Serious Mental Illness in Assisted Living Communities: Association with Nursing Home Placement

Abstract

Objectives:

Assess prevalence of serious mental illness (SMI) alone, and co-occurring with Alzheimer’s Disease and Related Dementias (ADRD), among Medicare beneficiaries in assisted living (AL). Examine the association between permanent nursing home (NH) placement and SMI, among residents with and without ADRD.

Design:

2018-2019 retrospective cohort of Medicare beneficiaries in AL. We used data from the Medicare Enrollment Database, the Medicare Beneficiary Summary File (MBSF), Minimum Dataset (MDS), and a national directory of state licensed AL communities. AL residents were identified using a validated, previously reported 9-digit zip code methodology.

Setting and Participants:

A cross-sectional study sample included 289,350 Medicare beneficiaries in 17,265 AL communities across 50 states and in the District of Columbia.

Methods:

The outcome was permanent NH placement: a continuous stay for more than 90 days. Key independent variable was presence of SMI - schizophrenia, bipolar disorder, and major depression. Other covariates included socio-demographic factors and presence of other chronic conditions, including ADRD. A linear probability model with robust standard errors, and AL-level random effects, was used to test the association between SMI diagnoses, ADRD, and their interactions, on NH placement.

Results:

More than half (55.65%) of AL residents had a diagnosis of SMI, among them 93.2% had major depression, 28.5% schizophrenia, and 22.2% bipolar disorder. Individuals with schizophrenia and bipolar disorder, had significantly lower probability of NH placement, a 32% and a 15% decrease relative to the cohort mean, respectively. Placement risk was significantly greater for residents with ADRD compared to those without, increasing for those who also had schizophrenia or bipolar disorder, 12.9% and 1.5% relative to the sample mean, respectively.

Conclusion and Implications:

Presence of schizophrenia and bipolar disorder, in conjunction with ADRD, significantly increase the risk of long-term NH placement, suggesting that ALs may not be well-prepared to care for these residents.

Brief summary:

This study examined presence of serious mental illness among Medicare beneficiaries who reside in assisted living communities, and the relationship between mental illness, dementia and permanent nursing home placement.

Introduction

Nursing homes (NH) have long been de facto institutional care settings for older individuals with serious mental illness who are unable to live independently in the community.^1,2 As assisted living (AL) communities have become increasingly popular alternatives to nursing homes, available evidence suggests that a similar pattern may be emerging in this residential setting.³

A recent study reported that the prevalence of serious mental illness in ALs has increased between 2007 and 2017, from 7.4% to 11.4%, with a significant variation across AL communities and states.⁴ A review of the literature showed the prevalence of psychiatric disorders in ALs to range from 26% to 33%, depending on the study cohort and the methods used.⁵ A report from the American Association of Homes and Services for the Aged estimated that 38% of AL residents had SMI.⁶ Another study of ALs in Maryland reported that 26.3% of residents had non-cognitive psychiatric diagnosis.⁷ Behavioral symptoms such as verbal or physical aggression, which are known to accompany depression, psychosis, and dementia, have been shown to be quite common in ALs. Based on a sample of ALs in 4 states, over 1/3 of the residents were reported to exhibit such symptoms at least once per week.⁸ Furthermore, AL is currently the largest residential care setting for individuals with Alzheimer’s disease and related dementias (ADRD), who account for almost half of all Medicare beneficiary residents,⁹ many of whom also suffer from depression and other co-morbid mental illness.¹⁰ The influence of SMI and its associated behavioral symptoms on AL residents’ ability to age in place is not clear, but there is significant cause for concern because ALs are frequently not incentivized to right-size the services they offer to meet the needs of their clients, particularly those with higher levels of care needs.¹¹ One study of 193 AL communities found that residents who were depressed had a higher rate of discharge to nursing homes than residents without depression.¹² Another study of 348 residents from 37 communities in one state showed that those with non-ADRD psychiatric disorders were more likely to be discharged to NHs compared to residents without psychiatric diagnoses.¹³

AL residents with SMI, and those with SMI and ADRD, have greater requirements for medical and behavioral supervision.^4,8 These care needs are likely to challenge the ALs’ original social model of care.^14,15 Today, only half of all ALs report providing any mental health or counseling services (rendered by AL employees or arranged with an outside service provider).¹⁶ While AL dementia care units are expected to provide care more supportive of the needs of residents with mental and behavioral health issues, only a minority of the residents who have these needs actually reside in ALs with dementia care units,¹⁷ and often these units have been reported to fall short on mental care provision.¹⁸ A number of studies have shown that mental health needs in ALs are under addressed and that mental health services utilization is low.^8,19-21

Existing studies of mental illness in ALs have been based on a variety of estimates and interpolations, using different methodologies for identifying AL residents and different definitions of SMI. Most were based on small, non-random or convenience samples.^7,8,13 While these studies have focused on residents with some SMI conditions, they have not specifically examined all SMI diagnoses and their associations with NH placement, particularly among residents who concurrently were diagnosed with ADRD.

