Nursing home admissions for persons with dementia: Role of home‐ and community‐based services

Abstract

Objective

To examine the relationship between Medicaid home‐ and community‐based services (HCBS) generosity and the likelihood of nursing home (NH) admission for dually enrolled older adults with Alzheimer's disease and related dementias (ADRD) and their level of physical and cognitive impairment at NH admission.

Data Sources

National Medicare data, Medicaid Analytic eXtract, and MDS 3.0 for CY2010‐2013 were linked.

Study Design

Eligible Medicare–Medicaid dual beneficiaries with ADRD were identified and followed for up to a year. We constructed two measures of HCBS generosity, breadth and intensity, at the county level for older duals with ADRD. Three binary outcomes were defined as follows: any NH placement during the follow‐up year for all individuals in the sample, high (vs. not high) physical impairment, and high (vs. not high) cognitive impairment at the time of NH admission for those who were admitted to an NH. Logistic regressions with state‐fixed effects and county random effects were estimated for these outcomes, respectively, accounting for individual‐ and county‐level covariates.

Data Extraction Methods

The study sample included 365,310 community‐dwelling older dual beneficiaries with ADRD who were enrolled in fee‐for‐service Medicare and Medicaid between October 1, 2010, and December 31, 2012.

Principal Findings

Considerable variations of breadth and intensity in county‐level HCBS were observed. We found that a 10‐percentage‐point increase in HCBS breadth was associated with a 1.4 (p < 0.01)‐percentage‐point reduction in the likelihood of NH admission. Among individuals with NH admission, greater HCBS breadth was associated with a higher level of physical impairment, and greater HCBS intensity was associated with a higher level of physical and cognitive impairment at NH admission.

Conclusions

Among community‐dwelling duals with ADRD, Medicaid HCBS generosity was associated with a lower likelihood of NH admission and greater functional impairment at NH admission.

What is known on this topic

• Medicaid expanded its home‐ and community‐based services (HCBS) in the last few decades to shift the provision of long‐term care from the institutional setting to the community.

• Greater HCBS spending has been found to be associated with reduced or delayed nursing home admission or placement among older adults in general.

• Older adults with Alzheimer's disease and related dementias (ADRD) are at a particular high risk for nursing home admission.

What this study adds

• Greater HCBS generosity, including a higher level of breadth and intensity, was associated with a reduced likelihood of nursing home admission among Medicare–Medicaid dually enrolled older adults with ADRD.

• Among duals with ADRD who were admitted to nursing homes, HCBS generosity, that is, breadth and intensity, was associated with a higher level of physical and cognitive impairment at nursing home admission.

• Efforts to improve HCBS access and to increase service intensity may help duals with ADRD avoid or delay institutionalization and maintain living in the community.

1. INTRODUCTION

More than 10% of people aged 65 years and older have Alzheimer's disease and related dementias (ADRD), and this prevalence increases to 32% among those 85 years and older. ¹ Many persons with ADRD have functional impairment and multiple chronic conditions in addition to cognitive impairment, and they often develop behavioral and psychological symptoms during the course of illness. While persons with ADRD generally rely on the support from informal caregivers to maintain their daily living in the community, ¹ , ² such support may not meet their increasing care needs as the illness progresses. Indeed, almost all community‐dwelling ADRD patients had one or more unmet safety, medical, or personal needs. ³ Therefore, although most prefer to live in the community, ⁴ they are at high risks of nursing home (NH) placement, ⁵ , ⁶ which is associated with further declines in their health status and emotional challenges for families. ⁷ The risk of NH placement is especially high among low‐income individuals with ADRD, who generally are unable to afford assistance and services needed to support their community living.

