Trends in Serious Mental Illness in US Assisted Living Compared to Nursing Homes and the Community: 2007-2017

Abstract

Objectives:

Little is known about the prevalence of serious mental illness (SMI) in assisted living (AL) communities in the United States. Trends in the prevalence of SMI in the AL community were examined over time and in relationship to characteristics such as dual eligibility and health conditions. Within- and between-state variability of SMI in AL was also examined.

Design:

Samples of Medicare beneficiaries who lived in the 48 contiguous states were created: those who resided in the community, in a nursing home (NH), and in AL on December 31st of each year (2007-2017). We conducted univariate analysis to display the trends in SMI over time in AL compared to NHs and the community. To demonstrate intrastate variability, we examined the prevalence of SMI for each state. We described within-state variability using a Lorenz curve and GINI coefficients.

Results.

The prevalence of SMI in AL increased by 54%, rising from 7.4% in 2007 to 11.4% in 2017. Residents with SMI were more likely to be dually eligible for Medicare and Medicaid than residents without SMI. The prevalence of SMI in AL ranged from to 3.2% in Wyoming to 33.1% in New York. Approximately 10% of AL communities had over half of the sample’s AL residents with SMI.

Conclusions.

Given the increased proportion of residents with SMI in AL, research is needed into the mental health and social care needs of this population. Research is needed to uncover reasons for variations across and between states.

Introduction

Over 800,000 residents live in assisted living (AL) communities in the United States (1). AL communities provide housing, personal care, at least two meals a day, and oversight 24 hours a day, but are not required to provide nursing services. AL settings were historically referred to as “facilities,” but that term has fallen out of favor due to the institutional connotation of that term. The preferred term is now “communities.” Although AL communities are intended to be more “home-like” than nursing homes (2) they are home to residents with neuropsychiatric care needs. Approximately 71% of AL residents have cognitive impairment (3), 28% have depression (1), and as of 2010, 8% had serious mental illness (SMI) (4), defined in this study as bipolar disorder or schizophrenia. Newer figures regarding SMI in AL are unknown, and various factors may have caused the prevalence to increase over time. After the move to deinstitutionalize the population residing in state psychiatric institutions in the 1960s and 1970s, many NHs have become home to a large proportion of residents with SMI (5). Because many individuals with SMI do not require nursing care, it is natural to expect that they have also moved to AL.

Previously, though, individuals with SMI may have not had access to AL given that many are Medicaid recipients and that the AL industry (separate from historic small “mom and pop” homes) traditionally has served individuals paying privately for their care (6). After Olmstead v. L.C. in 1999, public entities were mandated to provide programs in the least restrictive setting possible to individuals with disabilities (7). States responded in varying degrees by reallocating Medicaid funding toward home and community-based service (HCBS) such as home health, personal care services, and AL waivers (7). Subsequently, the number of low-income older adults dually enrolled in Medicare and Medicaid served in AL has increased dramatically, from an estimated 60,000 individuals in 2000 to over 300,000 in 2014 (8). Currently, 48 states allow for Medicaid covered services in AL, up from 21 states in 1996 (8). Thus, dually-eligible beneficiaries with SMI may increasingly have access to AL (6).

Understanding the prevalence of SMI in AL is important because AL residents who exhibit behaviors common to those with SMI diagnoses (e.g., agitation and aggression) face stigma and isolation from other residents and require appropriate care (9). Unfortunately, AL staff are often untrained in how to best manage these behaviors (9,10,11), which may explain why AL residents with SMI had a shorter time to discharge to a NH than those without a diagnosis (12). Data related to NHs are further informative, in that individuals with SMI are admitted to lower quality nursing homes than residents without SMI (13,14). In addition, residents who reside in NHs with high proportions of residents with SMI are more likely to receive worse quality of care (15,16,17,18), which is believed to relate to stressed resources (17,18). Having a better understanding of the prevalence of SMI in AL can help call attention to potential resource and care needs for this population.

Also important to understand is state variability in SMI prevalence, given that AL communities are state regulated and there are no federal requirements to identify or treat residents with SMI as there are for NHs (9). States determine and enforce regulations related to admission, retention, and care needs of AL residents (19), which likely contributes to geographic variability in SMI among AL residents. Differences in other state policies, such as Medicaid eligibility, likely also contribute to between-state variability. In addition, there is likely intrastate variability given that states have more than one licensure categories for AL, often reflecting different levels of service intensity provided and populations served (20). Finally, individual AL communities differ regarding whether they accept Medicaid residents (17).

