Risk of Nursing Home Use Among Older Americans: The Impact of Psychiatric History and Trajectories of Cognitive Function

Abstract

Objectives

Mental illness and cognitive functioning may be independently associated with nursing home use. We investigated the strength of the association between baseline (1998) psychiatric history, 8-year cognitive function trajectories, and prospective incidence of nursing home use over a 10-year period while accounting for relevant covariates in U.S. adults aged 65 and older. We hypothesized that self-reported baseline history of psychiatric, emotional, or nervous problems would be associated with a greater risk of nursing home use and that cognition trajectories with the greatest decline would be associated with a subsequent higher risk of nursing home use.

Methods

We used 8 waves (1998–2016) of Health and Retirement Study data for adults aged 65 years and older. Latent class mixture modeling identified 4 distinct cognitive function trajectory classes (1998–2006): low-declining, medium-declining, medium-stable, and high-declining. Participants from the 1998 wave (N = 5,628) were classified into these 4 classes. Competing risks regression analysis modeled the subhazard ratio of nursing home use between 2006 and 2016 as a function of baseline psychiatric history and cognitive function trajectories.

Results

Psychiatric history was independently associated with greater risk of nursing home use (subhazard ratio [SHR] 1.26, 95% confidence interval [CI] 1.06–1.51, p < .01), net the effects of life course variables. Furthermore, “low-declining” (SHR 2.255, 95% CI 1.70–2.99, p < .001) and “medium-declining” (2.103, 95% CI 1.69–2.61, p < .001) trajectories predicted increased risk of nursing home use.

Discussion

Evidence of these associations can be used to educate policymakers and providers about the need for appropriate psychiatric training for staff in community-based and residential long-term care programs.

Aging, Assisted living, Cognition trajectories, Life course health, Mental illness

Cumulative disadvantage and cumulative inequality across the life course can result in poorer health and greater functional and cognitive limitations later in life (Dannefer, 2003; Douthit & Dannefer, 2006; Ferraro et al., 2009; Ferraro & Shippee, 2009). Serious and often chronic mental illnesses have been identified as conditions experienced across the life course that create disparities in cognitive function later in life (Brown, 2010; Brown & Wolf, 2018). Although interest in geriatric mental health continues to grow in the field of gerontology, the influence of psychiatric problems on the use of institutional long-term care is not well documented. Recent studies indicate increasing proportions of mentally ill nursing home residents, although staff are not sufficiently trained to care for this population (Muralidharan et al., 2019). New nursing home admissions diagnosed with a serious mental illness are younger than nonmentally ill admissions, are more functionally independent, and require less care, but are also more likely to transition to long-stay status (Andrews et al., 2009; Becker et al., 2002; Grabowski et al., 2009).

Previous research using the Health and Retirement Study (HRS) analyzed the relationship between self-reported psychiatric problems and cognitive function in U.S. older adults and found that when controlling for ascribed characteristics, early-life indicators, and later-life demographic, socioeconomic, and health characteristics, self-reported psychiatric history was significantly associated with lower levels of cognitive function and steeper rates of cognitive decline among older adults (Brown, 2010). Subsequent research utilizing HRS-linked Medicare files indicated that older adults diagnosed with a severe mental illness (schizophrenia, bipolar disorder, or major depressive disorder) had a greater probability of being diagnosed with dementia, when controlling for age, sex, race, and educational attainment (Brown & Wolf, 2017).

This study extends this previous research by examining the relationship between self-reported psychiatric history, cognitive function trajectories, and nursing home utilization among U.S. older adults participating in the HRS. This study addresses the following research question: Do a history of psychiatric problems or mental illness and cognitive functioning trajectories predict the incidence of nursing home utilization, net the effect of selected life course variables?

This study addresses several gaps in the literature. While previous research has assessed the association between dementia, mental health (primarily depression), and nursing home use, there have not been any studies with a prolonged follow-up of 18 years or more, while also accounting for differential declines in cognition, despite a plethora of evidence indicating that cognition declines accelerate with age (Hajek et al., 2015; Luppa et al., 2008). In addition, most studies on cognition and nursing home use focus on the cognitive status of patients living in nursing homes (Emerson et al., 2017; Sverdrup et al., 2018). This study further contributes to the literature by assessing how different trajectories of cognitive decline predict nursing home use. This is particularly valuable as it provides insight into the mental and cognitive health of older adults who utilize nursing home facilities at very old ages. Moreover, by addressing these knowledge gaps, we believe that we are able to provide a more comprehensive picture of these influences on nursing home use which, along with other similar studies, can inform social and organizational policies regarding resources needed in nursing home facilities.

