Quality of Nursing Homes Admitting Working-Age Adults With Serious Mental Illness

Abstract

Objective:

This study examines the association of nursing home quality and working-age nursing home residents with serious mental illness.

Methods:

This cross-sectional study used 2015 national Minimum Dataset 3.0 and Nursing Home Compare data. A logistic mixed-effects model estimated the likelihood (adjusted odds ratios [AORs] and 95% confidence intervals [CIs]) of a working-age nursing home resident having serious mental illness, by NHC health inspection quality rating. The variance partition coefficient (VPC) was calculated to quantify the variation in serious mental illness attributable to nursing home characteristics. Measures include serious mental illness (i.e., schizophrenia, bipolar, other psychotic disorders), health inspection quality rating (1 star (below average) to 5 stars (above average)), and other sociodemographic and clinical covariates.

Results:

Of the 343,783 working-age adults (22–64 years) newly admitted to a nursing home in 2015 (n=14,307 facilities), 15.5% had active serious mental illness. The odds of a working-age resident having serious mental illness was lowest among above average quality nursing homes, as compared to below-average quality nursing homes (aOR 5-star versus 1-star: 0.78; 95% CI: 0.73 – 0.84). The calculated VPC from the full model was 0.1105.

Conclusions:

Evidence indicates an association between below-average nursing homes and working-age nursing home residents with serious mental illness. This suggests that those with serious mental illness may experience inequitable access to nursing homes of above-average quality. Access to alternatives to care, integration of mental health services in the community, and improved mental healthcare within nursing homes may aid in addressing these disparities.

Introduction

Each year in the United States, one in 25 adults experience serious mental illnesses, such as schizophrenia, bipolar disorder, and other psychotic disorders;¹ losing an estimated $193.2 billion earnings.² About 7.7% of those aged 18–25 years and 5.9% of those 26–49 years have a serious mental illness.¹ Since the advent of the deinstitutionalization movement, a dramatic depopulation and closing of state psychiatric hospitals has occurred, with the hope of fostering community integration.³ A significant proportion of those with serious mental illness continue to live in facilities that are considered institutions, such as nursing homes.³

In the U.S., nursing homes are primarily designed to care for older adults with cognitive and physical limitations, rather than the specialized care needed for severe psychiatric illness. Despite this, many people with serious mental illness reside in nursing homes. While more recent data are unavailable,⁴ as of 2004, approximately 13% of nursing home residents had a diagnosis of schizophrenia and 4% had a diagnosis of bipolar disorder.⁵ About 2% of people of all ages with serious mental illness resided in nursing homes between 2008–2009.³ Working-age adults with serious mental illness may represent a growing population of nursing home residents with unique care needs. In 2015, 16.5% of U.S. nursing home residents were under the age of 65 years.⁶ Although this age group may account for nearly half of all people with serious mental illness admitted to nursing homes,⁷ research on the psychiatric care needs of these residents is sparse.

People with serious mental illness are more likely to be admitted to lower quality nursing homes.^8–11 To our knowledge, of studies examining nursing home quality for residents with serious mental illness, only one examined residents younger than 65 years.⁸ To address this knowledge gap, this study sought to explore the association between nursing home quality and the presence of serious mental illness (i.e., schizophrenia, bipolar, other psychotic disorders) in working-age adults (22–64 years) admitted to a nursing home.

Methods

Study design

This was a cross-sectional study approved by the Institutional Review Board of the University of Massachusetts Medical School.

Data sources

The Minimum Dataset (MDS) 3.0 is a government-mandated assessment of residents in Centers for Medicare and Medicaid (CMS) certified nursing homes in the United States.¹² It is completed for each resident by nursing home staff and the resident at admission, quarterly, and when health changes occur. Assessments include demographic, clinical, cognitive, and functional characteristics, and receipt of pharmacological and nonpharmacological therapies. MDS 3.0 2015 data were merged to CMS’s five-star quality rating system, Nursing Home Compare (NHC).¹³ NHC rates nursing homes from one to five stars across four domains: health inspections, staffing, quality of care, and overall. Ratings are calculated using facility characteristics, such as number and severity of deficiencies, staffing hours, and percent of residents who improve in function. MDS 3.0 admission assessments were merged to the most recent quarterly NHC data at resident admission date using CMS certification number.