In the last several decades, Medicaid has been shifting the provision of long‐term services and supports (LTSS) from institutional to the community setting through the expansion of home‐ and community‐based services (HCBS) programs. The increase in Medicaid's investment in HCBS has been substantial, and it has exceeded the spending on NHs since 2013, accounting for the majority of Medicaid LTSS spending. ⁸ Medicaid HCBS programs cover many types of services, such as transportation, personal care, case management, home health aide, and respite care, and provide support for persons with high‐care needs to maintain their community living. Indeed, studies have suggested that Medicaid HCBS may reduce or delay NH admission. ⁹ , ¹⁰ , ¹¹ , ¹² , ¹³ For example, one study examined the association between state‐ and county‐level HCBS generosity (both spending and participation) and NH admissions among Medicaid 1915(c) waiver enrollees and suggested that higher spending on HCBS waiver programs was associated with a lower risk of NH placement among waiver participants, while the role of HCBS participation is unclear. ¹¹ Other studies focused on the spending of Medicaid HCBS (e.g., average waiver expenditure or HCBS spending as a percent of LTSS spending) and found that greater HCBS expenditures were associated with the reduction in the proportion of low‐care needs in NH residents. ⁹ , ¹⁰ In addition, it has also been suggested that a greater use of community‐based services, such as home health and home modifications, may reduce admissions to skilled nursing facilities after an acute event. ¹⁴

Medicare–Medicaid dually enrolled individuals with ADRD use a variety of Medicaid HCBS, such as durable medical equipment, transportation, and personal care, to maintain community living. ¹⁵ However, there is limited evidence on the relationship between HCBS and NH admissions among persons with ADRD. Most existing studies on HCBS use have targeted older adults in general, ¹¹ , ¹² , ¹³ and only few studies were focused on population with ADRD using data from 1980s to 1990s. For example, one study suggested that a higher percentage of Medicaid LTSS spending on HCBS was related to delayed NH admission for unmarried non‐Latino white individuals with Alzheimer's disease. ¹⁶ Another study showed that the use of HCBS provided by a 1915(c) waiver program in Georgia led to delayed NH placements for persons with ADRD. ¹⁷ Since these studies were focused on specific sub‐populations or geographic regions and were based on older datasets, the implications from these studies for current HCBS policies are limited. Moreover, while studies have shown substantial variations across states in the types and volume of services provided to enrollees, ¹⁸ , ¹⁹ , ²⁰ the HCBS generosity measures used in the literature were constructed among the general population and thus may not capture services most relevant to the ADRD population.

Motivated by these existing gaps in the literature, our objective was to examine the relationship between Medicaid HCBS generosity, measured by its breadth and intensity, and NH admissions among community‐dwelling duals with ADRD. As the level of impairment of physical and cognitive status at NH admission may suggest the timing of NH entry, ¹¹ we also examined the relationship between HCBS generosity and the level of physical and cognitive impairment at NH admission. We employed national Medicaid and Medicare data, as well as NH assessments for the time period 2010–2013. By using individual Medicaid claims data, we created HCBS generosity measures specifically for older duals with ADRD at the county level. More specifically, two dimensions of HCBS generosity, that is, breadth and intensity, were measured in this study (details below). These measures have some advantages over the prior measures ⁹ , ¹⁰ because they match the scope of the study population (i.e., population with ADRD), capture variations in HCBS generosity within a state, and disentangle the relationship between a broader HCBS coverage versus more comprehensive services and NH admissions among duals with ADRD.

2. METHODS

2.1. Data

We linked the national Medicaid Analytic eXtract (MAX) personal summary (PS) file, Medicare beneficiary summary file (MBSF) base segment and chronic condition segment, Medicare Provider and Analysis Review (MedPAR), the minimum dataset (MDS) 3.0, and Area Health Resource File (AHRF) data from 2010 to 2013. The MAX PS file contains information on Medicaid enrollees' demographic characteristics, Medicaid enrollment status (e.g., monthly in fee‐for‐service [FFS] or managed care), and service utilization and costs. At the time of this study, MAX data were available nationally from 2010 to 2012 (except for KS and ME for 2010 and ID for 2011) and were available in 28 states in 2013. ²¹ The MBSF base segment includes monthly sociodemographic data, such as Medicare–Medicaid dual status, Medicare Advantage coverage, date of death, and others. The chronic condition segment of MBSF includes 27 chronic conditions, including the diagnosis of ADRD, derived from Medicare claims. ²² The MDS data are a comprehensive assessment tool that is required for all residents in Medicare‐ and/or Medicaid‐certified NHs. The assessments contain detailed information on individual demographics and health status, such as physical functional status and cognitive status at multiple time points, including at admission time and throughout the NH stay. MedPAR file provides information on inpatient stay and was used to identify the hospitalizations in the year prior target NH admission. AHRF is publicly available and includes data on population characteristics, economics, and health facilities at the county level. It was used to construct county‐level covariates in this study.

