics | |||
| Age, median (IQR), y | 58.1 (54.5-61.6) | 60.0 (55.5-63.8) | 57.5 (54.1-61.0) |
| Gender | |||
| Female (cisgender) | 107 (24) | 24 (21) | 83 (25) |
| Male (cisgender) | 338 (75) | 92 (79) | 246 (74) |
| Transgender | 4 (1) | 0 | 4 (1) |
| Race and ethnicity | |||
| Black | 359 (80) | 91 (78) | 268 (80) |
| Latino | 21 (5) | 5 (4) | 16 (5) |
| White | 48 (11) | 17 (15) | 31 (9) |
| Multiracial/otherb | 21 (5) | 3 (3) | 18 (5) |
| Marital status | |||
| Married/partnered | 50 (12) | 11 (11) | 39 (13) |
| Divorced/separated | 163 (36) | 41 (35) | 122 (37) |
| Widowed | 43 (11) | 12 (12) | 31 (10) |
| Never married | 151 (37) | 35 (35) | 116 (38) |
| Educational attainment | |||
| More than high school education | 240 (54) | 60 (52) | 180 (55) |
| High school education or GED | 87 (20) | 25 (22) | 62 (19) |
| Less than high school education or GED | 118 (27) | 31 (27) | 87 (26) |
| Usual occupation | |||
| Higher executives, business managers, administrative personnel, and clerical and sales workers | 69 (15) | 23 (20) | 46 (14) |
| Skilled manual workers | 106 (24) | 30 (26) | 76 (23) |
| Semiskilled workers | 148 (34) | 36 (31) | 112 (34) |
| Unskilled laborer | 118 (27) | 30 (26) | 88 (27) |
| Total years homeless before study entry, mean (SD) | 6.8 (8.9) | 6.1 (8.6) | 7.1 (9.0) |
| First episode of homelessness at ≥50 y | 201 (45) | 62 (53) | 139 (42) |
| Lifetime substance use and mental health problems | |||
| Regular drinking | |||
| Childhood (<18 y) | 260 (59) | 66 (58) | 194 (59) |
| Adulthood (18-49 y) | 188 (42) | 52 (44) | 136 (41) |
| Regular drug use | |||
| Childhood (<18 y) | 52 (12) | 18 (15) | 34 (10) |
| Adulthood (18-49 y) | 262 (58) | 69 (59) | 193 (58) |
| Mental health problem | |||
| Childhood (<18 y) | 436 (98) | 112 (97) | 324 (98) |
| Young adulthood (18-25 y) | 203 (45) | 48 (41) | 155 (47) |
| Psychiatric hospitalization during lifetime | 91 (34) | 23 (32) | 68 (34) |
| Incarcerated in state or federal prison during lifetime | 168 (37) | 49 (42) | 119 (36) |
| Incarcerated in jail during lifetime | 299 (66) | 85 (73) | 214 (64) |
| Baseline values of measures collected every 6 mo | |||
| No. of confidants | |||
| 0-5 | 416 (92) | 109 (93) | 307 (92) |
| ≥6 | 34 (8) | 8 (7) | 26 (8) |
| Self-rated general health | |||
| Good, very good, or excellent | 198 (44) | 40 (34) | 158 (47) |
| Fair or poor | 252 (56) | 77 (66) | 175 (53) |
| Medical history | |||
| Hypertension | 259 (58) | 74 (63) | 185 (56) |
| Cardiac disease | 16 (4) | 6 (5) | 10 (3) |
| Congestive heart failure | 36 (8) | 15 (13) | 21 (6) |
| Stroke or transient ischemic attack | 54 (12) | 22 (19) | 32 (10) |
| Diabetes | 70 (16) | 30 (26) | 40 (12) |
| Chronic lung disease (asthma or COPD) | 328 (73) | 86 (74) | 242 (73) |
| Liver disease | 103 (23) | 36 (31) | 67 (20) |
| Kidney disease | 21 (5) | 7 (6) | 14 (4) |
| HIV/AIDS | 23 (5) | 9 (8) | 14 (4) |
| Cancer | 32 (7) | 11 (9) | 21 (6) |
| Body mass indexc | |||
| <18 | 11 (2) | 4 (3) | 7 (2) |
| 18-24.99 | 200 (44) | 48 (41) | 152 (46) |
| 25-29.99 | 130 (29) | 34 (29) | 96 (29) |
| >30 | 109 (24) | 31 (26) | 78 (23) |
| Impairment in activities of daily living | 176 (39) | 42 (36) | 134 (40) |
