Examining Fall Risk among Formerly Homeless Older Adults Living in Permanent Supportive Housing

Abstract

Although permanent supportive housing (PSH) has been credited with a decline in the number of chronically homeless adults in the United States since 2007, the extent to which PSH can accommodate the needs of a prematurely aging population, including reducing the likelihood of falls, is unclear. The objective of this study is to examine the prevalence and correlates of falls with a sample of 237 tenants (45-80 years old) from two PSH programs in Los Angeles from January 1, 2017, to August 10, 2017. We also explore the location and severity of fall-related injury using a sub-sample of 66 tenants. Standard surveys queried demographics, health status, history of homelessness, and falls. Multivariable logistic regression assessed the correlates of falling in the past year. More than half of the sample had fallen and more than 40% had multiple falls in the past year. Functional impairment, frailty, and persistent pain were all associated with increased fall risk. For the 66 tenants who provided more detailed fall information, more than 40% fell at home and of those, nearly half fell in their bathroom. Fall-related injuries were common, with more than one third of the subsample experiencing serious injury. These findings suggest that fall prevention is needed in PSH but that more research is needed to understand the degree to which individual and environmental risk factors are contributing to falls.

Falls are common among older persons living independently in the community (Verma et al., 2016) and are associated with disabilities, impaired quality of life, and increased health care burden and costs (Bergen, 2016). Falls are also common among adults who have experienced chronic homelessness, a population that represents a stable, aging cohort whose average age is approaching 60 years old and whose functional age is significantly older than its chronological age (Brown et al., 2017; Culhane, Metraux, Byrne, Stino, & Bainbridge, 2013). Permanent supportive housing (PSH) using a housing first approach is an evidence-based intervention that provides immediate access to independent housing along with support services, such as mental health services, medical linkage, and case management services, to adults who have experienced chronic homelessness (U.S. Interagency Council on Homelessness, 2010). In 2018, the inventory of PSH units in the United States had the capacity to house 361,386 individuals, making it the largest component of the housing assistance system for people affected by homelessness (Office of Community Planning and Development, 2018). Although PSH has been credited with a decline in the number of chronically homeless adults in the United States since 2007 (Office of Community Planning and Development, 2018), the extent to which PSH can accommodate the needs of a prematurely aging population, including reducing the likelihood of falls, is unclear (Henwood, Katz, & Gilmer, 2015; National Academies of Sciences, Engineering, and Medicine, 2018).

Approximately one third of a population-based sample of homeless adults aged 50 or older reported falling in the past 6 months, which is similar to fall rates among representative community samples that are approximately 2 decades older (Brown et al., 2017). Individual risk factors that increase fall risk and are highly prevalent among adults who have experienced chronic homelessness include multiple chronic medical conditions, substance use, malnutrition, vision loss and other functional loss (e.g., impairment of activities of daily living, or ADLs), frailty, and persistent pain (Ivers et al., 2002; Kojima, 2015; Leveille et al., 2009; Meijers et al., 2012; Pynoos, Steinman, & Nguyen, 2010; Rubenstein, 2006; Tinetti et al., 1988). Individual fall risk factors may be even higher among formerly homeless adults living in PSH because of a growing practice known as vulnerability indexing, wherein homeless individuals with a higher risk of mortality due to medical conditions receive priority for placement in PSH (Henwood et al., 2019; Henwood, Lahey, Rhoades, Winetrobe, & Wenzel, 2018). PSH may also introduce a variety of environmental risk factors since the average age of the chronically homeless population was mid-30s when the housing first approach was originally developed in the 1990s (Padgett, Henwood, & Tsemberis, 2016) and now is approaching 60 years old (Culhane, Metraux, Byrne, Stino, & Bainbridge, 2013). In a study of 25 PSH tenants, Gutman et al. (2017) found that home safety concerns specific to an aging population had not been addressed in PSH apartments. Such factors can jeopardize PSH tenants’ ability to age in place (Henwood et al., 2015).