Background

Health disparities in later life are a result of variability in life course experiences between individuals and groups, as effects of cumulative disadvantage and cumulative inequality (Dannefer, 2003; Ferraro et al., 2009). Structural disadvantage across the life course, often the result of interlocking power relations and social factors that create disparities in health through hierarchical stress and social selection, can result in poorer health and greater functional and cognitive limitations later in life (Douthit & Dannefer, 2006; Link & Phelan, 2000; McMullin, 2000). Serious and often chronic mental illnesses have been identified as conditions that create disparities in cognitive function later in life (Brown, 2010; Brown & Wolf, 2018).

Older adults with serious mental illness may be at greater risk of institutionalization in skilled nursing facilities (Andrews et al., 2009; Becker et al., 2002). Depression has been identified as a risk factor for functional disability among mentally ill older adults (Conus et al., 2014; Deschênes et al., 2015), and depression, anxiety, and other neuropsychiatric symptoms have been associated with limitations in activities of daily living (ADLs) and instrumental activities of daily living in older adults with cognitive impairment with or without dementia (Okura et al., 2010). Functional disability, in turn, places older adults at greater risk of nursing home admission (Middleton et al., 2018; Wergeland et al., 2015).

The numbers of nursing home residents with serious mental illness have been increasing in the general nursing home population, as well as in the Veterans Affairs nursing home system (Bowersox et al., 2013; Fullerton et al., 2009; Van Rensbergen & Nawrot, 2010). Five percent of all nursing home residents in 2004 were 65 or older and had a mental illness diagnosis (Bagchi et al., 2009). The proportion of new nursing home admissions with mental illness equals or exceeds new admissions with dementia (Fullerton et al., 2009), and the number of residents in Veterans Affairs nursing homes diagnosed with a serious mental illness almost doubled between 1999 and 2007 (Bowersox et al., 2013).

A history of psychiatric problems has been associated with nursing home admission at younger ages, even though these admissions generally involve fewer comorbidities and less functional dependence (Andrews et al., 2009; Aschbrenner et al., 2011). In 2005, approximately 27.4% of 1.2 million new admissions across the United States had at least one mental illness diagnosis (anxiety, depression, bipolar disorder, or schizophrenia), and the average age at first admission for people with serious mental illness (bipolar disorder or schizophrenia) was 62, compared to age 77 for all new admissions (Grabowski et al., 2009). New admissions with mental illness were also more likely to transition to long-stay status or to remain in the facility at least 90 days after admission (Grabowski et al., 2009). This combination of earlier admission and longer stay may result in prolonged use of nursing home facilities, which is associated with greater long-term care costs for older adults with mental illness.

This shift in nursing home populations continues, even though we have historically provided insufficient services to older adults with serious mental illness in institutional long-term care (Linkins et al., 2006). Research has shown that nursing home staff were ill-equipped to care for residents with psychiatric disorders and that this inability to provide adequate psychiatric care was a systemic problem in the nursing home industry (Muralidharan et al., 2019; Tariot et al., 1993). Furthermore, evidence links greater proportions of seriously mentally ill residents in nursing homes to poorer quality of care for all residents (McGarry et al., 2019; Rahman et al., 2013). Mentally ill residents in skilled nursing facilities often receive poorer quality of care (Grabowski et al., 2010). Care for mentally ill nursing home residents also varies across states, with only six states having regulations specifically addressing the care of residents with serious mental illness (Street et al., 2013).

The HRS is well suited to examine the relationship between psychiatric history, cognitive function, and nursing home utilization. The HRS includes a representative sample of U.S. adults aged 65 and older whose cognitive function was measured using a modified version of the Telephone Interview for Cognitive Status (TICS). At baseline, 10,766 adults 65 years and older reported on whether they had ever been told by a physician that they had psychiatric problems. Of these, 8,353 had a TICS score at baseline, plus one or more additional TICS measures up to 2006. The final analytical sample included a total of 5,628 respondents who were alive and participated in the 2006 wave, which we used as the entry point for the competing risks regression models. This analysis provides insight into how psychiatric history and differential cognitive function trajectories may prospectively affect the likelihood of living in nursing home facilities. This study utilizes data from the years 1998 through 2016 of the HRS to explore the hypotheses that:

1. Individuals with a self-reported history of psychiatric, emotional, or nervous problems will have a greater risk of nursing home use.