Motivated by these gaps in the literature, the objectives of our study were twofold. First, using recent national-level data, we assessed the prevalence of SMI diagnoses alone, and co-occurring with ADRD, among Medicare beneficiaries residing in ALs. Second, we examined the relationship between SMI diagnoses, with and without ADRD, and NH placement of AL residents.

Methods

Data Sources and Study Sample

We used CY 2018-2019 data from the Medicare Enrollment Database, the Medicare Beneficiary Summary File (MBSF), Minimum Dataset (MDS), and a national directory of state licensed AL communities.²² Medicare beneficiaries who were AL residents were identified using a validated, previously reported methodology in which 9-digit zip codes were employed to identify Medicare beneficiaries residing in AL during the study period.²³

Each resident was observed for up to two years. To allow for sufficient time to track placement in a nursing home, we limited the study sample to residents already in AL before October 2019. We excluded residents enrolled in the Medicare Advantage plans because their diagnostic data is not consistently accurate in the chronic conditions segment of the MBSF. Residents who died during the study period were included in the analysis if death occurred prior to long-stay nursing home placement. A sensitivity analysis was also conducted in which only residents alive throughout the study period were included, but as the results were not measurably different, these findings were not included in this study.

We created a cross-sectional study sample that included Medicare beneficiaries who were AL residents across 50 states and in the District of Columbia. The study sample consisted of 289,350 AL residents in 17,265 communities.

Outcomes

The outcome was permanent NH placement, which was defined as a continuous stay in a facility for more than 90 days, based on the MDS.

Key independent co-variates

We used the Chronic Conditions Warehouse (CCW) and the Other Potentially Disabling Conditions (OPDC) segments of the MBSF to identify residents who had SMI diagnosis and/or were diagnosed with ADRD. These segments were created using algorithms that search through Medicare claims to identify specific conditions.^24,25 AL residents with SMI were identified as having schizophrenia spectrum disorders, bipolar disorder, and major depression. We followed this definition adopted in prior research,²⁶ because these conditions were shown to confer an increased risk of early nursing home placement in other populations.^27,28

Other covariates

Other covariates included a resident’s age, gender, and race/ethnicity. Race was categorized as White, Black or Other (non-Hispanic), and ethnicity as Hispanic or not, following a validated classification from the Research Triangle Institute.²⁹ The category of “other” includes Asians, Alaskan Natives, and other groups that were too few in numbers to include as separate categories. We also included an indicator to identify beneficiaries who were dually eligible for both Medicare and Medicaid (dual) or for Medicare-only, based on their continuous enrollment during the study period. To control for the risks associated with death, we included a count of days alive during the study period. In addition to SMI and ADRD, we also included other chronic and disabling conditions that are particularly prevalent among individuals with SMI (e.g., depressive moods and anxiety, chronic kidney disease, congestive heart failure, and others).

Statistical Analysis

In a bivariate analysis, stratified by presence/absence of SMI, we examined the characteristics of AL residents. We also plotted the prevalence of SMI diagnosis alone and co-present with ADRD, and the corresponding rates of NH placement.

As our base model we employed a linear probability regression with robust standard errors to test the association between SMI diagnoses, ADRD, and their interactions with NH placement, controlling for other covariates. This regression model included AL-level random effects. As a sensitivity analysis, to account for AL-level covariates, we also run a model with AL-level fixed effects and another model with state AL staffing regulations. The associations between SMI, ADRD and NH placement were not significantly different in these sensitivity models and, therefore, only the based model is presented in this study.

All statistical analyses were performed using SAS 9.4 and Stata 16.1. The study was reviewed and approved by the Institutional Review Board.

RESULTS

Characteristics of AL residents with and without SMI

Characteristics of the study sample, stratified by presence/absence of SMI, were depicted in Table 1. The overall, unadjusted NH placement for all AL residents was 6.66%, 4.64% for those without SMI and 8.27% for those with SMI. More than half (55.65%) of AL residents had a diagnosis of SMI. They tended to be younger (81.6 vs. 85.6 years on average) and were more likely to be Black (5.1% vs. 3.36%) and Hispanic (2.84% vs. 1.87%) compared to residents without SMI. They were also more than twice as likely dually eligible for Medicare and Medicaid than Medicare-only (29.4% vs. 11.3%) and had higher mortality, as measured by days alive (476 for those with SMI vs. 502 for those without).