In this study, we examined the prevalence and correlates of falls in a sample of 237 older adults aged 45 and older who are living in PSH. Given that the average age of the target population for PSH (i.e. chronically homeless adults) is approaching 60 (Culhane et al., 2013) and that this population experiences decades early mortality (Henwood, Byrne, and Scriber, 2015), it is a safe assumption that the majority of PSH tenants are aged 45 or older. Based on the literature, we hypothesized that falls among PSH tenants will be associated with multiple chronic health conditions, visual impairment and functional loss (e.g., ADL impairment), food insecurity, frailty, and persistent pain (Ivers et al., 2002; Kojima, 2015; Leveille et al., 2009; Meijers et al., 2012; Pynoos, Steinman, & Nguyen, 2010; Rubenstein, 2006; Tinetti et al., 1988). Using a subsample (n = 66), we will also explore the location and severity of fall-related injury among PSH tenants.

METHODS

Sample Recruitment

Residents aged 45 or older in two PSH programs in Los Angeles County were recruited for participation in this study. Less than 10% of residents in either program was younger than 45. Residents were invited to participate in the study through a mailed letter explaining the purpose of the study and inviting them to contact the research team. The letter also indicated that residents could enroll in the study when interviewers were on site at their respective buildings. For the larger PSH program, which had more than 1,000 residents, determining who received invitations was done by sampling individuals aged 45 or older using a random number generator. Based on an expected frequency of geriatric syndromes of 80%, which was a conservative estimate derived from Brown et al., (2013) who found 92% prevalence of geriatric syndromes among homeless adults aged 50 or older, and 95% confidence limits, we calculated needing a sample of 203 residents in order to obtain prevalence estimates for this population. We sampled with replacement so that when individuals declined to participate or did not respond, another randomly selected participant was chosen. To ensure representation across different age ranges, randomization was stratified by age groups (i.e., 45-49, 50-54, 55-59, 60-64, 65-69, 70-74, 75-79, and 80+) and the number of people selected from each group was proportionate to the overall age distribution at the agency. Since the second partner agency had just over 100 total residents, we invited all 100 residents who were aged 45 or older to participate. For both methods, agency staff members assisted the study team in engaging with eligible residents.

Data Collection

All interested selected residents were screened to confirm study eligibility before proceeding with the informed consent process. Participants were eligible if they were at least 45 years old, spoke English, and were currently living in PSH provided by either of the two partner agencies. Age eligibility was reduced from 50 to 45 years old before data collection began due to provider feedback indicating that they perceived tenants younger than age 50 were experiencing some geriatric conditions. Screenings occurred over the phone or in person. Eligible residents then completed the informed consent process in which a trained study interviewer reviewed an information sheet and completed a screening for delirium; verbal consent was required. The study, which did not collect identifiable information including waiving written consent, received expedited approval from the institutional review board at the University of Southern California. Each interview took about 1.5 hours to complete, and participants received $25. Trained interviewers, who were graduate students from schools of social work, arts and sciences, and occupational therapy, administered surveys and recorded screening and test results on a study-supplied iPad. All interviewers were required to complete IRB approved human subjects training and received additional training on the measures used for the study as well as on trauma-informed interviewing.

During recruitment, 506 residents from both programs were invited to participate and 275 (54%) were screened for eligibility by the research team. Of 243 residents who were eligible, three participants scheduled interviews but did not show up and could not be rescheduled, and three participants did not pass the delirium screener. This resulted in a final sample of 237 participants. Because investigators identified high rates of reported falls midway through the study, questions on the location and the number of serious fall-related injury were added to the survey instrument. This resulted in additional fall related information from a subsample of 66 participants. Data collection occurred between January 1, 2017 and August 10, 2017.