2. Cognitive function trajectories with greater decline will be associated with a subsequent higher risk of nursing home use.

Data and Methods

HRS Data

Utilizing 18 years of the HRS data, this project used a two-step analytic approach to (a) identify cognitive function trajectories and then (b) examine the hazard of nursing home use over the subsequent 10 years as a function of psychiatric history and cognitive function trajectories. Models of the relationship between self-reported psychiatric history and nursing home utilization control for selected early-life, demographic, health status, and health behavior variables, using data from the 1998, 2000, 2002, 2004, 2006, 2008, 2010, 2012, 2014, and 2016 data years of the HRS. RAND Center for the Study of Aging cleans, processes, and streamlines HRS data to allow for comparability across waves. We used their publicly available RAND HRS files as the source for our primary variables of interest.

Sample Construction

Respondents were included in the study if they (a) were aged 65 or older, (b) indicated whether a physician had ever told them that they had psychiatric problems at baseline (n = 10,766), (c) had two or more TICS scores between 1998 and 2006 including one at baseline (to allow for assessment of intercept and slope of trajectories; n = 8,353), and (d) were alive and participated in the 2006 wave, which we used as the entry point for the competing risks regression models (n = 5,628).

HRS respondents were administered the TICS items beginning with their first interview after reaching age 65. Participants who did not complete the cognition interview during a particular interview, or whose cognition data were provided by a proxy, were excluded from that wave of the study. To assess for risk of nursing home use between 2006 and 2016, we included only those respondents who were alive and participating in the study in the 2006 wave. We then used the competing risk analysis to address mortality bias but not selective attrition from 2006 through 2016.

Dependent Variable

A dichotomous variable for nursing home utilization was recorded at each interview based on the respondent report of residing in a nursing home at the time of interview (yes = 1). Nursing home incidence was defined based on the first report of nursing home utilization starting in 2006.

Independent Variables

Psychiatric history

Rather than inquiring about specific diagnoses or categories of mental disorders, the HRS asked participants if a physician had ever told them that they had psychiatric, emotional, or nervous problems, and if they are now getting psychiatric or psychological treatment for these problems (Brown, 2010). Participants who scored as “refused” or “don’t know” on either of these questions (less than 10% in each wave) were assigned the modal value of zero or “no.” These questions provide general mental health information, which is used to construct a baseline measure of psychiatric history in 1998. This baseline measure allows us to distinguish individuals reporting a history of these problems from those who may be experiencing later-life onset or mental health issues related to aging. At baseline, 610 participants reported a history of having ever been seen by a physician for a psychiatric, emotional, or nervous disorder.

Cognition scores

Participants were administered a modified version of the TICS items beginning with their first interview after reaching age 65. The TICS was designed based on Folstein’s Mini-Mental Status Examination, a commonly used instrument for assessing cognitive impairment in clinical settings, which could be reliably administered by telephone (Herzog & Wallace, 1997). For the HRS, the TICS was modified to measure six tasks with a maximum score of 35 points, evaluating memory and executive function and weighting fluid cognitive measures more heavily than in the original instrument (Freedman et al., 2001). The TICS was modeled after the state-of-the-art understanding of the dimensions of cognition in the late 1980s, and its validity has been previously documented (Zsembik & Peek, 2001). The TICS measures used for this study were cleaned and processed to allow for comparability across waves by the RAND Center for the Study of Aging (Bugliari et al., 2016).

Other variables

Controls include inherent characteristics (gender), early-life disadvantage and health, later-life socioeconomic status (SES; relationship status, educational attainment, and household income), multimorbidity (0, 1, 2 or more of diabetes, hypertension, cardiovascular disease, lung disease, cancer, arthritis, and stroke), health behaviors, ADL limitations, and depressive symptoms as measured by an abridged Center for Epidemiologic Studies—Depression (CES-D) scale. The construction of these variables is described previously (Brown, 2010).