| Impairment in instrumental activities of daily living | 145 (32) | 40 (34) | 105 (32) |
| Cognitive impairment | 82 (18) | 24 (21) | 58 (18) |
| Falls in prior 6 mo | 146 (33) | 42 (36) | 104 (31) |
| Short Physical Performance Battery score <10 | 209 (46) | 63 (54) | 146 (44) |
| Binge drinking in prior 6 mo | 98 (22) | 27 (23) | 71 (21) |
| Drug use problem | 177 (39) | 48 (41) | 129 (39) |
| Smoking status | |||
| Current | 91 (20) | 19 (16) | 72 (22) |
| Former | 54 (12) | 15 (13) | 39 (12) |
| Never | 305 (68) | 83 (71) | 222 (67) |
| Depression in past 6 mo | 180 (67) | 40 (61) | 140 (69) |
| Psychiatric hospitalization in past 6 mo | 19 (22) | 5 (23) | 14 (21) |
| Incarcerated in jail in past 6 mo | 56 (12) | 15 (13) | 41 (12) |
Abbreviations: COPD, chronic obstructive pulmonary disease; GED, general educational development.
Percentages may not add to 100% due to rounding.
The other category includes Asian individuals.
Calculated as weight in kilograms divided by height in meters squared.
For common causes of death, we examined whether each diagnosis had been previously reported. We compared death rate over 4-year follow-up among participants enrolled before February 28, 2016, to that in Oakland by calculating standardized mortality ratios (SMRs). We used 4-year follow-up to avoid calculating SMRs after onset of the COVID-19 pandemic. In sensitivity analyses, we examined deaths over 2-year and 3-year follow-up. For groups cross-defined by age group (45-54, 55-64, 65-74, and 75-84 years) and sex, we calculated the 1-year mortality rate in Oakland. We defined this rate as number of observed deaths (obtained from California Department of Public Health annual mortality data for 2016-2020) divided by estimated population size (5-year American Community Survey for 2019).31 We extrapolated this 1-year rate to the analytic window of interest and multiplied by the number of at-risk participants to obtain the expected number of deaths. We summed expected numbers across age groups, obtaining sex-specific expected death counts. Finally, we divided sex-specific observed death counts in HOPE HOME by sex-specific expected death counts. We conducted analyses using SAS, version 9.4 (SAS Institute), and Stata, version 17 (StataCorp).
Results
Participant Characteristics
Of the 450 included participants, median (IQR) age at baseline was 58.1 (54.5-61.6) years, 107 (24%) were women, and 360 (80%) were Black (Table 1). Median (IQR) follow-up time through December 31, 2021, or date of death was 55 (38-93) months for the overall cohort and for the cohorts enrolled in 2013 through 2014 and 2017 through 2018 were 80 (47-95) months and 45 (26-49) months, respectively.
A total of 117 (26%) participants died through December 31, 2021. Of these, 101 deaths occurred among the 350 participants enrolled in 2013 through 2014 and 16 among the 100 participants enrolled in 2017 through 2018. Additionally, 93 deaths occurred among men and 24 among women. Forty-five deaths occurred after March 2020. Median (IQR) age at death was 64.6 (60.3-67.5) years (Figure).
Figure. Age at Time of Death Among 117 Homeless Adults Who Died During Follow-up.