Eligibility criteria

Residents aged 22–64 years (working-age adults) newly admitted to a nursing home in 2015 were included (Online supplement). Admission was determined using the CMS definition¹² (see Online supplement for MDS item numbers). Thus, residents admitted after acute care were not included. A new admission was further defined by excluding those who 1) were admitted from another nursing home or swing bed or 2) had a prior nursing home stay within 90 days of admission assessment. When more than one admission assessment was conducted, we randomly selected one for analysis. We excluded: 1) those without a NHC match (n=1,700); 2) comatose residents (n=1,174); and 3) those with missing information on key variables (n=39,822). The final sample included 343,783 persons and 14,307 facilities.

Outcome measure

The outcome measure was a binary indicator of serious mental illness defined as schizophrenia, bipolar disorder, or other psychotic disorder (see Online supplement for MDS item numbers).⁷ A diagnosis is recorded in MDS if it is physician-documented and it directly impacts the resident’s health or medical treatments.¹²

Exposure measure

The key exposure was whether a nursing home was of below-average quality, as measured by the health inspection domain of the NHC star ratings. One star indicates much below average quality, three stars indicate average quality, and five stars indicate much above average quality.¹³ The health inspection rating is based on federal regulations, is a standardized survey process, uses the three most recent years of inspections, and is weighted by the scope, severity, and timing of deficiencies.¹³ Major metrics for evaluating quality of care are incorporated, including staffing, nursing home environment, medication management, and resident quality of life (13–15), and the NHC rating moderately correlates with consumer ratings (16). The distribution of the health inspection rating is set so that for every state, 10% of nursing homes receive five stars, 20% receive one star, and the remaining facilities are distributed evenly across the other categories (13). Although the number of star ratings is based on the number of nursing homes in a state, the number of residents across these nursing homes is unknown and likely influenced by many factors. It is important to understand how many residents receive care in below-average facilities.

Covariate measures

Sociodemographic (age, sex, race-ethnicity, marital status, and admission source), functional (activities of daily living, cognitive impairment), and clinical covariates (psychiatric and other) were obtained from the MDS admission assessment.

Age was categorized into four groups (22–34, 35–44, 45–54, and 55–64 years) and sex included male or female. Race/ethnicity categories were limited to non-Hispanic White, non-Hispanic Black, Hispanic, or a composite category including American Indian, Alaska Native, Asian, Native Hawaiian, Pacific Islander, or multiracial. Marital status included never married, married, widowed, separated/divorced. Admission source included community (private home or apartment, board-and-care home, assisted living, or group home), acute hospital, psychiatric hospital, and other (inpatient rehabilitation facility, intellectual or developmental disability facility, or hospice).

Physical function included three levels (independent/limited assistance required, extensive assistance required, and dependent/total dependence) as defined using the activities of daily living self-performance hierarchy scale.¹⁷ The Cognitive Function Scale indicated intact cognitive function, mild, moderate, or severe impairment.¹⁸ A set of binary variables for the presence of other active disorders was included (dementia/Alzheimer’s, diabetes, cardiovascular, musculoskeletal, neurological other than dementia/Alzheimer’s, pulmonary, cancer, gastrointestinal, genitourinary, infection, malnutrition, depression, anxiety disorder, and post-traumatic stress disorder).

Nursing home structural characteristics (urbanicity, ownership, and number of certified beds) were defined using NHC data. Urbanicity was defined via the Census Bureau’s urban-rural classification system.¹⁹ Ownership type included for-profit, government, or non-profit as designated in NHC. The number of certified beds was categorized into quartiles (<63 beds, 63–99 beds, 100–127 beds, ≥128 beds). None of the covariates examined contribute to the health inspection star rating.