Analysis Plan

This study began with univariate and bivariate descriptions of participants in the HRS data set, using measures of nursing home utilization, sociodemographic characteristics, indicators of childhood disadvantage, health indicators and health behaviors, sensory impairments, ADL limitations, and depressive symptoms. Sample characteristics were summarized using frequencies and means. Bivariate analyses compared participants with a history of psychiatric, emotional, or nervous problems in 1998 to their counterparts who reported no such history, by examining differences in percentages (chi-square) or mean values of descriptive characteristics by psychiatric history.

This study employed a two-step analytic approach: to (a) identify cognitive function trajectories and then (b) assess the association between psychiatric history and the newly constructed trajectories with the incidence of nursing home use. Using the Stata plugin “traj” for group-based trajectory modeling, we constructed distinct trajectories of cognitive functioning (1998–2006) using latent class mixture models that estimated the number and size of the trajectories and assigned probability of latent membership to individuals in the sample. Models were first run with an intercept-only model, after which linear, quadratic, and cubic growth factors were added to determine the forms of the growth model (Duncan & Duncan, 2009). The best-fitting trajectories were modeled with their own linear functional form and direction simultaneously. We tested between two and seven models, and model selection was determined using (a) Bayesian information criterion, which allows for comparisons of both the number of trajectories and functional forms of each trajectory, (b) confidence intervals (CIs), (c) reasonable sample sizes in each identified trajectory, and (d) average posterior probabilities for which a value above 0.7 indicated a good fit (Nagin & Odgers, 2010). Models were initially run without covariates to establish the best-fitting model, after which age, gender, and race were included as covariates in the final trajectory model.

To account for potential bias from censoring the latent failure time to nursing home use in individuals who die before this event, we used the competing risk analysis to determine the cumulative incidence of nursing home use, with death considered a competing event. Using Fine–Gray subdistribution hazard models (Fine & Gray, 1999) we examined the association of (a) psychiatric history and (b) the newly constructed trajectories with nursing home use between 2006 and 2016, while adjusting for participant characteristics in 2006. We ran six hierarchical models: Model 1, the crude model, predicts risk of nursing home use as a function of psychiatric history and the cognitive function trajectory classes identified in the latent class analysis; Model 2 includes age, sex, and race (inherent characteristics); Model 3 includes childhood factors—disadvantage and childhood poor health (control for missing); Model 4 includes later-life SES—education, household income, and relationship status; Model 5 includes health behaviors; and Model 6, the fully specified model, includes health status—multimorbidity, ADL limitations, and CES-D score. The main effects in Models 1 through 5 are presented in the Supplementary Table. Table 2 presents the fully adjusted results of Model 6. To examine potential effect modification, interactions between psychiatric history and cognition trajectories were tested; however, these interactions were not significant, and no further data stratification was conducted. All analyses were performed using Stata14 MP Software (Stata, College Station, TX).

Results

There were various differences between participants who reported a history of psychiatric, emotional, or nervous problems and their counterparts who reported no such history. Table 1 displays these differences for baseline (1998) and time-varying characteristics (1998 and 2006). For example, reports of a history of psychiatric, emotional, or nervous problems varied based on sex (13.5% of women vs. 6.9% of men, p < .000), childhood health (20.6% poor to fair vs. 10.2% good to excellent, p < .000), and educational attainment (15.4% less than high school vs. 9.3% high school or equivalent and 8.5% more than high school, p < .000). Individuals with a history of psychiatric, emotional, or nervous problems were more likely to have CES-D scores indicative of depressive symptoms (p < .000).

Table 1.