Measures

Falls

A previous published study using the current dataset found that almost 57% (n = 134) of the sample experienced a fall in the past year (Henwood et al., 2019). Falls were self-reported by participants and operationalized as “any event where any part of your body above your ankle hit the floor or ground. Also include falls that might have occurred on stairs” (Cummings, Nevitt, & Kidd, 1988). Additional fall related questions given to the sub-set of 66 participants (49% of those with a reported fall) included self-reported fall location (inside your home’s bathroom; inside your home somewhere besides the bathroom; inside someone else’s home; inside another building; inside another location; outside), and whether the respondent had experienced an injury related to the fall, and if so, what type of injury (Cummings et al., 1988). Fall locations inside someone else’s home, inside another building, and inside another location were collapsed into a single category indicating inside but not in the respondent’s home. The open-ended question assessing type of injury sustained during falls was coded to indicate serious injury, defined as falls leading to medical attention (e.g., fractures, joint dislocation, head injury, and lacerations; Bhasin et al., 2018). Six people who reported fall-related injuries had injuries that could not be post hoc categorized as serious versus not and were coded as missing.

Possible Correlates

ADLs were measured using five items from the Katz ADL Scale (bathing, dressing, eating, transferring, and toileting; Katz, Downs, Cash, & Grotz, 1970), and impairment was defined as self-reported difficulty performing one or more ADL. Frailty was measured according to the Fried criteria (Fried et al., 2001) and included unintentional weight loss, grip weakness, exhaustion, slow gait, and low physical activity; participants were coded as frail if they met three or more of these criteria. The extent to which pain interfered with general activities was measured on a scale of 0 to 11, with higher scores indicating pain interfered more with general activity and 0 indicating no pain; participants were coded into tertiles of pain severity, including those with no pain (Krebs et al., 2009).

Food security was assessed with the US Department of Agriculture’s Adult Food Security Module (Bickel et al., 2000); we dichotomized responses as high or marginal versus low or very low food security. Using response options adapted from literature with persons experiencing homelessness (Bassuk et al., 1998; Hwang, 2001; National Health Care for the Homeless Council, 2011), we assessed for all lifetime diagnosis of chronic health conditions and grouped diagnoses into 8 categories: cardiovascular conditions and diseases (e.g., heart attack, high cholesterol); musculoskeletal disorders (e.g., osteoporosis, rheumatoid arthritis); autoimmune diseases (e.g., Lupus, HIV/AIDS); neurological disorders (e.g., epilepsy, Alzheimer’s); eye and vision problems (includes self-reported vision impairment and eye disorders such as cataracts, glaucoma, and macular degeneration); kidney/liver diseases (e.g., cirrhosis, failing kidneys); cancer; and mental health disorders. Other measures included substance use in the past 30 days (three variables: use of alcohol to intoxication, marijuana use, and illicit substance use; McLellan et al., 1992); demographics (age, race, gender, and education); housing tenure (how long participants had lived in PSH); and lifetime years of homelessness.

Analytic Methods

Basic descriptive statistics (e.g., sample size, percentage, mean, and standard deviation) are presented for all variables. Because of the limited sample size of 66 for the fall-related items added later in the study—location and injuries—we present only descriptive results for these items. A multivariable logistic regression model assessed the correlates of falling in the past year and included ADL impairment, frailty, food insecurity, chronic health conditions including visual impairment, substance use, and pain. The model also controlled for demographics, housing history, and lifetime duration of homelessness. Because ADL impairment and frailty were highly correlated (χ² = 43.46, p < .001), for the purposes of analysis they were combined into a single variable: 0 = neither ADL impairment nor frailty, 1 = ADL impairment without frailty, 2 = frailty without ADL impairment, or 3 = both ADL impairment and frailty. Variance inflation factors were calculated in Stata 14.0 post-regression to ensure there were no multicollinearity issues. The sample size for the logistic regression model (233) is slightly lower than the overall sample (237) because of missing data. The dataset analyzed for the current study is available from the corresponding author on request.

RESULTS

Sample Characteristics

As shown in Table 1, the overall average age of the sample was 57.7 years old (SD = 6.4), with 63% of the sample identifying as male and 61% identifying as African American. Nine percent of the sample (n = 22) was younger than 50 years old, and 34% had less than a high school education. The average length of residence in a PSH unit was 4.8 years (SD = 3.7). The mean lifetime years of homelessness was 7.9 (SD = 8.1).