Currently, there is no information regarding historical trends and state variability of the prevalence of SMI in AL. Therefore, the goal of this study is to examine the prevalence of SMI in AL, as well as whether it has changed over time, and in relationship to characteristics such as dual eligibility and health conditions. We compare rates to community and NH samples to ensure that increases in prevalence are not simply a reflection of increased recognition of SMI among healthcare providers. We also explore between- and within-state variability of SMI in AL.

Methods

Data

We obtained data from the Medicare Master Beneficiary Summary File (MBSF), a ZIP Code History file, OASIS home health assessment data, Medicare Part B claims, a national list of state licensed AL communities, and a Residential History File (RHF). The MBSF provided demographic information such as the beneficiary’s race and age, and the Chronic Condition Warehouse (CCW) section of the MBSF provided information about chronic conditions (21). The addresses of the AL communities came from a national list of state licensed settings collected by our study team. The ZIP code history file provided a person’s ZIP code for each day within a calendar year. The Residential History File provided a person’s location of care per day within each calendar year using a combination of Minimum Data Set assessments, Medicare claims, and the Home Health Outcome and Assessment Information Set (22).

We used a validated methodology to identify AL communities with at least 25 beds (23), which has been applied in other studies (24,25,26). Using a combination of home health OASIS assessment data, Part B claims, the ZIP code history file, and our list of licensed AL communities, we created a finder file of unique 9-digit ZIP codes representing a licensed AL community (23). This finder file was linked to the MBSF. This methodology makes it only possible to accurately identify residents living in large ALs. Consistent with previous work, we defined a large AL as a community of 25+ beds licensed to serve an older adult population (23,27). For additional explanation of the validated methodology to identify large ALs, please see the Appendix.

Sample

We created cross-sectional samples of Medicare beneficiaries who lived in the 48 contiguous states: those who resided in the community, in a NH, and in AL on December 31st of each year (2007-2017). We defined our community cohorts as individuals who did not live in NHs or 25+ bed AL communities. Our AL community cohorts consisted of individuals who lived in 25+ bed AL communities.

To be included in our study, individuals had to be alive, not in a hospital on December 31st, and not enrolled in MA. To identify our community cohorts, we used a random 5% sample of Medicare beneficiaries in each year. We excluded individuals residing in a NH and individuals in AL communities. To identify our AL samples, we excluded individuals residing in NHs and in the community. To identify our NH samples, we excluded individuals who were residing in large ALs and individuals in the community. We additionally excluded residents in ALs and NHs with fewer than 10 residents per year and those that did not remain in operation throughout the time period, resulting in residents living in 5,961 ALs and 14,059 NHs.

The CCW included measures for chronic physical and mental health conditions that were created using algorithms that searched Medicare claims data for specific diagnosis codes. Consistent with previous research in the nursing home setting (15,16) and using measures from the CCW (21), we operationalized SMI as a diagnosis of schizophrenia or bipolar disorder. Our measure of schizophrenia included major types of schizophrenia as well as schizoaffective disorder and schizophreniform disorder. Our bipolar disorder measure included diagnosis for bipolar I, bipolar II, and “other” bipolar disorders. For information regarding the diagnosis codes used, please see Supplementary Table 1 in the Appendix.

For descriptive analysis of AL residents, we included an indicator of whether the beneficiary was dually enrolled in Medicare and Medicaid on December 31st of 2007 and 2017. We also reported on individuals’ age group (≤64, 65-74, 75-84, 85-94, and 95+), sex, and reason for Medicare eligibility. We examined the prevalence of the diagnosis of the following conditions given their differential prevalence among individuals with SMI (28,29):Alzheimer’s Disease and Related Dementias (ADRD), atrial fibrillation, cancer, chronic kidney disease, chronic obstructive pulmonary disease, diabetes, heart failure, hyperlipidemia, hypertension, ischemic heart disease, obesity, and stroke .