Sample Characteristics by Psychiatric History, HRS 1998 and 2006

		1998			2006
	Total sample (N = 5,628)	No history of psychiatric problems (n = 5,018	History of psychiatric problems (n = 610)	p > t	No history of psychiatric problems (n = 5,018	History of psychiatric problems (n = 610)	p > t
Baseline characteristics		%	%		%	%
Race/ethnicity
White	81.1	89.4	10.6	****
Hispanic	6.5	85.5	14.5	****
Black	11.0	89.7	10.3	****
Other	1.4	86.4	13.6	****
Total sample	100.0	89.2	10.8
Early-life characteristics
Male	39.7	93.1	6.9	***
Female	60.3	86.5	13.5	***
Total sample	100.0	89.2	10.8
Childhood disadvantage index (mean, SD)	0.35 (0.30)	0.348 (0.30)	0.398 (0.31)	***
Childhood health
Good to excellent	94.1	89.8	10.2	***
Poor to fair	5.9	79.4	20.6	***
Total sample	100.0	89.2	10.8
Educational attainment
Less than high school	30.0	84.6	15.4	***
High school	34.4	90.7	9.3	***
More than high school	35.6	91.5	8.5	***
Total sample	100.0	89.2	10.8	***
Later-life characteristics
Age distribution (years)
<75	67.3	89.1	10.9		87.8	12.2
75–85	29.4	89.5	10.5		89.7	10.3
>85	3.3	86.6	13.4		88.9	11.1
Total sample	100.0	89.2	10.8		89.2	10.8
Relationship status
Not partnered	64.8	90.9	9.1	***	91.9	8.1	***
Married or partnered	35.2	85.9	14.1	***	86.4	13.6	***
Total sample	100.0	89.2	10.8	***	89.2	10.8	***
Household income (mean, SD)	42,298.56 (92,144.37)	43,184.67 (94,570.18)	35,009.14 (68,655.13)	*	39,511.87 (51,335.00)	2,090.84 (85,310.17)
Ever drinks alcohol
No	51.3	87	13	***	87.2	12.8	***
Yes	48.7	91.4	8.6	***	91.9	8.1	***
Total sample	100.0	89.2	10.8	***	89.2	10.8	***
Smokes tobacco products
No	91.0	89.6	10.4	**	89.4	10.6	*
Yes	9.0	85.2	14.8	**	85.5	14.5	*
Total sample	100.0	89.2	10.8	**	89.1	10.9	*
Chronic health conditions
None	20.1	92.6	7.4	***	93.0	7.0	***
One	33.6	91.2	8.8	***	92.9	7.1	***
Two or more	46.3	86.2	13.8	***	87.7	12.3	***
Total sample	100.0	89.2	10.8	***	89.2	10.8	***
CES-D score
Less than 4	83.8	91.9	8.1	***	91.9	8.1	***
4 or more	12.9	72.0	28.0	***	76.3	23.7	***
Missing	3.3	88.1	11.9	***	85.8	14.2	***
Total sample	100.0	89.2	10.8	***	89.2	10.8	***
ADL limitations
None	88.6	90.9	9.1	***	91.1	8.9	***
One or more	11.4	75.4	24.6	***	84.2	15.8	***
Total sample	100.0	89.2	10.8	***	89.2	10.8	***
Cognition trajectories
Low-declining	8.2	78.6	21.4	***	78.6	21.4	***
Medium-declining	23.4	87.7	12.3	***	87.7	12.3	***
Medium-stable	38.3	91.2	8.8	***	91.2	8.8	***
High-declining	30.0	90.6	9.4	***	90.6	9.4	***
Total sample	100.0	89.2	10.8	***	89.2	10.8	***

Note: ADL = activities of daily living; CES-D = Center for Epidemiologic Studies—Depression; HRS = Health and Retirement Study.

*p < .05, **p < .01, ***p < .000, ****p < .10.

Individuals reporting psychiatric problems in 1998 (Table 1) were more likely to be classified into the worst faring trajectories, that is, the low- and medium-declining trajectories (21.4% and 12.3%, respectively, p < .000). Finally, almost one fifth of the sample (18.5%) reported living in a nursing home between 2006 and 2016. The incidence rate of living in a nursing home (not reported here) was significantly higher among respondents who reported a history of psychiatric, emotional, or nervous problems (0.038, 95% CI 0.024–0.027) than among those who did not report this history (0.025, 95% CI 0.032–0.044).

Latent class mixture modeling identified four distinct trajectory classes of cognitive function over 8 years (1998–2006; Figure 1). The four identified trajectory classes in the best-fitting model were high cognitive function at baseline and modest declination over time (high-declining), medium cognitive function at baseline and stable over time (medium-stable), medium cognitive function at baseline and declining over time (medium-declining), and low cognitive function at baseline and declining over time (low-declining). The high-declining trajectory was used as the referent category in all subsequent analyses because, relative to the other three trajectory classes, those classified into the high-declining trajectory group had the highest estimated mean scores at baseline and maintained the highest estimated mean cognition scores between 1998 and 2006.