Association of Participant Characteristics With Mortality
In multivariable analyses, increased risk of mortality was associated with first homelessness at 50 years and older (adjusted hazard ratio [aHR], 1.62; 95% CI, 1.13-2.32) and homelessness (aHR, 1.82; 95% CI, 1.23-2.68) or institutionalization (aHR, 6.36; 95% CI, 3.42-11.82) at any follow-up compared with being housed (Table 2). Similarly, fair or poor self-rated health (aHR, 1.64; 95% CI, 1.13-2.40) and diabetes (aHR, 1.55; 95% CI, 1.06-2.26) were associated with mortality.
Table 2. Association of Participant Characteristics With Mortality (n = 450).
| Characteristic | Adjusted hazard ratio (95% CI) | |
|---|---|---|
| Full modela | Parsimonious modelb | |
| Sociodemographics | ||
| Agec | 1.01 (0.98-1.05) | NA |
| Not married/partneredc | 1.49 (0.84-2.64) | NA |
| First episode of homelessness at ≥50 yd | 1.49 (1.05-2.15) | 1.62 (1.13-2.32) |
| Regular drug use in childhood (<18 y) | 1.49 (0.93-2.38) | NA |
| Time-varying characteristics | ||
| Housing status | ||
| Housed | 1 [Reference] | NA |
| Homeless | 1.76 (1.20-2.58) | 1.82 (1.23-2.68) |
| Staying in skilled nursing facility or jail/prison >3 mo | 6.12 (3.44-10.89) | 6.36 (3.42-11.82) |
| Incarcerated in jail | 1.82 (1.07-3.10) | NA |
| No. of confidants | ||
| ≥6 | 1 [Reference] | NA |
| 0-5 | 1.74 (0.87-3.46) | NA |
| Self-rated general health of fair or poor | 1.63 (1.13-2.36) | 1.64 (1.13-2.40) |
| Medical history | ||
| Cardiac disease | 1.55 (0.99-2.42) | NA |
| Congestive heart failure | 1.43 (0.90-2.28) | NA |
| Stroke or transient ischemic attack | 1.46 (0.96-2.23) | NA |
| Diabetes | 1.59 (1.10-2.29) | 1.55 (1.06-2.26) |
| Cancer | 1.26 (0.74-2.16) | NA |
| Liver disease | 1.73 (1.20-2.50) | NA |
| Cognitive impairment | 1.70 (1.11-2.60) | NA |
| Falls, any | 1.28 (0.88-1.86) | NA |
| Binge drinking | 0.60 (0.32-1.13) | NA |
| Mental health problems | 1.94 (0.76-4.98) | NA |
Abbreviation: NA, not applicable.
Only variables with univariate P ≤ .20 were entered in the full model.
Only variables significantly contributing to the fit of the model are retained in the adjusted parsimonious model.
Time varying.
Baseline.
Causes of Death
We obtained death certificates for 104 of 117 decedents (88.9%). The most common causes of death were heart disease (n = 17 [14.5%]), cancer (n = 17 [14.5%]), and drug overdose (n = 14 [12.0%]) (Table 3). Chronic lower respiratory diseases were the cause of 11 (9.4%) deaths and chronic liver disease of 8 (6.8%). Of participants with cancer as a primary cause of death, 7 (41.2%) had previously reported a diagnosis of cancer, and 5 (29.4%) and 3 (37.5%) previously reported heart disease and liver disease, respectively. Of 46 participants who died after March 2020, 3 had COVID-19 as a cause of death.
Table 3. Causes of the 117 Deaths in the Cohort .
