Service Availability in Assisted Living and Other Community-Based Residential Settings at the End of Life

Abstract

Background: Community-based residential settings (e.g., assisted living facilities and retirement communities), are increasing, where vulnerable older adults are living as they age and die. Despite prevalent serious illness, functional impairment, and dementia among residents, the mix and types of built-in services available are not known.

Objective: To classify older adults in community-based residential settings by the types of services available and examine associations between service availability and hospice use and location of death.

Design: Pooled cross-sectional analysis.

Setting: Medicare Current Beneficiary Survey data (2002–2018).

Subjects: U.S. adults 65 years of age and older, who lived in a community-based residential setting and died between 2002 and 2018 (N = 1006).

Measurements: Availability (yes/no) of nursing care, medication assistance, meals, laundry, cleaning, transportation, and recreation.

Results: Our sample resided in assisted living facilities (32.0%), retirement communities (29.0%), senior citizen housing (13.7%), continuing care facilities (13.5%), and other (11.8%). Four classes of individuals with distinct combinations of available services were identified: 48.2% lived in a residence with all measured services available; 29.1% had availability of all services, except nursing care and medication assistance; 12.6% had availability of only recreation and transportation services; and 10.1% had minimal/no service availability. Of the 51.8% of older adults residing in settings without clinical services, more than half died at home and fewer than half died with hospice.

Conclusions: The majority of older adults who die in community-based residential settings do not have access to built-in clinical services. Palliative care training for staff in these settings may be warranted, given variable rates of hospice use and high rates of home death.

Introduction

Community-based residential settings enable older adults to live and die in the community, despite complex health needs and functional impairment.¹ These settings include assisted living facilities, retirement communities, continuing care facilities, and senior citizen housing, among other types, and exclude private residences and nursing homes. The use of these alternatives to institutional long-term care has the potential to improve quality of life and promote function and independence for individuals as they age. An estimated 28,900 community-based residential settings existed in the United States in 2017.² The National Academies of Science, Engineering, and Medicine's most recent report³ highlighted that these care settings are critical to the quality of life and care of those with serious illness and provide important residential alternatives to nursing home care.

Despite the proliferation of community-based residential settings and the vulnerability of those who reside in these settings, there is no uniform requirement regarding availability of services, staff training, admission policies, or provision of third-party services such as hospice. States have the authority to regulate some types of assisted living facilities, yet even within assisted living facility regulation, there is substantial variability in requirements and oversight.^4,5 For example, in Alabama, individuals cannot require skilled nursing care or hospice services if living in an assisted living facility, whereas in Utah, facilities must provide personal care, food service, housekeeping, laundry, maintenance, activity programs, assistance with self-administration of medication, and arrange for necessary medical care, including hospice.⁴ Consequently, standard definitions of the mix of services provided within each type of community-based residential setting (e.g., within assisted living facilities or within retirement communities) do not exist.

As the number of individuals living in community-based residential settings continues to increase alongside emerging evidence, suggesting substantial functional impairment and rates of serious illness including dementia, it is critical to understand the extent of built-in services that exist to support residents.^1,3 Our knowledge gap regarding service availability in community-based residential settings is particularly important, given that an estimated 10% of older adults who die each year in the United States reside in these settings.¹ Although hospices play an integral role in supporting individuals at the end of life in community-based residential settings,^1,6–8 the extent to which hospice complements existing built-in services or fills gaps in care is important to understand to ensure that the needs of older adults living in these settings are met.

Thus, our goal was to classify older adults in community-based residential settings by the mix of services available and examine associations between service availability and hospice use and location of death. We analyzed data from the nationally representative Medicare Current Beneficiary Survey (MCBS) regarding older adults who were living in residential community settings and died between 2002 and 2018. Study results will inform improved end-of-life care planning for older adults and may help identify service gaps in supporting older adults living in community-based residential settings at the end of life.