Figure 1.

Cognitive function trajectory classes, Health and Retirement Study 1998–2006. TICS = Telephone Interview for Cognitive Status.

Within the 10-year follow-up period (39,385 person-years), 1,040 cases of nursing home utilization were reported. The crude absolute incidence rates of nursing home use per 1,000 person-years were higher among older adults with psychiatric history (0.038) than among those without this history (0.025). Figure 2 shows the competing risk curves (crude cumulative incidence function) of nursing home use as a function of psychiatric history at baseline in 1998. Similarly, the absolute incidence rates of nursing home use per 1,000 person-years were higher among older adults with low- and medium-declining trajectories (0.052 and 0.042, respectively) and lower among those with medium-stable and high-declining trajectories (0.024 and 0.014, respectively). Competing risks curves also indicated a higher incidence of nursing home use in these groups, as shown in Figure 3. Overall, there were significant differences in cumulative incident curves of nursing home between trajectory classes (χ ² = 257.5, p < .001) and psychiatric history (χ ² = 23.13, p < .001).

Figure 2.

Competing risk curves (crude cumulative incidence function) of nursing home use as a function of psychiatric history at baseline (1998).

Figure 3.

Competing risk curves (crude cumulative incidence function) of nursing home use as a function of cognitive function trajectory class (2006–2016).

Preliminary analysis showed that psychiatric history alone (not displayed), when modeled with all controls, was associated with a 28% greater risk of nursing home use (subhazard ratio [SHR] 1.28, 95% CI 1.28–1.54, p < .001). The effect of psychiatric history as a predictor of nursing home use remained significant when the cognitive function trajectory classes were included in the models. Table 2 displays the results of the hierarchical models predicting the risk of nursing home use as a function of psychiatric history and cognitive function trajectory classes. The independent effect of psychiatric history remained relatively stable while the effect sizes of the three cognitive function trajectories were moderated by the addition of each set of controls. In particular, the addition of controls for later-life health status—number of chronic health conditions, ADL limitations, and depression symptoms—reduced the risk of nursing home use for the low-declining and medium-declining trajectory classes. Model 6—the fully specified model—shows that after adjusting for all relevant covariates, respondents who reported at baseline ever being told by a physician that they had psychiatric, emotional, or nervous problems had a 26% increased risk of nursing home use between 2006 and 2016 (SHR 1.263, 95% CI 1.06–1.51, p < .01). Independent of the significant effect of psychiatric history, cognitive function trajectory classes were associated with increased risk of nursing home use in the low-declining (2.25, p < .000) and medium-declining (2.10, p < .000) trajectory classes, relative to the high-declining class.

Table 2.

Risk of Nursing Home Use as a Function of Psychiatric History and Cognitive Function Trajectories, Health and Retirement Study 2006–2016

	SHR	95% CI		p < t
Trajectory group (ref: High-declining)
Low-declining	2.25	1.70	2.99	***
Medium-declining	2.10	1.69	2.61	***
Medium-stable	1.50	1.24	1.80	***
History of psychiatric problems	1.26	1.06	1.51	**
Baseline characteristics (1998 values)
Gender (ref: Male)
Female	1.46	1.26	1.69	***
Race/ethnicity (ref: White)
Hispanic	0.33	0.23	0.48	***
Black	0.55	0.44	0.68	***
Other	0.73	0.47	1.15
CDI score	0.92	0.74	1.13
Poor child health	1.26	1.00	1.58	*
Educational attainment (ref: High school)
Less than high school	0.90	0.76	1.07
More than high school	0.94	0.81	1.09
Later-life characteristics (2006 values)
Age (ref: Younger than 75)
75–85 years	1.39	1.11	1.75	***
More than 85 years	1.71	1.33	2.20	***
Partnered	1.37	1.19	1.59	**
Household income	1.00	1.00	1.00
Ever drink alcohol	0.97	0.84	1.11
Ever smoke	0.65	0.47	0.91	*
Chronic diseases (ref: None)
One condition	0.94	0.72	1.23
Two or more conditions	0.81	0.64	1.03	**
ADLs (ref: None)
One or more	1.37	1.19	1.58	***
CES-D score (ref: Less than 4)
Between 4 and 8	0.97	0.81	1.16
Missing	2.30	1.93	2.74	***

Notes: ADL = activities of daily living; CDI, childhood disadvantage index; CES-D = Center for Epidemiologic Studies—Depression; SHR = subhazard ratio. Exponentiated coefficients; 95% confidence intervals in brackets.