| Causes of deatha | Deaths, No. (%) |
|---|---|
| Heart disease | 17 (14.5) |
| Cancer | 17 (14.5) |
| Liver and intrahepatic bile ducts | 4 (3.4) |
| Trachea, bronchus, and lung | 3 (2.6) |
| Prostate | 2 (1.7) |
| Colon, rectum, and anus | 2 (1.7) |
| Bladder | 1 (1.0) |
| Stomach | 1 (1.0) |
| Non-Hodgkin lymphoma | 1 (1.0) |
| Multiple myeloma and myeloproliferative disorders | 1 (1.0) |
| Corpus uteri and uterus | 1 (1.0) |
| All other and unspecified malignant neoplasms (leiomyosarcoma) | 1 (1.0) |
| Drug overdose | 14 (12.0) |
| Polysubstanceb | 6 (5.1) |
| Cocaine | 3 (2.6) |
| Amphetamines | 3 (2.6) |
| Opioids | 2 (1.7) |
| Chronic lower respiratory diseases | 11 (9.4) |
| Chronic liver disease and cirrhosis | 8 (6.8) |
| Cerebrovascular disease | 7 (6.0) |
| Accidents | 6 (5.1) |
| Pneumonia | 5 (4.3) |
| Nephritis, nephrotic syndrome, and nephrosis | 4 (3.4) |
| COVID-19 | 3 (2.6) |
| Diabetes | 3 (2.6) |
| HIV | 3 (2.6) |
| Homicide | 2 (1.7) |
| Other digestive diseases (gastrointestinal hemorrhage) | 2 (1.7) |
| All other diseases (amyloidosis) | 1 (1.0) |
| Psychoactive substance use disorder | 1 (1.0) |
| Missing | 13 (11.1) |
Causes of death based on the International Statistical Classification of Diseases and Related Health Problems, Tenth Revision underlying cause of death codes.
Of the 6 polysubstance-related deaths, 1 was attributed to alcohol and cocaine and 1 to cocaine and methamphetamine; the causes of the other 4 deaths were not further specified.
Standardized Mortality Ratios
Among participants enrolled in 2013 through 2014, the age-standardized SMR over 4-year follow-up was 3.5-fold higher (95% CI, 2.5-4.4) than the general population (Table 4). Mortality rates for men were 3.1-fold higher (95% CI, 2.1-4.2); those for women were 5.2-fold higher (95% CI, 2.1-8.2). Sensitivity analyses using varying follow-up periods showed similar results (eTable in the Supplement).
Table 4. Age-Standardized Mortality Ratios for All-Cause Deaths by Sex.
| Sex and age group | No. | Standardized mortality ratio (95% CI)a | |
|---|---|---|---|
| Observed deaths | Expected deaths | ||
| Male, y | |||
| 45-54 | 4 | 0.93 | NA |
| 55-64 | 21 | 7.18 | NA |
| 65-74 | 11 | 3.01 | NA |
| 75-84 | 0 | 0.36 | NA |
| Total | 36 | 11.47 | 3.1 (2.1-4.2) |
| Female, y | |||
| 45-54 | 1 | 0.13 | NA |
| 55-64 | 5 | 1.37 | NA |
| 65-74 | 5 | 0.51 | NA |
| 75-84 | 0 | 0.12 | NA |
| Total | 11 | 2.14 | 5.2 (2.1-8.2) |
| Total male and female | 47 | 13.61 | 3.5 (2.5-4.4) |
Abbreviation: NA, not applicable.
Deaths calculated based on participants enrolled prior to February 28, 2016, who died over a 4-year follow-up.
Discussion
In this population-based cohort study among homeless older adults, premature mortality was common, with an all-cause mortality rate 3.5-fold higher than the general population. The most common causes of death were heart disease, cancer, and drug overdose. Factors associated with mortality included late-life homelessness, ongoing homelessness or institutionalization, and self-rated fair or poor health.
The present findings confirm and extend prior work showing that homeless populations experience disparities in mortality. The average age of death among homeless adults is 42 to 52 years.11 In a retrospective cohort study of adults who received care at a homeless health care organization, mortality rates in individuals aged 25 to 44 years were 9-fold higher than the general population, while those in individuals aged 45 to 64 years were 4.5-fold higher.5,6 Other US6,32 and international studies show similar findings.33 Despite differences in study design, we identified a similar mortality ratio.