Table 1.

Demographics, Frailty, Falls, and Other Characteristics (N = 237)

	% (n)	M (SD)
Age (range: 45-80)		57.7 (6.4)
Women	36.7 (87)
Race and ethnicity
African American	61.0 (144)
White	18.2 (43)
Latino/a	7.2 (17)
Multiracial or other	13.6 (32)
Less than high school education	34.2 (81)
Housing and homelessness
Lifetime years of homelessness (range: 0.08-45)		7.9 (8.1)
Years in PSH (range: 0.08-22)		4.8 (3.7)
Number of chronic health conditions (range: 0–21)		6.2 (3.8)
Any chronic physical health condition diagnosis	92.0 (218)
Any chronic mental health condition diagnosis	78.1 (185)
Activities of daily living
Impairment only	12.3 (29)
Frailty only	15.7 (37)
Both	29.4 (69)
Neither	42.6 (100)
Pain interference with general activity (range: 0–11)		4.4 (4.5)
Pain interference tertiles
First tertile	39.7 (94)
Second tertile	34.2 (81)
Third tertile	26.2 (62)
Substance use (past 30 days)
Days of alcohol use to intoxication (range: 0-30)		2.4 (6.6)
No alcohol use	77.6 (184)
Days of marijuana use (range: 0-30)		6.1 (11.0)
No marijuana use	67.5 (160)
Days of illicit substance use (range: 0-1)		0.2 (0.4)
No illicit substance use	84.0 (199)
Low or very low food security	67.1 (159)
Chronic Health Problems
Cardiovascular conditions	69.2 (164)
Musculoskeletal conditions	44.7 (106)
Autoimmune disorders	21.5 (51)
Neurological conditions	17.7 (42)
Vision impairment or eye/vision disorders	70.9 (168)
Kidney/liver conditions	30.4 (72)
Cancers	8.9 (21)
Mental health disorders	76.8 (182)
Any fall in the past year	56.8 (134)
Number of falls
0	43.22 (102)
1	15.7 (37)
2	15.7 (37)
3	11.9 (28)
4	3.8 (9)
5+	9.8 (23)

Nearly one third of the sample (29%) had both ADL impairment and frailty (16% had frailty without ADL impairment, 12% had ADL impairment without frailty, and 43% had neither). The average pain interference score was 4.4 (SD = 4.5). Two-thirds of the sample (67%) reported low or very low food security. Chronic health problems were common, with 77% reporting mental health disorder diagnoses, 71% vision impairment or eye/vision disorders, 69% cardiovascular conditions, 45% musculoskeletal conditions, 30% kidney/liver conditions, 22% autoimmune disorders, 18% neurological conditions, and 9% cancers. In terms of substance use during the prior 30 days, the mean number of days participants had consumed alcohol to intoxication was 2.4 (SD = 6.6; range: 0-30; 78% reported no alcohol use), marijuana use was 6.1 (SD = 11.0; range: 0-30; 68% reported no marijuana use), and illicit substance use was 0.2 (SD = 0.4; range: 0-1; 84.0% reported no illicit substance use). More than half of the sample (57%) reported one or more falls in the prior year, with 16% reporting one, 16% reporting two, 12% reporting three, 4% reporting four, and 10% reporting five or more falls.

For the subset of participants who completed additional fall-related items (see Table 2), 70% reported some injury associated with a fall and 35% reported a serious injury. In terms of location, 41% reported that their falls had occurred exclusively outside, 15% only in their own home, 9% only in another inside location, and 33% in multiple locations. Among the 29 respondents who reported a fall at home (44% of the subsample), 21% of those were only in the bathroom, 52% were only in another location inside their home, and 28% fell in both in-home locations.

Table 2.