*p < .05, **p < .01, ***p < .001.

Discussion

The results of these analyses support the first hypothesis, by showing that psychiatric history was independently associated with greater risk of nursing home use net the effects of a variety of life course variables in the model. Furthermore, the second hypothesis is supported by showing that participants in the “low-declining” and “medium-declining” cognition function trajectory classes faced subsequently higher risk of nursing home use. These findings reinforce studies reporting higher rates of health care utilization, including nursing home admissions, among older adults with a history of schizophrenia (Bartels et al., 2003; Hendrie et al., 2014), and a greater likelihood of nursing home admissions among older veterans with more severe mental health problems like schizophrenia and bipolar disorder (Miller & Rosenheck, 2006).

These results elaborate on the previous findings that a history of psychiatric, emotional, or nervous problems was related to poorer cognitive function after age 65 (Brown, 2010) and that older adults diagnosed with serious mental health disorders face a higher probability of being diagnosed with some sort of dementia (Brown & Wolf, 2018). By modeling latent classes of cognition trajectories, we have identified four trajectories of decline that are differentially related to reporting a history of psychiatric problems, with individuals in the lower functioning, less stable cognition trajectory classes being more likely to have reported this history.

Our study assessed 18 years of data and showed clear evidence of cumulative disadvantage. Participants with a history of psychiatric problems in 1998 were more likely to become nursing home residents over time. Similarly, participants with less favorable cognitive functioning trajectories between 1998 and 2006 were more likely to become nursing home residents between 2006 and 2016. This effect was graded, indicating that over time those whose cognitive trajectories started low and decreased further over time were at an increasing health disadvantage relative to those whose intercepts and slopes started and remained relatively high, a result that is not attenuated after accounting for factors across the life course. We further found evidence of increased risk of nursing home use in participants who had poor childhood health (Table 2). Such early disadvantage is likely to affect social determinants of health and wealth which, in turn, affect health outcomes (Ferraro & Shippee, 2009).

As noted previously, history has shown that older adults with serious mental illness receive insufficient services in traditional institutional long-term care (Linkins et al., 2006). Given that individuals with a history of psychiatric problems and unstable trajectories of cognitive decline are at greater risk of nursing home use, nursing home staff should be prepared for the inevitability of providing care to this population. However, nursing home staff are historically ill-equipped to care for residents with psychiatric disorders, and the inability to provide adequate psychiatric care had been proven to be a systemic problem in the nursing home industry (Muralidharan et al., 2019; Tariot et al., 1993). These inadequacies have led to poorer quality of care for all residents in skilled nursing facilities with larger proportions of mentally ill residents (McGarry et al., 2019; Rahman et al., 2013), as well as poorer quality of care for mentally ill residents themselves (Grabowski et al., 2010). Because states differ in regulations addressing the care of residents with serious mental illness, the quality of care they receive also varies across states (Street et al., 2013).

Limitations

These findings should be considered in the context of the limitations of this study. First, these models do not include midlife measures of factors that may contribute to nursing home use in later life, such as episodes of acute illness, or data showing early onset and therefore potential greater severity of chronic conditions. Second, these analyses rely on self-reports of psychiatric, emotional, or nervous problems. These problems may be underreported due to the social stigma faced by individuals with psychiatric disorders and mental health concerns. This underreporting may have affected the strength of the relationships found in these data between psychiatric history, cognition trajectories, and risk of nursing home use. Third, as noted in previous analyses using the measure of psychiatric history (Brown, 2010), HRS data do not provide information about specific psychiatric disorders. Previous studies have indicated that cognitive functioning in later life can be differentially affected by different psychiatric disorders (Brown & Wolf, 2018; Gildengers et al., 2009; Kessing & Nilsson, 2003; Leinonen et al., 2004; Zorrilla et al., 2000). Therefore, the less-specific data employed in this project limit our ability to determine which types of psychiatric disorders are playing a stronger role in influencing cognitive function and the risk of nursing home use in this sample. The inability to distinguish between episodic, acute, or chronic psychiatric disorders may have resulted in an overestimation of the impact of psychiatric history on nursing home use. Finally, these models do not account for earlier nursing home use, so we do not know the patterns of sicker individuals who died before 2006.