Statistical Analysis

We described the sample and facility-level characteristics using data collected on individual residents, stratified by serious mental illness. An absolute difference of ≥5 percent was considered notable rather than using statistical significance due to the large sample size. We quantified the association between quality of nursing home and presence of serious mental illness using a logistic mixed effects model. The model included two levels: the individual-level and the random effects of facilities. Dementia and neurological conditions were excluded due to multicollinearity. Covariates were included based on clinical significance and previous studies.^8,9 State cluster effect was accounted for using state dummy variables to control for variations in health inspection surveying and policies related to mental health. Variations in health inspection surveying may occur due to differences in state licensing laws, variations in the survey process by state surveyors, and Medicaid policies that impact nursing home eligibility rules, payment, and enforcement of quality.¹³

Conceptually, the outcome could be the quality rating of the nursing home chosen by the resident. Fitting a model on a group-level outcome (i.e., the quality rating) with aggregated individual-level exposure and confounders but without accounting for individual-level variabilities may introduce bias (20). To overcome this micro-macro multilevel situation, we defined the outcome as a binary indicator of serious mental illness (i.e., individual level) and the primary exposure as nursing home quality (i.e., group level) and fitted a logistic mixed-effects model to explore the association between nursing home quality and presence of serious mental illness while adjusting for individual-level covariates and similarity in the individuals’ nursing homes. The adjusted exposure odds ratio (OR) for the presence of serious mental illness is the same as the adjusted outcome OR for nursing home quality (21).

We first ran an empty model with the outcome and facility-level random effects, then sequentially added the individual-level fixed effects, the facility-level fixed effects, and state dummy variables. The final model included all covariates listed in Tables 1 and 2 (other than dementia and neurological conditions), state dummy variables, and random effects of facilities. We present adjusted odds ratios (aOR) and 95% confidence intervals (CIs) of the full model, focusing on the primary exposure in Table 1, and the full model results in Supplemental Table S1. We also report the variance partition correlation coefficient (VPC) of the final model to examine the variation attributed to unobserved differences between nursing homes. SAS 9.4 software was used.

Table 1.

Characteristics of working-age residents newly admitted to US nursing homes in 2015, by presence of serious mental illness

Characteristic	Residents with serious mental illness (n = 53,104)		Residents without serious mental illness (n = 290,679)
	N	%	N	%
Sociodemographic covariates
Age in years
22–34	2694	5.1	8463	2.9
35–44	4539	8.6	17606	6.1
45–54	14338	27.0	68181	23.5
55–64	31533	59.4	196429	67.6
Female	27270	51.4	143629	49.4
Race/ethnicity
Non-Hispanic White	37755	71.1	195995	67.4
Non-Hispanic Black	10810	20.4	65139	22.4
Hispanic	3460	6.5	21361	7.4
American Indian, Alaska Native, Asian, Native	1079	2.0	8184	2.8
Hawaiian, Pacific Islander, or Multi-racial
Marital status
Never married	28141	53.0	108247	37.2
Married	8243	15.5	88551	30.5
Widowed	3202	6.0	21752	7.5
Divorced/separated	13518	25.5	72129	24.8
Admitted from
Community	3245	6.2	11841	4.1
Acute hospital	43946	84.1	270109	94.0
Psychiatric hospital	4552	8.7	1183	<1
Other	538	1.0	4376	1.5
Functional covariates
Physical function
Independent/ limited assistance required for ADLs	20129	37.9	73539	25.3
Extensive assistance required for ADLs	25528	48.1	157740	54.3
Dependent/ total dependence for ADLs	7400	14.0	59151	20.4
Cognitive function
Intact	35213	66.3	222709	76.6
Mild impairments	10532	19.8	40386	13.9
Moderate impairments	5682	10.7	18713	6.4
Severe impairments	1677	3.2	8872	3.1
Clinical covariates
Comorbidities
Hypertension	31548	59.4	184866	63.6
Cardiovascular	22284	42.0	147761	50.8
Diabetes	18866	35.5	116130	40.0
Gastrointestinal	19661	37.0	99651	34.3
Pulmonary	17694	33.3	77705	26.7
Neurological	16230	30.6	81913	28.2
Dementia/ Alzheimer’s	4685	8.8	11224	3.9
Musculoskeletal	14103	26.6	88741	30.5
Genitourinary	8649	16.3	61249	21.1
Infection	11219	21.1	64912	22.3
Cancer	2989	5.6	26981	9.3
Malnutrition	1730	3.3	12528	4.3
Psychiatric comorbidities
Depression	23437	44.2	104162	35.8
Anxiety disorder	19561	36.8	67916	23.4
Post-traumatic stress disorder	1621	3.1	2516	<1