Analysis

We conducted univariate analysis to display the trends in the point prevalence of residents with SMI on December 31st of each year for the AL, nursing home, and community samples. We also described characteristics (e.g., age, sex, and dual status) of residents in AL who had SMI compared to those who did not during the years 2007 and 2017 using χ² tests of statistical significance. Additionally, we examined the proportion of AL residents with SMI in the United States for each state to demonstrate interstate variability. We used a Lorenz curve to understand the distribution of SMI across ALs in the United States. This curve is often used to describe inequalities, and depicts the relationship between the cumulative percentage of AL residents on the x-axis and the cumulative proportion of residents with SMI on the y-axis. The line of equality is drawn with a slope of 1, which represents a perfectly even distribution. Values further away from this line indicate an unequal distribution, representing high concentrations of SMI within a few communities. Finally, we calculated GINI coefficients for each state to describe interstate variation in the proportion of individuals with SMI in each state; the GINI coefficient is a measure of inequality that is expressed as a ratio of the area between the Lorenz curve and the perfect equality line divided by the total area under the equality line (30). Values closer to 1 indicate more uneven distributions of SMI within states.

Analyses were conducted with SAS v.9.4, and STATA v.16.1. Additional information about the data and methods used for these analyses can be found in the (BLINDED).

Results

Trends in the prevalence of SMI in three settings

As shown in Figure 1, the prevalence of SMI increased in all three settings over the 10-year study period. The prevalence of SMI in AL increased by 54%, rising from 7.4% in 2007 to 11.4% in 2017. The increase was the most marked for the nursing home population, with the prevalence increasing 77%, from 10.5% in 2007 to 18.6% in 2017. The prevalence of SMI in the community increased by 39%, rising from 2.8% in 2007 to 3.9% in 2017. The prevalence of SMI in AL was consistently lower than the NH prevalence and greater than the community prevalence. The total number of AL residents with SMI grew from 15,592 in 2007 to 26,747 in 2017, a 72% increase. Approximately 73% of AL communities housed at least 1 resident with SMI. The median percentage of residents in our sample with SMI in each community was 5.6% with an interquartile range of 14.

FIGURE 1.

Changes Over Time in the Percentage of Medicare Beneficiaries with Serious Mental Illness, by Residence (2007-2017)

Notes: Data from the 2017 Medicare Master Beneficiary Summary File and Chronic Conditions Segment. Samples represented fee-for-service Medicare beneficiaries in each setting on December 31 of each year. For the nursing home and assisted living samples, we only included residents in continuously operating AL communities (i.e., open from 2007-2017).

Characteristics of AL Residents with SMI

Table 1 shows that residents with SMI were much more likely to be in a younger age group category; in 2017, 36% of AL residents with SMI were under age 65 compared to 5% of AL residents without SMI (the proportion younger than 65 among the SMI group was even higher in 2007 [43%]). Beneficiaries with SMI were also more likely to be nonwhite and Hispanic. Further, individuals with SMI were much more likely than residents without a diagnosis of SMI to be dually enrolled in Medicare and Medicaid: compared to about 20% of residents without SMI, approximately 75% of residents with SMI were dually enrolled in both 2007 and 2017. Residents with a diagnosis of SMI were also more likely to be eligible for Medicare due to disability rather than age when compared to residents without an SMI diagnosis. The prevalence of chronic conditions varied by diagnosis of SMI. For example, residents with SMI were more likely to have COPD (28% and 32% in 2007 and 2017, respectively) compared to residents without SMI (14% and 16%). Residents with SMI were also more likely to have a diagnosis of obesity or diabetes than residents without a diagnosis of SMI. Approximately 30% of residents with SMI had 6 or more chronic conditions compared to 20% of residents without SMI.

TABLE 1.

Characteristics of Medicare Beneficiaries in Assisted Living, by Presence of a Serious Mental Illness Diagnosis, by Year (2007 vs 2017)