Most prior studies have included people who access homeless health care and used medical records to identify factors associated with mortality retrospectively.6,12 These studies define homelessness at one point in time, without accounting for its dynamic nature. We found that many participants regained housing and that remaining homeless was associated with mortality. Staying in an institution was associated with increased risk, although this may reflect reverse causality: limited end-of-life options for persons experiencing homelessness may lead to reliance on skilled nursing facilities to provide end-of-life care. Self-reported fair or poor health was associated with mortality, suggesting that self-report provides valuable prognostic information. Diabetes was associated with mortality, consistent with prior research.12
Those with first homelessness at 50 years and older had elevated mortality risk compared with those with earlier homelessness. Several factors may explain this. First, people 50 years and older with earlier homelessness have survived a period of homelessness. Newly homeless individuals may not have access to resources, or resilience, from prior lived experiences of homelessness. Second, late-life homelessness may represent a “health shock” that leads to worse control of chronic conditions, precipitating health decline, and premature mortality. Third, illness may be on the pathway to late-life homelessness. Individuals with late-life homelessness are more likely to have a crisis leading to homelessness; serious illness is a common precipitant.20 In these cases, illness leads to job loss, eviction, and homelessness. Illness may then lead to death while homeless.
The most common causes of death were heart disease, cancer, and drug overdose, consistent with prior research.5 Relatively few participants reported diagnoses of cancer, heart disease, and liver disease before dying of these conditions. This may reflect difficulties accessing health care while homeless, leading to delayed diagnosis and a short lag time between diagnosis and death. For heart disease, this discrepancy may reflect its overreporting as a cause of death.34
Although the pace of deaths accelerated during the COVID-19 pandemic, only 3 participants died of COVID-19. SARS-COV-2 infection spread rapidly in homeless shelters.35 However, most participants experiencing homelessness during the pandemic stayed in unsheltered settings or noncongregate “shelter-in-place” hotels, where SARS-COV-2 spread was rare.35 The higher mortality during the pandemic may reflect the consequences of disruptions in health care, drug supply, and access to substance use treatment, plus cohort aging.
The high proportion of Black participants reflects the 3- to 4-fold elevated risk of homelessness among Black US adults, owing to effects of structural racism in housing, employment, education, and criminal justice systems.36 Homelessness may be one of many mediators between structural racism and elevated mortality risk among Black US adults.
These findings underscore the need to recognize the elevated mortality risk in homeless older adults and deliver interventions to prevent and end homelessness. Older adults face the highest risk of severe rent burden in the US and a high risk of homelessness.37 Preventing homelessness for this population is key, given the increased mortality risk associated with late-life homelessness. For those with long-standing homelessness, studies support the use of permanent supportive housing, which is subsidized housing associated with voluntary supportive services and offered on a Housing First basis, without preconditions (eg, sobriety, engagement in care).38 Permanent supportive housing is highly effective at maintaining housing among those with long-term homelessness,39 although efforts are needed to adapt permanent supportive housing to meet older adults’ needs.4 For those with new-onset homelessness, housing subsidies and navigation services can be deployed. While people remain homeless, risk can be mitigated through outreach and tailored medical care. The high risk of death from chronic disease highlights the need for end-of-life care tailored to the needs of homeless-experienced adults.40
Limitations
This study has several limitations. We may have underestimated out-of-state deaths. We were unable to obtain death certificates for 13 individuals, including 3 who died in states requiring that a legal surrogate request certificates. Given the relatively small number of deaths, we may have lacked power to detect some factors associated with mortality. We sampled from a city with a larger than average Black population; 26% of the Oakland population is Black compared with 12.4% nationally.41 Nearly 80% of the cohort was Black, approximately 3 times that in Oakland. This overrepresentation is comparable with other large cities.42 The interval between the last interview and month of death varied, which could result in misclassification and bias for time-varying variables. However, health status is unlikely to change over short time periods.43 We categorized prolonged jail/prison stays as an institutional setting with skilled nursing facility stays, which may not be comparable. However, prolonged incarcerations were rare (n = 2), and results were similar when we classified these separately. We did not adjust SMRs for race because homelessness is likely on the causal pathway for elevated mortality among Black US adults.