Missing data for admission from (serious mental illness n=823, without n=3170), physical function (serious mental illness n=47, without n=249), hypertension (serious mental illness n=10, without n=40), cardiovascular (serious mental illness n=15, without n=55), diabetes (serious mental illness n=8, without n=21), gastrointestinal (serious mental illness n=7, without n=36), pulmonary (serious mental illness n=9, without n=23), neurological (serious mental illness n=10, without n=23), dementia/ Alzheimer’s (serious mental illness n=1, without n=7), musculoskeletal (serious mental illness n=5, without n=34), genitourinary (serious mental illness n=3, without n=22), infection (serious mental illness n=11, without n=79), cancer (serious mental illness n=5, without n=50), malnutrition (serious mental illness n=2, without n=6), depression (serious mental illness n=14, without n=38), anxiety disorder (serious mental illness n=3, without n=30), and post-traumatic stress disorder (serious mental illness n=3, without n=2).

Table 2.

Nursing home residents with or without serious mental illness, by characteristics of nursing homes to which they were newly admitted in 2015

Characteristic	Residents with SMI (n = 53,104)		Residents without SMI (n = 290,679)
	N	%	N	%
Ownership
For-profit	45932	86.5	233814	80.4
Government-owned	1562	2.9	9693	3.3
Nonprofit	5610	10.6	47172	16.2
Urbanicity
Urban	46053	86.7	252225	86.8
Rural	7051	13.3	38454	13.2
Number of Medicare-/Medicaid- certified beds
Quartile 1 (low): < 63 beds	3982	7.5	26761	9.2
Quartile 2: 63 to < 100 beds	10153	19.1	51528	17.7
Quartile 3: 100 to < 128 beds	14246	26.8	83217	28.6
Quartile 4 (high): ≥ 128 beds	24723	46.6	12973	44.4

Results

Of 343,783 working-age adults admitted to a nursing home in 2015, 66.3% (N=5,227,962) were between the ages of 55 and 64, 50.3% (N=5,172,884) were male, and 68.0% (N=5,233,750) were non-Hispanic White. Of this total sample, 15.5% (N=553,104) had a documented active serious mental illness diagnosis. Of the 14,307 nursing homes identified, 76.8% (N=510,987) were in an urban setting, 81.4% (N=511,641) were for profit, and 44.8% (N=56,409) had $128 beds. About one-quarter of nursing homes (N=53,874, 27.1%) had no working-age residents with serious mental illness; a higher percentage of these nursing homes were rural, smaller, and had nonprofit ownership and slightly higher health inspection ratings.

Characteristics of working-age residents with serious mental illness

Table 1 shows that 15.5% of residents with serious mental illness were married versus 30.5% of those without. While few residents without serious mental illness were admitted from a psychiatric hospital, 8.7% of residents with serious mental illness were. Residents with serious mental illness were less likely to have physical functional dependencies (limited assistance required: 37.9% versus 25.3%) and were less frequently cognitively intact (66.3% versus 76.6%). Few working-age residents had severe cognitive impairment (~3%) or dementia/Alzheimer’s (8.8% among those with serious mental illness, 3.9% among those without).

Clinical comorbid conditions of working-age adults tended to be similar among those with and without serious mental illness. More than half had hypertension, and significant proportions of the two samples (42.0%–50.8%) had cardiovascular disorders. One-third of those with serious mental illness had pulmonary disorders, compared to about one-quarter of those without. Depression and anxiety disorder were common, but more so in those with serious mental illness (depression: 44.2% versus 35.8%; anxiety disorder: 36.8% versus 23.4%).