Methods

Study design and sample

We conducted a pooled cross-sectional study using data from the MCBS from 2002 to 2018. These data exclude survey results from 2014, which were not released by Centers for Medicare & Medicaid Services (CMS). The MCBS employs a rotating panel design. The stratified, three-stage area probability sample is representative of the Medicare population in total and within age groups. Each sampled beneficiary is interviewed in person up to three times per year over a four-year period. A key advantage of the MCBS is that it conducts interviews in the community and in long-term care facilities, allowing a continuous record regardless of the location of the respondent. The MCBS data are linked to Medicare claims data for Traditional Medicare beneficiaries.

Our sample consists of MCBS participants who self-/proxy reported living in a community-based residential setting and who died between 2002 and 2018. We define residential setting at the end of life as where an individual self-/proxy reported residing at the time of death and is distinct from location of death. Consistent with prior work using MCBS,⁹ we broadly defined community-based residences as places that are likely to provide long-term supportive services in a home-like setting. Our goal is to examine older adults who are living across a diverse and emerging group of community settings to better understand the services and support that they receive as they age. Respondents chose from the following list of settings: assisted living facility, retirement community, continuing care community, senior citizen housing, retirement apartment, personal care home, board and care home, rest/retirement home, staged living community, domiciliary, church-provided housing, and other. On average, individuals moved into their last residence 140 days before death. MCBS respondents who reported their last residence as a private residence or a nursing home, rehabilitation facility, hospital, inpatient hospice, or mental health facility were excluded from our sample.

Measures

MCBS surveys are conducted in-person, three times per year. All measures are self-reported or proxy reported. All measured variables are assessed either as of an individual's last MCBS interview date before death (15.6% of our sample and occurred an average of 35.6 days before death) or the after-death proxy interview (84.4% of our sample and occurred an average of 69.6 days following death).

Availability of services

We used seven dichotomous measures of service availability based on respondent self-report of whether or not each of the following services was available at their residence: nursing care, medication assistance, meals, laundry, cleaning, transportation, and recreation.

Demographic, clinical, and socioeconomic characteristics

We measured age at death, gender, race/ethnicity (white, black, Hispanic, Asian, or other), education (less than high school, high school, some college, or college graduate), marital status (married or not married), Medicaid coverage (yes/no), and region using self-report. We measured medical conditions using self-report. For our measure of dementia, we used an inclusive case definition developed for use with the MCBS data.¹⁰ We measured functional status based on self-/proxy-reported difficulty with six basic activities of daily living (ADL) (i.e., walking, feeding, dressing, toileting, bathing, and transferring).

Hospice use and location of death

We define hospice care as interdisciplinary care provided by hospice agencies under the Medicare Hospice Benefit. We used Medicare hospice claims linked to MCBS data to identify hospice enrollment by MCBS participants. We used MCBS survey data in conjunction with linked Medicare claims data to identify location of death, measured as home (i.e., the community-based residential setting), hospital, nursing home, or inpatient hospice.

Analysis

We used latent class analysis to identify homogeneous groups of respondents based on the availability of services at their community-based residential setting.¹¹ These statistical models are appropriate when one posits that a population comprises two or more underlying, latent subgroups defined by the intersection of multiple observed characteristics (in this case, availability of the seven measured services).^11,12 Particular combinations of services define the latent subgroups. To identify the smallest number of classes needed to account for patterns of service availability, the number of latent classes was determined iteratively beginning with a one-class model and testing models of increasing numbers of classes with robust maximum likelihood estimation. The optimal number of classes was determined on the basis of goodness of fit statistics (i.e., Bayesian information criterion and Akaike information criterion).

Respondents were assigned to the class for which they had the highest probability of membership. For each class, the prevalence of respondents and the probability of each service being available are reported. Conditional probabilities of 0.70–1.00 were considered to indicate a high probability of service availability; 0.40–0.69, a moderate probability; and <0.40, a low probability. All estimates reported are weighted. All analyses take into account the sample weights and survey design variables available in MCBS.

We used descriptive statistics to characterize residents of each class. We then used sampling weights to determine population estimates of people self-reporting each type of residence, proportion of residents in each class enrolled with hospice, and location of death. This project was determined exempt from the Mount Sinai IRB.