Table 1.

AL resident study cohort stratified by presence of serious mental illness (SMI) (N=289,350)

	No SMI N=128321 (44.35%)	Has SMI N=161029 (55.65%)	P Value
Resident Characteristics
Resident Age
Mean (SD)	85.6 (9.56)	81.6 (12.1)	<0.001
Race/Ethnicity
NH White	117254 (91.4%)	144699 (89.9%)	<0.001
NH Black	4316 (3.36%)	8218 (5.10%)
Hispanic	2394 (1.87%)	4575 (2.84%)
Other	4357 (3.40%)	3537 (2.20%)
Male	49831 (38.8%)	51452 (32.0%)	<0.001
Medicare-Medicaid Dual	14442 (11.3%)	47343 (29.4%)	<0.001
Days Alive CY2018-2019: Mean (SD)	502 (246)	476 (249)	<0.001
Length of stay in AL (months): Mean (SD)	41.36 (30.32)	40.16 (29.97)	<0.001
Mental Conditions
Schizophrenia	0 (0%)	45818 (28.5%)	<0.001
Bipolar	0 (0%)	35679 (22.2%)
Major Depression	0 (0%)	150118 (93.2%)	<0.001
Depressive Mood/Anxiety	42214 (32.9%)	156419 (97.1%)	<0.001
ADRD	47882 (37.3%)	101239 (62.9%)	<0.001
Chronic Conditions
COPD	41504 (32.3%)	79143 (49.1%)	<0.001
Chronic Kidney Disease	64765 (50.5%)	100091 (62.2%)	<0.001
CHF	56715 (44.2%)	88513 (55.0%)	<0.001
Diabetes	48077 (37.5%)	81650 (50.7%)	<0.001
Osteoporosis	49254 (38.4%)	71533 (44.4%)	<0.001
Stroke	33669 (26.2%)	58372 (36.2%)	<0.001
Mobility Impairment	11595 (9.04%)	24735 (15.4%)	<0.001
Hypertension	113480 (88.4%)	152226 (94.5%)	<0.001
Outcome
NH Placement	5950 (4.64%)	13311 (8.27%)	<0.001

Notes:

1. Abbreviations: SMI, Severe Mental Illness; ADRD, Alzheimer Disease and Related Dementia; COPD, Chronic Obstructive Pulmonary Disease; CHF, Chronic Heart Failure.

Among those with SMI, almost all had a diagnosis of major depression (93.2%), 28.5% had schizophrenia and 22.2% bipolar disorder. Residents with SMI had a very large prevalence of depressive mood/anxiety (97.1%) and of ADRD (62.9%). Compared to those without SMI, residents with SMI had significantly higher prevalence of other chronic disorders (Table 1).

Prevalence of ADRD and SMI conditions and NH placement

In Figure 1, we depicted the unadjusted percent prevalence of ADRD and of SMI, by specific diagnoses, and showed the combined presence of SMI with ADRD. The same figure also showed, for each condition and the combination of conditions, the unadjusted proportion of AL residents with NH placement. Major depression was very prevalent in the study cohort, present among 51.8% of residents. Schizophrenia and bipolar disorder, although less frequent, were still very significant in this population. But the prevalence of schizophrenia and bipolar disorders largely seemed to have been due to their co-presence with neurodegenerative disorders such as ADRD. For example, while overall 15.83% of the study cohort had a diagnosis of schizophrenia, 11.30% had both schizophrenia and ADRD. Similarly, 12.33% of the study cohort had a diagnosis of bipolar disorder, but most of them had both bipolar illness and ADRD (8.10%).

Figure 1:

ADRD and Serious Mental Illness in Assisted Living: Unadjusted Prevalence and NH Placement

Note: The subgroups with ADRD in panel B are subgroups of the overall ADRD sample presented in panel A.

In Figure 1 (right-hand vertical axis), we also presented the unadjusted NH placement rate for individuals diagnosed with a specific condition. For example, while 10.88% of AL residents diagnosed with schizophrenia were placed in a NH over the study period, those with both schizophrenia and ADRD experienced a 12.70% NH placement rate, with 10.15% and 12.29%, for those with bipolar disorder alone and combined with ADRD, respectively.

Risk of NH placement for AL residents with SMI and with ADRD

In Table 2, we presented the results of the regression analysis for the risk of NH placement among AL residents during the study period. Compared to White residents, those who were Black, Hispanic, and Other had a significantly (p<0.01) lower risk of NH placement by 0.96, 2.56, and 1.00 percentage points (pp), respectively. Residents who were older and female had a higher risk of placement, as did those with a longer length of AL residence. Significantly, those with dual Medicare and Medicaid eligibility had 17.7 pp (p<0.01) higher NH placement risk compared to the Medicare only residents.