Serious mental illness, facility characteristics, and quality of nursing home

The proportion of residents with serious mental illness in for-profit facilities was 6.1% percentage points higher than the proportion of residents without serious mental illness in these facilities (Table 2). No meaningful differences by the presence of serious mental illness were observed in urbanicity and number of beds. Figure 1 shows the proportions of persons in both groups residing in the five quality-ranked facility categories. Of residents with serious mental illness, 55% resided in nursing homes that received below-average quality ratings (1 or 2 stars), compared to 49.7% of those without serious mental illness. Only 23.6% of residents with serious mental illness resided in nursing homes that received above average-quality ratings (4- or 5- stars), compared to 28.2% of residents without serious mental illness.

Figure 1.

Distribution of working-age residents with or without serious mental illness newly admitted to U.S. nursing homes in 2015, by health inspection star rating of nursing homes

Compared with a resident in a one-star nursing home, a resident in a five-star facility was significantly less likely to have serious mental illness (AOR=0.78), after the analysis was adjusted for individual- and facility-level covariates, state variations, and clustering in the nursing home (Table 3). A similar trend was seen for nursing homes with four stars (AOR=0.85), three stars (AOR=0.90), and two stars (AOR=0.95) (see online supplement for full model results). In total, 11% of the variation in the outcome was attributable to unobserved differences between nursing homes, after the analysis accounted for individual-level and facility-level covariates and state dummy variables (VPC=0.11).

Table 3.

Likelihood in 2015 that a working-age nursing home resident had serious mental illness, compared with a resident of a nursing home with a one-star quality rating

	Adjusted odds ratio	95% confidence intervals
Nursing Home Quality
1-star survey rating	reference
2-star survey rating	0.95	0.91–0.99
3-star survey rating	0.90	0.86–0.94
4-star survey rating	0.85	0.81–0.89
5-star survey rating	0.78	0.73–0.84

Note: VPC = 0.1105. Values based on SAS PROC GLMIMMIX; Estimation method: Laplace. Models adjusted for state and all covariates listed in Tables 1 and 2 (other than dementia and neurological conditions). See online supplement for full model results.

Discussion

Our findings show that working-age residents with serious mental illness may experience a disparity in nursing home quality, compared with those without serious mental illness. Slightly more than half (55.0%) of working-age residents with serious mental illness were admitted to one- or two-star nursing homes. Furthermore, among newly admitted working-age residents, we found an association between having serious mental illness and below-average nursing home quality, after the analysis was adjusted for individual- and facility-level factors.

Persons with serious mental illness are a disenfranchised population, with experts calling for serious mental illness to be designated as a health disparity.²² Individuals with serious mental illness experience greater than double the risk for early death, compared with the general population.^23,24 Marginalized groups such as racial/ethnic minorities²⁵, individuals with low education status²⁶, and those dually eligible for Medicare/Medicaid²⁷ tend to reside in lower quality nursing homes. In 2015, 26% of persons with serious mental illness were covered by Medicaid,²⁸ the predominant payer for nursing home care. The use of federal Medicaid funds for persons ages 22–64 to receive care in specialized psychiatric facilities is prohibited, potentially incentivizing nursing home admission.^29,30 Our analysis shows that working-age residents with serious mental illness are at risk for below-average nursing home care. Nursing homes with a high concentration of residents with serious mental illness have worse outcomes (e.g., use of physical restraints, any hospitalization, use of an indwelling catheter) facility-wide.¹¹ Nursing homes often lack the resources and trained staff to treat complicated mental health disorders³¹, potentially resulting in inadequate care. Improvement in psychiatric treatment within nursing homes could help address this disparity in nursing home quality. Policies which offer financial incentives that reward higher quality care for persons with serious mental illness, similar to current COVID-19 related incentive payments, may help improve quality of care for working-age adults with serious mental illness.¹¹