Results

Number of latent classes

We chose a four-class solution according to measures of model fit and because it was most conceptually meaningful. The analysis revealed four classes of beneficiaries with substantially different services available where they lived (Fig. 1). The largest class (“All Services”) contained 48.2% of the sample and comprised people who lived in settings where essentially all measured services were available. The next largest class (“Most Services Excluding Clinical”) contained 29.1% of the sample. These were people who lived in a setting with high probability of availability of services such as cleaning, meal preparation, laundry, recreation, and transportation, but low probability of clinical services such as medication management and nursing care. The third class (“Recreation- and Transportation-Based Services”) contained 12.6% of the sample. These people lived in a setting in which the primary services available were recreation and transportation. The fourth class (“Minimal Services”) contained 10.1% of the sample and comprised those who reported limited/no service availability.

FIG. 1.

Four classes generated according to service availability in community-based residential settings, 2002–2018. Source: Authors' analysis of data for 2002–2018 (excluding 2014) from the Medicare Current Beneficiary Survey. Community-based residential settings include assisted living facilities, retirement communities, continuing care communities, senior citizen housing, retirement apartments, personal care homes, board and care homes, rest/retirement homes, staged living communities, domiciliaries, and church-provided housing, and exclude private residences and nursing homes.

Characteristics of class members

The 1006 respondents who reported living in community-based residential settings in the year they died correspond to a population estimate of 2.6 million individuals, or ∼160,000 individuals per year. The average age at death was highest in the All Services class (86.1 years old) (Table 1). Gender and marital status differed across the classes. Unmarried women comprised 53.3% of the All Services class, 55.4% of the All Services Except Clinical class, 44.2 of the Recreation/Transportation class, and 29.1% of the Minimal Services class. Married females made up fewer than 10% of each class and did not significantly differ across classes.

Table 1.

Characteristics of Medicare Beneficiaries Living in Community-Based Residential Settings at the End of Life, 2002–2018

	Latent class generated according to service availability				p
	All services, N = 1.04 million	All services, except clinical, N = 927,000	Recreation and transportation services, N = 402,000	Minimal/no services, N = 264,000
Demographic and socioeconomic characteristics
Age (mean)	86.1	85.5	80.4	80.6	<0.001
Gender and marital status
Female unmarried	53.3	55.4	44.2	29.1	0.001
Male unmarried	23.9	20.8	20.2	27.5	0.633
Female married	7.9	8.5	5.6	9.5	0.719
Male married	14.9	15.3	30.1	33.9	<0.001
Race/ethnicity
Non-Hispanic white	92.0	88.9	86.3	80.6	0.081
Non-Hispanic black	3.0	4.0	4.8	1.9	0.750
Hispanic	2.8	4.2	1.6	6.7	0.365
Other/multiracial	2.1	2.8	7.4	10.9	0.018
College graduate	20.4	25.2	14.6	10.5	0.041
Medicaid coverage	26.7	19.9	36.0	25.7	0.061
Clinical and functional characteristics
Serious illness
Dementia	76.0	47.3	16.9	36.6	<0.001
Heart disease	44.5	49.0	54.5	43.5	0.224
Stroke	22.2	22.7	18.5	23.4	0.818
Lung disease	22.5	32.0	28.7	33.4	0.125
Cancer	31.3	46.0	40.7	44.6	0.015
Diabetes	31.5	33.5	32.4	48.4	0.047
Receive help with 3+ADLs	81.3	64.4	49.9	41.7	<0.001
Geographic characteristics
Metropolitan area	83.9	89.6	94.3	87.0	0.101
Region
Northeast	12.9	14.1	33.3	24.2	0.022
Midwest	29.6	28.4	17.6	8.3	0.008
South	32.8	28.8	28.2	37.7	0.616
West	24.7	28.6	20.9	29.8	0.565

Medicare Current Beneficiary Survey data are not available for 2014. Community-based residential settings include assisted living facilities, retirement communities, continuing care communities, senior citizen housing, retirement apartments, personal care homes, board and care homes, rest/retirement homes, staged living communities, domiciliaries, and church-provided housing, and exclude private residences and nursing homes.