Table 2.

Results from linear probability model examining the relationship between NH placement, mental illness and ADRD.

Variables	β estimate	SE	P	95% Conf. Interval
Demographics
Resident Age	0.0016	0.0001	0.0000	(0.0015, 0.0017)
Race/Ethnicity (Ref=NH White)
NH Black	−0.0096	0.0029	0.0010	(−0.0154, −0.0039)
Hispanic	−0.0256	0.0036	0.0000	−0.0325, −0.0186
Other	−0.0100	0.0027	0.0000	−0.0153, −0.0047
Female	0.0038	0.0010	0.0000	(0.0018, 0.0058)
Medicare-Medicaid Dual	0.1775	0.0026	0.0000	(0.1723, 0.1827)
Length of stay in AL (Months)	0.0007	0.0000	0.0000	(0.0006, 0.0007)
Days Alive CY2018-2019	−0.0002	0.0000	0.0000	(−0.0002, −0.0002)
Mental Conditions
ADRD	0.0188	0.0017	0.0000	(0.0155, 0.0221)
Schizophrenia	−0.0216	0.0027	0.0000	(−0.0269, −0.0164)
Bipolar disorder	−0.0102	0.0025	0.0000	(−0.0151, −0.0052)
Major depression	0.0047	0.0015	0.0010	(0.0018, 0.0076)
Depressive Moods/Anxiety	−0.0044	0.0013	0.0010	(−0.0071, −0.0018)
Interactions between ADRD and SMI
Schizophrenia x ADRD	0.0302	0.0033	0.0000	(0.0238, 0.0367)
Bipolar x ADRD	0.0112	0.0034	0.0010	(0.0046, 0.0179)
Depressive Moods/Anxiety x ADRD	0.0053	0.0025	0.0340	(0.0004, 0.0103)
Major Depression x ADRD	−0.0012	0.0024	0.6280	(−0.0059, 0.0036)
Indicators for Chronic Conditions
Chronic Kidney Diseases	0.0073	0.0010	0.0000	(0.0054, 0.0092)
COPD	−0.0029	0.0010	0.0030	(−0.0048, −0.0010)
CHF	0.0075	0.0010	0.0000	(0.0056, 0.0095)
Diabetes	0.0037	0.0010	0.0000	(0.0019, 0.0056)
Osteoporosis	0.0058	0.0010	0.0000	(0.0038, 0.0078)
Stroke	0.0043	0.0011	0.0000	(0.0021, 0.0065)
Mobility Impairment	0.0294	0.0017	0.0000	(0.0260, 0.0328)
Hypertension	0.0031	0.0013	0.0200	(0.0005, 0.0058)

Notes:

1. Abbreviations: NH, Non-Hispanic; AL, Assisted Livings; ADRD, Alzheimer Disease and Related Dementia; COPD, Chronic Obstructive Pulmonary Disease; CHF, Chronic Heart Failure.
2. All models controlled for AL-level random effects. Robust standard errors are reported.

Individuals with schizophrenia and those with bipolar disorders, but without ADRD, had significantly lower probability of permanent nursing home placement. The 2.16 pp lower risk of NH placement for residents with schizophrenia was equivalent to a 32% decrease relative to the cohort mean [(−2.16/6.66)*100], and a 1.02 pp lower risk for residents with bipolar disorder was a 15% decrease relative to the mean [(−1.02/6.66)*100]. For residents with major depression, the risk of placement was 0.47pp higher than for those without, or 7% greater than the sample mean. For residents with ADRD, but without SMI, the risk of NH placement increased 28% relative to the mean (or 1.88 pp; p<0.01). Among residents with ADRD, who also had a diagnosis of schizophrenia or bipolar disorder, the risk of NH placement was, respectively, 0.86 pp and 0.10 pp higher, or 12.9% and 1.5% increase relative to the sample mean. There was no statistically significant added risk of NH placement for residents with both ADRD and major depression.

DISCUSSION

To the best of our knowledge, this is a first national-level study to examine NH placement among AL residents with SMI. More than half of AL residents had a diagnosis of SMI, most due to major depression, with more than a quarter (28.5%) due to schizophrenia and more than one in five (22.2%) due to a bipolar disorder. We found that while the overall NH placement in the study cohort was 6.66%, the unadjusted rate for residents with SMI was considerably higher.