Conclusions
In this cohort study, we found that premature mortality was common among homeless older adults and that late-life homelessness and ongoing homelessness are key prognostic factors. These findings point to the urgent need for policy approaches to prevent and end homelessness among older adults in the US.
eTable. Standardized Mortality Ratios for All-Cause Deaths by Sex and Age Group
References
- 1.Hahn JA, Kushel MB, Bangsberg DR, Riley E, Moss AR. Brief report: the aging of the homeless population: fourteen-year trends in San Francisco. J Gen Intern Med. 2006;21(7):775-778. doi: 10.1111/j.1525-1497.2006.00493.x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Culhane D, Doran K, Schretzman M, et al. The emerging crisis of aged homelessness in the US: could cost avoidance in health care fund housing solutions? Int J Popul Data Sci. 2019;4(3). doi: 10.23889/ijpds.v4i3.1185 [DOI] [Google Scholar]
- 3.Brown RT, Kiely DK, Bharel M, Mitchell SL. Geriatric syndromes in older homeless adults. J Gen Intern Med. 2012;27(1):16-22. doi: 10.1007/s11606-011-1848-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Brown RT, Hemati K, Riley ED, et al. Geriatric conditions in a population-based sample of older homeless adults. Gerontologist. 2017;57(4):757-766. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Baggett TP, Hwang SW, O’Connell JJ, et al. Mortality among homeless adults in Boston: shifts in causes of death over a 15-year period. JAMA Intern Med. 2013;173(3):189-195. doi: 10.1001/jamainternmed.2013.1604 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Hwang SW, Orav EJ, O’Connell JJ, Lebow JM, Brennan TA. Causes of death in homeless adults in Boston. Ann Intern Med. 1997;126(8):625-628. doi: 10.7326/0003-4819-126-8-199704150-00007 [DOI] [PubMed] [Google Scholar]
- 7.Schinka JA, Bossarte RM, Curtiss G, Lapcevic WA, Casey RJ. Increased mortality among older veterans admitted to VA homelessness programs. Psychiatr Serv. 2016;67(4):465-468. doi: 10.1176/appi.ps.201500095 [DOI] [PubMed] [Google Scholar]
- 8.Richmond-Rakerd LS, D’Souza S, Milne BJ, Caspi A, Moffitt TE. Longitudinal associations of mental disorders with physical diseases and mortality among 2.3 million New Zealand citizens. JAMA Netw Open. 2021;4(1):e2033448. doi: 10.1001/jamanetworkopen.2020.33448 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Simons RL, Lei MK, Klopack E, Beach SRH, Gibbons FX, Philibert RA. The effects of social adversity, discrimination, and health risk behaviors on the accelerated aging of African Americans: further support for the weathering hypothesis. Soc Sci Med. 2021;282:113169. doi: 10.1016/j.socscimed.2020.113169 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Baggett TP, Rigotti NA. Cigarette smoking and advice to quit in a national sample of homeless adults. Am J Prev Med. 2010;39(2):164-172. doi: 10.1016/j.amepre.2010.03.024 [DOI] [PubMed] [Google Scholar]
- 11.O’Connell J. Premature mortality in homeless populations: a review of the literature. National Health Care for the Homeless Council . December 2005. Accessed July 31, 2022. http://sbdww.org/wp-content/uploads/2011/04/PrematureMortalityFinal.pdf
- 12.Hwang SW, Lebow JM, Bierer MF, O’Connell JJ, Orav EJ, Brennan TA. Risk factors for death in homeless adults in Boston. Arch Intern Med. 1998;158(13):1454-1460. doi: 10.1001/archinte.158.13.1454 [DOI] [PubMed] [Google Scholar]
- 13.Cawley C, Kanzaria HK, Zevin B, Doran KM, Kushel M, Raven MC. Mortality among people experiencing homelessness in San Francisco during the COVID-19 pandemic. JAMA Netw Open. 2022;5(3):e221870. doi: 10.1001/jamanetworkopen.2022.1870 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.von Elm E, Altman DG, Egger M, Pocock SJ, Gøtzsche PC, Vandenbroucke JP; STROBE Initiative . The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. Ann Intern Med. 2007;147(8):573-577. doi: 10.7326/0003-4819-147-8-200710160-00010 [DOI] [PubMed] [Google Scholar]
- 15.Burnam MA, Koegel P. Methodology for obtaining a representative sample of homeless persons: the Los Angeles Skid Row study. Eval Rev. 1988;12(2):117-152. doi: 10.1177/0193841X8801200202 [DOI] [Google Scholar]
- 16.Homeless Emergency Assistance and Rapid Transition to Housing Act of 2009, HR 1877, 111th Cong (2009). Pub L No. 111-22. Accessed July 31, 2022.