Nursing home admission may be inappropriate for many working-age residents with serious mental illness.³² Medicaid-certified nursing homes are required to evaluate potential residents for serious mental illness to ensure that they receive necessary services in the most appropriate setting.³³ Many residents in this study could complete activities of daily living independently or with limited assistance and few were cognitively impaired, indicating they have the functional capacity to live in a less restrictive environment and potentially obtain supported employment. Those with serious mental illness living in high support housing (such as nursing homes) report decreased quality of life as compared to those in supported housing, which have an increased focus on rehabilitation, personal choice, and improvements in social functioning.^34,35 Additionally, employment can improve quality of life in persons with serious mental illness³⁶; in the U.S. about 55% of working-age adults with serious mental illness are employed.³⁷ In addition to supported employment, recommended standards of institutional care for those with serious mental illness include cognitive behavioral therapy and family interventions.³⁸ Given the documented substandard treatment of mental illness in nursing homes³¹ and the focus on older adults with physical and cognitive impairments, it is unlikely that these standards are achieved. Enforcement of required evaluation and increased access to community-based alternatives to nursing homes could provide more appropriate support.

The relationship between health inspection ratings and admission of residents with serious mental illness may be impacted by factors such as ownership and size of the facility. For-profit and larger facilities admit a higher proportion of residents with serious mental illness, and both of these characteristics have been associated with worse quality and poorer resident outcomes.^39–41 For-profit facilities focus on profit maximization, while non-for-profits aim to fulfill a stated mission.⁴² Nonprofit facilities typically have certain tax exemptions and can reinvest excess revenues in improving quality of care.⁴² Because of their focus on earnings, for-profit facilities may prefer to admit residents that seemingly require less monitoring and resources, such as working-age adults with SMI.⁴³ Certain for-profit nursing homes may have also gained a reputation for treating the psychiatric population following deinstitutionalization.⁴³

To our knowledge, this is the first study to focus on residents less than the age of 65 years with serious mental illness and nursing home quality. We were limited to the data within the MDS 3.0 and NHC, which lack information regarding important characteristics such as payer-mix of nursing homes, Medicaid/Medicare eligibility, health coverage type, and county-level information. While we adjusted for state-level fixed effects in our model, more detailed geographical variations, such as market factors and urban-rural differences, may be particularly relevant to working age adults.^44–46 Longitudinal data and additional data sources such as Medicaid claims may be valuable in expanding our understanding of the quality of care received by these residents. While we utilized a mixed effects model to explore group-level and individual-level characteristics, future research should account for individual-group levels more accurately by utilizing micro-macro strategies of multilevel modeling such as a latent multilevel model.²⁰ This study is an important first step in better understanding quality of care for working-age adults with serious mental illness in nursing homes. In particular, the VPC of 0.11 highlights the largely unexplored role of facility-level factors and quality of nursing home care in those with serious mental illness.

Conclusions

More than a half a century after the beginning of deinstitutionalization, many individuals with serious mental illness continue to be housed in restrictive, low-quality institutions. Policies to protect those with serious mental illness consider institutionalization the last resort, to be used only if integrated community-based services are not a viable alternative.^47,48 The current mental health care system in the United States is likely failing those with serious mental illness, which may be the result of poor compliance with and oversight of these policies and of underinvestment in alternative placements that focus on psychosocial supports, such as supportive housing and supported employment. Our results show that a substantial number of working-age adults with serious mental illness receive care in nursing homes and that nursing home admission may not be indicated for many. Further, there is an association between the presence of serious mental illness and admission to below-average quality nursing homes. A commitment to providing high quality care to persons with serious mental illness is needed to confront the decades of mistreatment and poor outcomes experienced by this population.

HIGHLIGHTS.

• Many working-age adults with serious mental illness may be inappropriately admitted to nursing homes.

• Of persons ages 22–64 years admitted to a nursing home in 2015, 15.5% had a serious mental illness.

• More than half of working-age residents with serious mental illness were admitted to below-average nursing homes.