Fall Characteristics Among Subsample (N = 60-66)

	% (n)
Falls with any injury (n = 66)	69.7 (46)
Falls with serious injury (n = 60)	35.0 (21)
Fall locations (n = 66)
Outside only	40.9 (27)
Inside participant’s home only	15.2 (10)
Another inside location only	9.1 (6)
Multiple locations	33.3 (22)
Unknown location	1.5 (1)
Any fall at home	43.9 (29)
Locations among those who fell at home (n = 29)
Inside bathroom	20.7 (6)
Other location inside home	51.7 (15)
Both	27.6 (8)

Correlates of Falls

The multivariable logistic regression model (Table 3) revealed statistically significantly increased odds of past-year falls among those with ADL impairment without frailty (OR = 3.96; 95% CI = 1.41, 11.11) and among those with both ADL impairment and frailty (OR = 5.48; 95% CI = 2.24, 13.39), compared to respondents with neither frailty nor ADL impairment. Those in the highest tertile of pain that interfered with general activities were more likely than those in the lowest tertile to report past-year falls (OR = 3.63; 95% CI = 1.44, 9.16), whereas those in the middle tertile did not have statistically significantly greater odds of reporting falls. Demographic characteristics, housing and homelessness histories, food insecurity, chronic health conditions groupings, and substance use were not associated with falls. In exploratory analyses, we also assessed correlates of falls with serious injuries in the subsample that answered that question (n = 60; results not shown in tables). Given the small sample size, results should be considered preliminary, but among those who reported falls, women were less likely to report a fall with serious injury (OR = 0.26; 95% CI = 0.08, 0.89).

Table 3.

Multivariable Logistic Regression of Falls in the Past Year (N = 233)

	OR (95% CI)
Age	1.02 (0.96, 1.08)
Female gender	1.01 (0.50, 2.03)
Race and ethnicitya
African American	0.61 (0.25, 1.46)
Latino/a	2.77 (0.49, 15.55)
Multiracial or other	0.62 (0.20, 1.98)
Less than high school education	0.93 (0.47, 1.84)
Lifetime years of homelessness	1.02 (0.98, 1.06)
Years in PSH	1.01 (0.92, 1.11)
Activities of daily living
Impairment only	3.96 (1.41, 11.11)
Frailty only	2.08 (0.87, 4.97)
Both	5.48 (2.24, 13.39)
Pain interference with general activityb
Second tertile	0.96 (0.47, 1.98)
Third tertile	3.63 (1.44, 9.16)
Days of substance use (past 30 days)
Alcohol use to intoxication	1.01 (0.96, 1.06)
Marijuana use	0.98 (0.95, 1.01)
Illicit substance use	1.38 (0.55, 3.48)
Low or very low food security	0.93 (0.47, 1.82)
Chronic Health Problems
Cardiovascular conditions	1.64 (0.81, 3.32)
Musculoskeletal conditions	0.87 (0.44, 1.69)
Autoimmune disorders	0.72 (0.31, 1.63)
Neurological conditions	1.60 (0.60, 4.30)
Vision impairment or eye/vision disorders	0.56 (0.27, 1.14)
Kidney/liver conditions	0.95 (0.46, 1.98)
Cancers	1.58 (0.43, 5.86)
Mental health disorders	1.70 (0.79, 3.67)

^aWhite is omitted category.

^bFirst tertile is omitted category.

Bold indicates statistically significant relationship at p<0.05.

DISCUSSION

We found that formerly homeless persons over the age of 45 who are living in PSH had a high prevalence of falls, with more than half of our sample having fallen at least once and more than 40% falling more than once in the past year. The findings also show that among a sub-set of residents who were asked about fall-related injuries, more than one third reported experiencing serious injury. As hypothesized, we found that ADL impairment, frailty (when co-occurring with ADL impairment), and persistent pain were associated with increased fall risk, which is consistent with the literature on older adults (Kojima, 2015; Leveille et al., 2009; Pynoos, Steinman, & Nguyen, 2010; Tinetti et al., 1988). This suggests that efforts to address modifiable factors such as interventions designed to improve ADLs (Phelan, Williams, Penninx, LoGerfo, & Leveille, 2004), frailty (Puts et al., 2017), and pain management (Reid, Eccleston, & Pillemer, 2015) could potentially reduce fall risk.