	2007				2017
	Residents with SMI (n=15,592 )	Residents without SMI (n=221,857)	Pearson Chi Square (df)	p valuea	Residents with SMI (n=26,747)	Residents without SMI (n=208,315)	Pearson Chi Square (df)	p valuea
Age group			56,625.3 (4)	p<.001			39,845.8 (4)	p<.001
<65	43.4	3.8			35.6	4.7
65-74	23.8	9.5			30.6	14.6
75-84	21.5	32.4			20.1	25.8
85-94	10.4	47.3			12.0	44.6
95+	0.9	7.0			1.8	10.3
Sex (%)			2,152.5 (1)	p<.001			1,864.7 (1)	p<.001
Male	43.4	28.3			45.3	32.1
Female	56.6	71.8			54.7	67.9
Race (%)			7,321.2 (3)	p<.001			7,073.6 (3)	p<.001
White	79.9	93.8			76.7	91.6
Black	12.2	2.9			14.1	3.8
Hispanic	5.1	1.5			5.8	2.0
Other	2.8	1.8			3.4	2.5
Dually eligible for Medicare and Medicaid (%)	75.5	20.9	35,590.4 (1)	p<.001	75.1	19.9	15,830.7 (1)	p<.001
Reason for original entitlement			48584.1 (2)	p<.001		41266.3 (2)		p<.001
Agedb	55.3	96.0			38.6	88.4
Disabledb	44.5	3.9			61.3	11.5
End Stage Renal Disease only	0.2	0.1			0.1	0.1
Chronic Conditions (%)
Alzheimer’s Disease and Related Dementias	30.8	26.1	170.5 (1)	p<.001	48.9	34.9	2,008.8 (1)	p<.001
Atrial Fibrillation	4.7	14.5	1,162.2 (1)	p<.001	8.7	17.7	1,368.5 (1)	p<.001
Cancer	5.1	9.5	339.6 (1)	p<.001	6.2	10.3	455.2 (1)	p<.001
Chronic Kidney Disease	12.8	14.4	29.2 (1)	p<.001	38.9	33.4	325.9 (1)	p<.001
Chronic Obstructive Pulmonary Disease	27.6	13.6	2,308.8 (1)	p<.001	32.4	15.9	4,392.8 (1)	p<.001
Diabetes	35.9	23.2	1,275.9 (1)	p<.001	46.7	26.9	4,504.8 (1)	p<.001
Heart Failure	21.3	27.9	318.6 (1)	p<.001	27.4	26.9	3.6 (1)	p=.06
Hyperlipidemia	42.1	41.0	7.6 (1)	p=.01	56.0	51.8	164.3 (1)	p<.001
Hypertension	60.7	67.2	274.1 (1)	p<.001	75.6	72.6	110.6 (1)	p<.001
Ischemic Heart Disease	34.8	40.4	186.6 (1)	p<.001	39.6	37.6	41.7 (1)	p<.001
Obesity	13.1	2.7	4,947.3 (1)	p<.001	26.9	12.6	3,935.2 (1)	p<.001
Stroke	5.2	6.9	65.1 (1)	p<.001	7.6	7.5	0.05 (1)	p=08
Number of Chronic Conditions			121.3 (3)	p<.001			1,705.6 (3)	p<.001
Fewer than 2 Chronic Conditionsc	28.8	27.5			15.5	22.2
2-3 Chronic Conditionsc	34.0	37.4			26.4	31.1
4-5 Chronic Conditionsc	24.2	24.3			28.4	26.4
6+ Chronic Conditionsc	12.9	10.7			29.8	20.3

Notes: Data come from the 2017 Medicare Master Beneficiary Summary File and Chronic Conditions dataset. Individuals were enrolled in fee-for-service Medicare. Residents were in continuously operating assisted living settings on December 31 of each year.

^a.Comparison was between SMI and non-SMI

^b.Included those with diagnosis of End Stage Renal disease

^c.Condition counts were created from aforementioned listed conditions

Between and Within-State Variability of SMI in AL

We observed significant interstate variability in the prevalence SMI in AL (Figure 2). The prevalence ranged from to 3.2% in Wyoming to 33.1% in New York. The five states with the highest rates of SMI in AL were New York, North Carolina, Indiana, Connecticut, and Oklahoma, and the five states with the lowest rates (all less than 5%) were Wyoming, Delaware, Arizona, Oregon, and Montana. We also observed significant intrastate variability in the prevalence of SMI in AL communities. Figure 3 shows notable inequality based on the deviation from the line of equality; in this context, results indicated that 10% of AL communities had over half of the sample’s AL residents with SMI. In Figure 4, we observed that in many cases, states with higher proportions of SMI also had greater inequality in the distribution of SMI across AL communities. New York, Indiana, and Oklahoma, for example, had GINI coefficients above .5 and a 19-33% prevalence of SMI in AL.