Implications for Practice and Research

Current long-term care systems do not provide adequate resources to enable mentally ill older adults, particularly lower-income older adults, to remain independent in the community or to reside in assisted living facilities (ALFs), even though community-based settings are less costly and less restrictive than nursing home care (Becker et al., 2002). Older adults are frequently discharged from hospitals or rehabilitation facilities to ALFs, as these programs provide more supportive services than other types of community-based housing; however, ALF staff often lack both awareness of the needs of mentally ill residents and the training necessary to care for mentally ill residents (Becker et al., 2002; Morgan et al., 2016), and staffing levels are often insufficient to provide the level of care appropriate for residents with mental illness (Beeber et al., 2014). Those older adults with mental illness who do find themselves in ALFs report feeling stigmatized, being isolated due to shunning by others, or self-isolating to avoid conflict with other residents because of real or perceived mental illness (Morgan et al., 2016).

Unfortunately, mentally ill older adults who lack appropriate community-based options often end up in residential long-term care, or skilled nursing facilities, where they must rely on staff who also lack awareness of and training about their specific needs. Muralidharan et al. (2019) identified multiple ways in which the needs of mentally ill residents overlap with those of residents who have dementia, even though the origin of these needs is distinctly different, and recommend integrating training on the care of mentally ill residents with dementia care training because of these similarities. This approach to training should also be considered for training staff in assisted living and other community-based settings.

The disparity between the need for appropriate community-based services and supports for mentally ill older adults and the absence and inadequacy of these services and supports should be of concern to social workers and administrators in long-term care, as nursing homes are often the ultimate destination for many cognitively impaired and mentally disordered older adults, who are more likely to be cognitively impaired (Brown & Wolf, 2018; Grabowski et al., 2010). If other long-term care facilities resist admitting these same older adults because they are not capable of providing psychiatric care, nursing homes should be better equipped to care for them, and state health departments ultimately need to increase the number of nursing home beds to accommodate them. Similarly, if nursing homes can access existing regulations and resources through state health departments, or if organizations like LeadingAge (2020) and the Alzheimer’s Association (2020) lobby for the creation of resources to support the provision of acceptable standards of psychiatric care for these patients, all long-term care facilities should have access to those resources.

Future research should endeavor to capture specific diagnostic information about psychiatric disorders to better elucidate the relationship between these diagnoses and the trajectory classes of cognitive decline identified in this project. This can be accomplished by integrated administrative data sets with existing panel data sets like the HRS. Future studies should analyze larger and more diverse samples of older adults, which would allow for controlling race and ethnicity, and address potentially different outcomes experienced by different cultural groups. Future studies should identify subgroups of older adults who are more vulnerable to psychiatric disorders and to differential rates of cognitive decline after age 65. Models should account for the interactive effects of functional limitations, psychiatric disorders, and trajectories of cognitive decline when modeling the risk of nursing home use. To further identify any life course factors contributing to the risk of nursing home use, researchers should explore longitudinal data that provide young-adult and midlife factors that can serve as pathways for the impact of psychiatric, emotional, and nervous disorders on cognitive function and nursing home use in later life.

Conclusions

Evidence of the relationship between psychiatric history, cognitive decline, and risk of nursing home use can be utilized to enhance public understanding of the impact that psychiatric history has on the long-term care system. These findings can be used to educate policymakers and providers about the need for the regulation of mental health care, and for appropriate psychiatric training for staff, in community-based and residential long-term care programs. Industry leaders should advocate for regulations to invest in the development of, and arm staff with, appropriate training and resources to provide quality care to consumers with or without psychiatric and cognitive disorders.

Funding

None declared.

Conflict of Interest

None declared.

Acknowledgments

The authors would like to acknowledge the support of Syracuse University’s David B. Falk College of Sport and Human Dynamics and the Aging Studies Institute.

Author Contributions

M. T. Brown planned the study and wrote the manuscript. M. Mutambudzi designed and performed all statistical analyses and contributed to writing the manuscript.