ADL, activities of daily living (bathing, dressing, transfers, toileting, walking, and eating).

In terms of serious illness prevalence and functional impairment, the All Services class had the highest proportion of residents with dementia (76.0%) and with three or more ADL impairments (81.3%). In the All Services Except Clinical class, the most prevalent serious illnesses were heart disease (49.0%) and dementia (47.3%). and 64.4% had three or more ADL impairments. In the Recreation/Transportation class, the most prevalent serious illnesses were heart disease (54.5%) and cancer (40.7%), and 49.9% had three or more ADL impairments. In the Minimal/No services class, the prevalence of serious illnesses were more evenly distributed, including 48.4% with diabetes, 44.6% with cancer, and 43.5% with heart disease, and 41.7% living with three or more ADL impairments.

Relationship between self-reported residence and class membership

Overall, 32.0% of our sample reported living in an assisted living facility (52,700 individuals per year), followed by those reporting living in a retirement community (29.0%, or 47,600 individuals per year), continuing care facility (13.5%, or 22,200 individuals per year), senior citizen housing (13.7%, or 22,500 individuals per year), and personal care home (3.8%, or 6200 individuals per year), and fewer than 4% in each of the following: board and care home, rest home, adult/group home, stages living community, and domiciliary.

Where a respondent reported living was an imprecise marker for service availability (Fig. 2). For example, 63.1% of those living in assisted living facilities were in the All Services class, 34.2% were in the All Services Except Clinical class, 1.2% were in the Recreation/Transportation class, and 1.5% were in the Minimal/No Services class. Those who reported living in retirement communities were also represented across all four latent classes. Specifically, 6.4% of those living in retirement communities were in the All Services class, 39.8% were in the All Services Except Clinical class, 25.2% were in the Recreation/Transportation class, and 28.6% were in the Minimal/No Services class.

FIG. 2.

Relationship between self-reported residence and class membership. This figure depicts the proportion of individuals in each type of community-based residential setting with access to each of the four classes of services.

Relationship between class membership and hospice use and location of death

Hospice use was highest for those in the All Services class (59.3%, compared with 50.0% in the All Services Except Clinical class, 40.5% in the Recreation/Transportation class, and 44.8% in the Minimal/No Services class) (p = 0.010) (Table 2). The duration of hospice enrollment was highest in the All Services class (mean 81.9 days and median of 19.0 days) followed by the All Services Except Clinical class (mean 66.5 days and median 18.0 days), Recreation/Transportation class (mean 51.7 days and median 10.0 days), and the Minimal/No Services class (mean 27.6 days and median 6.0 days).

Table 2.

Hospice Use and Location of Death of Medicare Beneficiaries Living in Community-Based Residential Settings at the End of Life, 2002–2018

	Latent class generated according to service availability
	All services, N = 1.04 million	All services, except clinical, N = 927,000	Recreation and transportation services, N = 402,000	Minimal/no services, N = 264,000
Hospice use	59.3%	50.0%	40.5%	44.8%	*
Location of death
Home	70.2	54.9	50.6	60.6	**
Hospital	14.7	29.4	35.1	31.3	*
Nursing home	12.6	12.4	7.3	4.4
Inpatient hospice	2.5	3.3	7.0	3.6

^* p < 0.05; ^**p < 0.01 for chi-squared comparisons across latent class groups. Community-based residential settings include assisted living facilities, retirement communities, continuing care communities, senior citizen housing, retirement apartments, personal care homes, board and care homes, rest/retirement homes, staged living communities, domiciliaries, and church-provided housing, and exclude private residences and nursing homes.