FIGURE 2.

State Rates (%) of Serious Mental Illness among Medicare Beneficiaries Residing in Assisted Living in 2017 (n=235,062)

Notes: Data were from the 2017 Medicare Master Beneficiary Summary File and Chronic Conditions dataset. Residents were in assisted living on December 31, 2017. Individuals were enrolled in fee-for-service Medicare. Residents were included if they resided in continuously operating settings between 2007-2017.

FIGURE 3.

Lorenz Curve of the Proportion of Residents with Serious Mental Illness Plotted Against the Percentage of Assisted Living Settings in the Sample

Notes: Data were from the 2017 Medicare Master Beneficiary Summary File and Chronic Conditions. Residents were in assisted living on December 31, 2017. Individuals were enrolled in fee-for-service Medicare. The x-axis is the cumulative proportion of AL communities (n=5961); the y-axis is the cumulative percentage of the total number of residents with SMI (n=26,747). The line of equality, drawn at 45°, depicts perfect equality. The farther the line deviates from the line of equality, the more unequal the distribution. The shaded area represents the 95% confidence interval.

FIGURE 4.

The Percentage of Assisted living Residents with SMI Plotted Against the GINI Coefficient of Between-Community Inequality in SMI, by State

Notes: Data were from the 2017 Medicare Master Beneficiary Summary File and Chronic Conditions. Residents were in assisted living on December 31, 2017. Individuals were enrolled in fee-for-service Medicare.

The x-axis is the GINI coefficient, representing between-community inequality in SMI. The y-axis is the percentage of residents with SMI in a state. A coefficient of 0 depicts perfect equality. A coefficient of 1 indicates perfect inequality. The line represents a locally estimated scatterplot smoothing (LOESS) curve that best fit the data. The shaded area represents the 95% confidence interval.

Discussion

To our knowledge, our study is the first to examine trends and geographic variability in the prevalence of SMI in AL. We found that the prevalence increased over time at a rate faster than in the community albeit slower than in NHs. Over one in ten AL residents had SMI in 2017, a 54% increase in the same settings since 2007. Due to the overall increase in the number of residents in AL, the total number of residents with SMI in these settings grew 165%. This increased prevalence of SMI in AL has implications for the provision of mental health services in this setting, a topic that is largely underexplored other than in broad strokes (e.g., in 2016, 55% of ALs provided ‘mental health or counseling services’) (1).

Consistent with previous work (28,29), individuals with SMI in our sample had a higher prevalence of medical conditions such as obesity, COPD and diabetes despite being younger on average; they were also more likely to have 6 or more chronic conditions. The prevalence of health concerns may be a concern because, unlike NHs, AL communities are not required to provide round the clock nursing services, although 54% have an RN or LPN on staff (3). More research is needed into the mental health and medical care needs for this population.

The increase in the proportion of AL residents with SMI is likely related to the increase in Medicaid coverage options available. However, it should be noted that the highest increase in the proportion of residents with SMI occurred in nursing homes. Given that Medicaid coverage of AL varies by state, funding for AL services may remain suboptimal in many locations. Future research can empirically investigate the relationship between state Medicaid policy and changes in the prevalence of SMI in AL.

Interstate variability

Our study found substantial variability in the prevalence and concentration of SMI among states. Multiple factors could contribute to these differences. For example, states differ in policies regarding screening and admission of residents with behaviors that can result from SMI. Wyoming and Colorado do not allow residents who display “disruptive behaviors” to be admitted or retained in AL (31). Both states have a lower prevalence of SMI in AL than the national rate. Oregon, similarly, has a low rate of SMI in AL and does not admit or retain residents who display “behavior or actions that repeatedly and substantially interfere with the rights, health, or safety of residents of others”(31).

We found that residents with SMI were more likely to be Medicaid recipients, suggesting that access to Medicaid likely plays a role in access to AL care for individuals with SMI. The impact of state and federal policy on SMI prevalence seems operative, but it is not consistent. For example, Medicaid does not pay for AL services in Kentucky, Louisiana, and West Virginia (8), and while Louisiana and West Virginia have lower prevalence of SMI in AL than the national rate, Kentucky’s rate is higher than the national rate. Medicaid supplementation also varies by state. Room and board is not funded by Medicaid, so these fees must be paid in another manner (32). Some states allow outside parties, usually families, to supplement these costs. Texas and Oregon do not allow supplementation, which may in part explain their low rates of SMI in ALs (32). The Supplemental Security Income (SSI) program allows some individuals with SMI to pay for room and board. Unlike some states, Arizona and North Dakota do not augment their SSI programs, which may explain their lower rates of SMI in AL (32). Taken together, our findings suggest that additional investigation is needed to better understand the mechanisms driving the interstate variability in the presence of SMI among AL residents.