Consistent with prior work,⁴ compared to AL residents without SMI, those with SMI had significantly higher rates of other chronic conditions, most notably ADRD, chronic kidney disease, chronic heart failure, and many others. Also consistent with prior studies, AL residents with a diagnosis of schizophrenia or bipolar disorder had a higher unadjusted rate of NH placement. However, after adjusting for numerous demographic and health covariates, we found that the diagnoses of schizophrenia or bipolar disorder were associated with a lower risk of NH placement, while those with major depression had a significantly increased risk of discharge to NHs. Interactions between specific SMI diagnoses and ADRD showed that residents with schizophrenia or bipolar disorder and ADRD had an increased risk of NH placement compared to those with ADRD alone; however major depression co-occurring with ADRD did not increase the risk of placement.

These findings raise important questions regarding the provision of mental health services in ALs and their ability to adequately care for and support residents with SMI, and especially those with SMI in combination with ADRD. Although the prevalence of SMI and of ADRD in ALs is quite substantial, empirical evidence on the provision of mental health services to AL residents is rather sparse. Approximately half of all ALs do not provide any mental health services.¹⁶ Furthermore, AL staff may not be adequately prepared to care for residents with SMI. The proportion of ALs that require staff training relating to psychiatric disorders has been shown to be very small and insufficient.³⁰ Notably, some states, for example Florida, have specific mental health license requirements for staff training and case management supervision for ALs serving at least one resident with SMI. While the required training is very modest, at least 6 hours, it may represent a step in the right direction. At the same time, strict AL regulations may hamper care provision to residents with SMI. For example, in many states AL regulations require that only licensed professional staff administer medications.³¹ Lack of qualified licensed staff to monitor medication use would thus hinder AL’s ability to provide adequate care to residents with SMI. As individuals with SMI disproportionately reside in small communities (<25 beds), and these communities are less likely to have licensed staff, their ability to provide appropriate care is further compromised. Overall, the provision of mental health services to AL residents has been scant.^8,19,20 A recent study has shown that only 1 in 5 AL residents with ADRD and a co-morbid psychiatric illness had any mental health visits with a Medicare provider prior to the COVID-19 pandemic.²¹ Recent recommendations for medical and mental health care in ALs emphasize the need for personalizing care for these residents and for keeping AL providers accountable for their expertise claims in serving this aging population. Recently, the panel of experts made 43 recommendations for medical and mental care in ALs.³² Key among these recommendations were staff training for dementia and mental illness, on-site availability of nursing staff, having a mental health provider on staff, and conducting formal cognitive assessment of residents.

The difficulties that ALs face in managing the care and providing support to residents with serious and ongoing mental illness is evident in the increased risk of NH placement these residents face. Close to 70% of residents with ADRD had a co-morbid SMI condition, which for those with co-occurring schizophrenia increased NH placement risk by as much as 13%, doubling the risk relative to the sample mean. Despite high prevalence of SMI in ALs, it has been shown that awareness of SMI is limited both among staff and administrators, and that the management of problematic behaviors among these residents is haphazard at best, neither providing good care for residents with SMI nor sufficiently protecting the residents without SMI.⁶ While NH placement of residents with SMI may be a “solution” for ALs unable to provide the promised care and supports, it is likely not the solution that the consumers entering ALs had anticipated or hoped for, nor is it keeping ALs accountable for providing the promised personalized supportive care and facilitating aging in place.

Several study limitations should be mentioned. First, we were not able to include individuals enrolled in the Medicare Advantage who account for a significant proportion of AL residents (~35%). While our findings cannot be generalized to all AL residents, a recent comparison of traditional Medicare and the Medicare Advantage enrollees showed they have similar demographics and levels of chronic illness.³³ Second, we used segments of the MBSF to identify SMI and other chronic conditions, which may undercount most recent diagnoses. Third, while we observed considerable variation in SMI prevalence across AL communities, we did not have data on AL-level characteristics, such as staffing, presence of medical personnel, and others, that may influence NH placement. We tried to partially control for these variations by including AL-level random effects in the regression models. AL fixed effects models were also estimated, but the results were not different and were not presented.

Conclusions and Implications

SMI diagnoses are common in AL and are often co-occur with ADRD. We found that the presence of schizophrenia and bipolar disorders, in conjunction with ADRD, significantly increase the risk of long-term nursing home placement, suggesting that ALs may not be well-prepared to care for these residents. Policy reforms may be needed to ensure that AL residents with these conditions can adequately access the personal, memory, and mental health care needed to facilitate aging in place, or to ensure that ALs do not accept residents who may need to transfer to a more intensive care setting in a relatively short amount of time after AL admission.