For more than half of our sample, the location of death was their home (i.e., the community-based residential setting) (Table 2). The highest rate of home death was for those in the All Services class (70.2%) and this class also had the lowest rate of death in a hospital (14.7%). The proportion of individuals who died in a nursing home or an inpatient hospice did not significantly differ by latent class. Of the three classes residing in settings without clinical services, representing 51.8% of older adults in our sample, more than half died at home and fewer than half died with hospice.

Discussion

The estimated 160,000 Medicare recipients living in community-based residential settings at the end of life each year experience a wide range of available services that can be categorized into four distinct classes. Only one of these classes includes clinical support, despite high rates of dementia and functional impairment across classes. In addition, the type of residential setting (for example, “assisted living facility” or “retirement community”) did not consistently correspond to the types of services provided in those settings.

Importantly, fewer than half of individuals were living in a community-based residential setting in which clinical care (nursing and medication management) was available at the end of life. High rates of serious illness, dementia, and functional impairment among inhabitants of community-based residential settings were apparent in our study. While those living in settings without clinical care tended to be healthier and less impaired, they still had significant care needs. For example, 42% of those living in settings with Minimal/No Services available were impaired in three or more ADLs. As nursing homes continue to shift toward caring for those with the most significant needs, acuity level of populations living in all types of community-based residential settings will likely increase. Given this shift, further exploration of whether or not community-based residential settings adequately meet the clinical needs of their residents is essential. In particular, this inquiry should focus on how community-based residential settings respond when the health and function of residents decline.

Among community-based residential settings with availability of clinical services, more than half of residents enrolled with hospice. In fact, hospice use was most prevalent among those in community-based residential settings with All Services, likely reflecting the older, more impaired population living there. It may also be that community-based residential settings with greater service availability are more likely to have regulations requiring access to hospice care.

For those living in community-based residential settings without clinical services, the majority of residents did not enroll with hospice at the end of life, despite the fact that more than half died at home. This suggests that a significant proportion of Medicare beneficiaries live in residential settings with little to no clinical support at the end of life. Future research on potential barriers to hospice enrollment in these settings is needed. In addition, expansion of nonhospice palliative care to these settings may also help ensure that residents receive the support they need as they age.

Our results also underscore the need to broaden our understanding of what encompasses community-based residential settings beyond assisted living facilities. Although previous research has demonstrated variability in quality of care and end-of-life outcomes within assisted living facilities,^6–8,13 our study found that only one-third of individuals in community-based residential settings reported living in assisted living facilities. In addition, assisted living facilities themselves were heterogeneous with approximately one-third lacking availability of clinical services. Future study of community-based residential settings should consider the wider range of residence types included in our study and carefully characterize available services within these settings.

Limitations of our study include only examining the availability of services in residential settings and not the actual use of such services. Similarly, we did not assess the use of home-based medical care or visiting nurse services, which may complement the services provided by residences. We were also unable to identify and statistically adjust for whether or not individuals were from the same residential setting, although given that MCBS is a national sample, the likelihood of more than one resident from a given location is low. Finally, we cannot assess individuals' preferences for service availability or location of death.

Conclusion

Community-based residential settings play a key role in enabling the growing population of older adults with serious illness and functional impairment to remain in their communities, age in place, and die at home. Yet the range of services available to residents at the end of life varies and does not consistently correspond with the residence type. These findings underscore the importance of a nuanced understanding of available services to ensure older adults and their families have sufficient support at the end of life. Our results suggest the need for greater standardization of available services and access to outside services such as hospice, particularly in community-based residential settings that lack built-in clinical support. In addition, palliative care training for staff in residences without built-in clinical care may be warranted, given high rates of home death.

Authors' Contributions

All authors were involved in the study concept and design. All authors were involved in analysis and interpretation of data. All authors were involved in preparation of article.

Presented at the American Academy of Hospice and Palliative Medicine Annual Research Meeting 2019 and Academy Health Annual Research Meeting 2019.

Funding Information

Claude D. Pepper Older Americans Independence Center pilot award, National Institute of Nursing research R01NR018462.

Funders played no role in the study's design, methods, data collection, analysis, and preparation of the article.

Author Disclosure Statement

No competing financial interests exist.