Intrastate variability

Our study found 10% of AL communities housed over half of the residents with SMI in our sample. There are likely multiple factors contributing to this high concentration of residents with SMI in a small number of AL communities. For one, states license AL communities differently. Florida, for example, has a separate license category called “limited mental health;” under this license, ALs can provide specialized behavioral care to 3 or more residents with mental illness. Not surprisingly, settings with a limited mental health license in FL care for a higher proportion of residents with SMI (20). Similarly, New York has a specific designation for AL communities called “transitional adult homes.” These AL communities have a capacity of over 80 beds and a resident population of at least 25% individuals living with SMI (33).

Differences in whether particular settings accept Medicaid residents also likely affect intrastate variability in SMI. This is important given that research in NHs suggests that residents with SMI are disproportionately located in the lower quality facilities in part due to higher reliance on Medicaid reimbursement (17). Medicaid reimbursement rates in AL are also often very low (32). We cannot make statements regarding the quality of AL communities that serve residents with SMI, but there is cause to examine this topic due to research in NHs suggesting that high concentrations of residents with SMI are associated with lower quality of care( 15,16). Research is needed to determine adequate reimbursement requirements for settings that care for individuals with SMI.

Other market factors, such as the availability of other residential alternatives, may play a role in explaining intrastate variation in the prevalence of SMI among AL communities. One study found that the proportion and increase over time in the prevalence of SMI in NHs was higher in facilities closer to a psychiatric hospital (15). Another state-specific study did not find a significant relationship between psychiatric bed supply and the prevalence of SMI in NHs (34).

Our study is not without limitations. Our methodology is dependent on identifying residents in larger AL communities (25+ beds), limiting its generalizability to smaller settings. Smaller settings tend to have a higher proportion of residents with SMI and residents who are covered by Medicaid (17), therefore the prevalence of residents with SMI in all ALs is likely to be higher than what we report. In addition, our cohort consists only of residents who receive fee-for-service Medicare; therefore, it is not representative of residents who are enrolled in Medicare Advantage, representing approximately ⅓ of all Medicare beneficiaries (35). Given that Medicare Advantage enrollees tend to be wealthier and have better self-reported mental health status on average (36), our estimates of SMI in AL may be reduced if these individuals were to be included. Given the differences in MA enrollee characteristics as well as the variation in MA plans’ benefit packages, future research examining SMI in AL specific to the MA-enrolled population is warranted. Further, our study is descriptive. Thus, research is needed that examines potential reasons for the trends and variability observed.

We identified SMI using CCW measures, which have been found to undercount the true prevalence of SMI (34). However, we do not expect that this undercount differed over time. We compare rates to the community and to nursing homes in an effort to account for temporal changes (e.g., the change from ICD-9 to ICD-10 codes) that would influence individuals in all settings. However, our estimates could still be affected. Although we focus on intrastate variability, federal policies, such as the introduction of the Affordable Care Act in 2010, may also contribute to the increases observed.

Conclusions

Our study found that the prevalence of SMI in AL has increased over time at a rate faster than in the community. Our study also found significant within and between-state variability in the presence of SMI in AL. Work is needed to tease out the reasons for this variability. Finally, our study found a high concentration of residents within 10% of AL communities. Research is needed to ensure that training and resources are adequate to meet the needs of residents in these settings.

Highlights.

• What is the primary question addressed by this study?

  Has the prevalence of serious mental illness (SMI) in assisted living (AL) increased in recent years?

• What is the main finding of this study?

  Rates of SMI in AL increased between 2007 and 2017 at a rate faster than the community but slower than in nursing homes. We found substantial between and within state variability in the percentage of residents with a diagnosis of SMI.

• What is the meaning of the finding?

  Given the increased prevalence of SMI in AL, research is needed into the care needs of this population.