- 17.Kochanek KD, Murphy SL, Xu J, Arias E. Deaths: final data for 2017. Natl Vital Stat Rep. 2019;68(9):1-77. [PubMed] [Google Scholar]
- 18.Hollingshead AB, Redlich FC. Social class and mental illness: a community study. Am J Public Health. 2007;97(10):1756-1757. doi: 10.2105/AJPH.97.10.1756 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.McLellan AT, Kushner H, Metzger D, et al. The Fifth Edition of the Addiction Severity Index. J Subst Abuse Treat. 1992;9(3):199-213. doi: 10.1016/0740-5472(92)90062-S [DOI] [PubMed] [Google Scholar]
- 20.Brown RT, Goodman L, Guzman D, Tieu L, Ponath C, Kushel MB. Pathways to homelessness among older homeless adults: results from the HOPE HOME Study. PLoS One. 2016;11(5):e0155065. doi: 10.1371/journal.pone.0155065 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Tsemberis S, McHugo G, Williams V, Hanrahan P, Stefancic A. Measuring homelessness and residential stability: the residential time-line follow-back inventory. J Community Psychol. 2007;35(1):29-42. doi: 10.1002/jcop.20132 [DOI] [Google Scholar]
- 22.Gielen AC, McDonnell KA, Wu AW, O’Campo P, Faden R. Quality of life among women living with HIV: the importance violence, social support, and self care behaviors. Soc Sci Med. 2001;52(2):315-322. doi: 10.1016/S0277-9536(00)00135-0 [DOI] [PubMed] [Google Scholar]
- 23.National Health and Nutrition Examination Survey: NHANES questionnaires, datasets, and related documentation. Centers for Disease Control and Prevention . Accessed July 31, 2022. https://wwwn.cdc.gov/nchs/nhanes/Default.aspx
- 24.Katz S. Assessing self-maintenance: activities of daily living, mobility, and instrumental activities of daily living. J Am Geriatr Soc. 1983;31(12):721-727. doi: 10.1111/j.1532-5415.1983.tb03391.x [DOI] [PubMed] [Google Scholar]
- 25.Sullivan G, Dumenci L, Burnam A, Koegel P. Validation of the brief instrumental functioning scale in a homeless population. Psychiatr Serv. 2001;52(8):1097-1099. doi: 10.1176/appi.ps.52.8.1097 [DOI] [PubMed] [Google Scholar]
- 26.Bland RC, Newman SC. Mild dementia or cognitive impairment: the Modified Mini-Mental State examination (3MS) as a screen for dementia. Can J Psychiatry. 2001;46(6):506-510. doi: 10.1177/070674370104600604 [DOI] [PubMed] [Google Scholar]
- 27.Guralnik JM, Simonsick EM, Ferrucci L, et al. A short physical performance battery assessing lower extremity function: association with self-reported disability and prediction of mortality and nursing home admission. J Gerontol. 1994;49(2):M85-M94. doi: 10.1093/geronj/49.2.M85 [DOI] [PubMed] [Google Scholar]
- 28.Babor TF, Higgins-Biddle JC, Saunders JB, Monteiro MG. AUDIT: the Alcohol Use Disorders Identification Test: guidelines for use in primary health care. World Health Organization . November 18, 2001. Accessed July 31, 2022. https://www.who.int/publications/i/item/WHO-MSD-MSB-01.6a
- 29.Humeniuk R, Henry-Edwards S, Ali R, Poznyak V, Monteiro M, Maristela G. The Alcohol, Smoking and Substance Involvement Screening Test (ASSIST): manual for use in primary care. World Health Organization . January 1, 2010. Accessed July 31, 2022. https://www.who.int/publications/i/item/978924159938-2
- 30.Al-Delaimy WK, Edland S, Pierce JP, et al. California Tobacco Survey (CTS) 2008. University of California, San Diego . Accessed July 31, 2022. https://library.ucsd.edu/dc/object/bb5086895c