We did not find that multiple chronic conditions, visual impairment, food insecurity, or substance use including alcohol was associated with increased fall risk, as previously reported in the general population (Ivers et al., 2002; Meijers et al., 2012; Rubenstein, 2006). This may be due to lack of statistical power or to high rates of these conditions in PSH as compared to the general population. It should be noted, however, that this sample reported low rates of substance use, which is consistent with other research on older tenants in PSH (Shibusawa & Padgett, 2009). Further research with a larger sample is warranted.

Although this study focused primarily on individual risk factors as opposed to environmental risk factors, we note that among those who reported fall location, 44% fell at home, of whom nearly half fell in their bathroom. This underscores the importance of home modifications for fall prevention (Pynoos et al., 2010; Pynoos et al., 2012) and has implications for the design of new PSH projects. Future studies could also address whether modifications in and around PSH buildings are needed and feasible to reduce fall risk (Gutman et al., 2017). Interventions that address both individual and environmental risk factors for falls may be most effective, such as the Community Aging in Place for Better Elderly Living model that involves clinical interventions from a nurse and occupational therapist along with home modifications (Szanton & Gitlin, 2016), which has not been tested in PSH. As a result of this study, we are planning to pilot-test this model to determine what adaptations, if any, are needed in PSH for a population that has experienced a lifetime of cumulative adversity (Padgett, Smith, Henwood, & Tiderington, 2012).

Limitations

Although this study represents an important step toward planning for the changing needs of people who have experienced chronical homeless and accelerated aging (Brown, Thomas, Cutler, & Hinderlie, 2013; Henwood et al., 2019), we note several limitations. First, fall related questions were self-report and due to difficulties with recall are likely to reflect an undercount of actual falls (Griffin et al., 2019). Second, the study is cross-sectional, and thus we report associations but could not examine causation; future work may employ a longitudinal design to better understand factors contributing to fall risk. Third, we did not measure some important potential fall risk factors, including various types of medication; given that the majority of participants have a lifetime diagnosis of mental illness, it is likely that many were taking psychotropic medications that are associated with falls (Hartikainen, Lonnroos, & Louhivuori, 2007). While PSH often provides medication management, it was not assessed in this study. Also, not all of our study measures hypothesized as correlates of past-year falls were assessed using the same past-year timeframe (e.g., substance use was only measured in the past 30 days); changes in participant activity and behavior may therefore account for some of the null findings between these measures and falls. Finally, although the study used validated measures, self-report items such as past diagnoses and substance use were not confirmed or objectively measured, although previous research showed that self-reports of homeless adults are reliable (Gelberg & Siecke, 1997).

Conclusion

This study demonstrates that fall prevention is needed in PSH. Given the high rate of injury and costs associated with medical care from these injuries (Bergen, 2016), fall prevention efforts in PSH have the potential to improve quality of life while reducing health care utilization. Although fall prevention strategies that are cost effective and easily deployed have been developed (Pynoos et al., 2012), no known studies have explored applying these strategies to tenants living in PSH, which is becoming increasingly important give that the target population is an aging cohort with an average age approaching 60 years old (Culhane, Metraux, Byrne, Stino, & Bainbridge, 2013). Evidence-based home modification and clinical interventions (Szanton & Gitlin, 2016; Stark et al., 2018) may have to be adapted for PSH tenants, who have experienced cumulative adversity and accelerated aging (Brown et al., 2013; Henwood et al., 2019; Padgett et al., 2012; Padgett, Tiderington, Smith, Derejko, & Henwood, 2016).

What is known about this topic:

• Adults who have experienced chronic homelessness represent a stable, aging cohort whose average age is approaching 60 years old

• Permanent supportive housing effectively ends chronic homelessness.

• Older adults who experience chronic homelessness prematurely age and have high rates of falls.

What this paper adds:

• PSH tenants have a high prevalence of falls and serious fall-related injuries are common.

• Individual risk factors including functional impairment, frailty, and persistent pain are associated with increased fall risk in PSH tenants

• More research on environmental risk factors for falls in PSH is needed.