- 31.IPUMS USA. University of Minnesota . Accessed July 31, 2022. https://usa.ipums.org/usa/
- 32.Hibbs JR, Benner L, Klugman L, et al. Mortality in a cohort of homeless adults in Philadelphia. N Engl J Med. 1994;331(5):304-309. doi: 10.1056/NEJM199408043310506 [DOI] [PubMed] [Google Scholar]
- 33.Cheung AM, Hwang SW. Risk of death among homeless women: a cohort study and review of the literature. CMAJ. 2004;170(8):1243-1247. doi: 10.1503/cmaj.1031167 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Coady SA, Sorlie PD, Cooper LS, Folsom AR, Rosamond WD, Conwill DE. Validation of death certificate diagnosis for coronary heart disease: the Atherosclerosis Risk in Communities (ARIC) Study. J Clin Epidemiol. 2001;54(1):40-50. doi: 10.1016/S0895-4356(00)00272-9 [DOI] [PubMed] [Google Scholar]
- 35.Montgomery MP, Paulin HN, Morris A, et al. Establishment of isolation and noncongregate hotels during COVID-19 and symptom evolution among people experiencing homelessness: Atlanta, Georgia, 2020. J Public Health Manag Pract. 2021;27(3):285-294. doi: 10.1097/PHH.0000000000001349 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Paul DW, Knight KR, Olsen P, Weeks J, Yen IH, Kushel MB. Racial discrimination in the life course of older adults experiencing homelessness: results from the HOPE HOME Study. J Soc Distress Homeless. 2020;29(2):184-193. doi: 10.1080/10530789.2019.1702248 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Housing America’s older adults 2019. Joint Center for Housing Studies of Harvard University . 2019. Accessed July 31, 2022.https://www.jchs.harvard.edu/sites/default/files/reports/files/Harvard_JCHS_Housing_Americas_Older_Adults_2019.pdf
- 38.Housing first in permanent supportive housing. US Department of Housing and Urban Development . Accessed July 31, 2022. https://files.hudexchange.info/resources/documents/Housing-First-Permanent-Supportive-Housing-Brief.pdf
- 39.Stergiopoulos V, Hwang SW, Gozdzik A, et al. ; At Home/Chez Soi Investigators . Effect of scattered-site housing using rent supplements and intensive case management on housing stability among homeless adults with mental illness: a randomized trial. JAMA. 2015;313(9):905-915. doi: 10.1001/jama.2015.1163 [DOI] [PubMed] [Google Scholar]
- 40.Kushel MB, Miaskowski C. End-of-life care for homeless patients: “she says she is there to help me in any situation”. JAMA. 2006;296(24):2959-2966. doi: 10.1001/jama.296.24.2959 [DOI] [PubMed] [Google Scholar]
- 41.Supplementary tables on race and Hispanic origin: 2020 Census redistricting data (P.L. 94-171). US Census Bureau . August 12, 2021. Accessed July 31, 2022. https://www.census.gov/data/tables/2020/dec/2020-redistricting-supplementary-tables.html
- 42.CoC homeless populations and subpopulations reports. US Department of Housing and Urban Development . Accessed July 31, 2022. https://www.hudexchange.info/programs/coc/coc-homeless-populations-and-subpopulations-reports/
- 43.Feenstra M, van Munster BC, MacNeil Vroomen JL, de Rooij SE, Smidt N. Trajectories of self-rated health in an older general population and their determinants: the Lifelines Cohort Study. BMJ Open. 2020;10(2):e035012. doi: 10.1136/bmjopen-2019-035012 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
eTable. Standardized Mortality Ratios for All-Cause Deaths by Sex and Age Group