2.2. Cohort

The study cohort included community‐dwelling older adults with ADRD who were enrolled in Medicare and full‐Medicaid benefits (i.e., duals), regardless of HCBS use. We restricted the cohort to those who were enrolled in Medicare and Medicaid FFS. More specifically, Medicaid‐managed care enrollees who enrolled in comprehensive managed care plans, managed LTSS, and Programs of All‐Inclusive Care for the Elderly were excluded because the variable of interest, HCBS generosity, was constructed based on FFS HCBS users and thus may not capture the services covered through managed care. Medicare Advantage enrollees were excluded because hospitalizations for this subgroup may not be accurately recorded. The identification of ADRD was based on MBSF chronic condition segment—we included individuals who had the diagnoses of “Alzheimer's disease” or “ADRD or senile dementia” prior to the baseline date, as discussed below. We further excluded individuals who were younger than 65 years old. In addition, we excluded individuals who died within 1 year of the selected baseline date (defined below) because they were likely very different than those who did not die with regard to their care needs and services utilization. As this study focused on community‐dwelling individuals, we excluded those who were in a NH for any time within 180 days prior to the baseline date. At the county level, we excluded counties with fewer than 100 eligible duals with ADRD because the county‐level HCBS measurement may not be reliable for a very small population. As a result, 3.6% of individuals in our sample resided in these counties and were excluded.

We set the eligible date as the date when an individual became dually enrolled in Medicare and Medicaid FSS between October 1, 2010, and December 31, 2012, or October 1, 2010 (MDS 3.0 was implemented on October 2010) if an identified individual was dually enrolled prior to this date. An individual could be observed up to 3 years after the eligible date. As we are mainly focused on 1‐year outcome, we randomly selected 1 year as the index “episode” if an individual was observed for 2 or 3 years. The beginning date of the index “episode” was set as the baseline date (e.g., if an individual had the eligible date as January 11, 2010 and the second episode was randomly selected, the baseline date would be January 11, 2010, and the follow‐up period will be January 11, 2010 to October 31, 2012). We did not include individuals who became dual in 2013 because the follow‐up period would be less than 365 days. The final analytical sample included 364,310 dually enrolled community‐dwelling older ADRD patients in 1890 counties.

2.3. Variables

Three dichotomous outcomes were examined in this study. The first outcome was whether an individual had an NH admission during the follow‐up period (dichotomous variable), determined based on MDS 3.0 data. The other two outcomes were defined among those who had an NH admission, including physical and cognitive functioning status at the time of admission, based on the MDS 3.0. Physical function was measured by activities of daily living (ADL) scale, ranging from 0 to 28. ²³ We defined individuals with ADL greater or equal to 19 (47% of the sample) as highly impaired, otherwise as not highly impaired. Cognitive impairment was measured on the cognitive function scale, which ranges from 1 to 4. ²⁴ Individuals with chronic fatigue syndrome (CFS) scores of 3 or 4 (48% of the sample) were defined as high impairment, otherwise as not highly impaired. We also conducted sensitivity analyses using different cut‐offs to dichotomize ADL and CFS.

The main independent variable was the generosity of Medicaid‐funded HCBS. Based on MAX data, we constructed two county‐level measures of HCBS generosity, breadth and intensity, among older duals with ADRD. Breadth and intensity were measured annually based on the utilization and cost of Medicaid FFS HCBS, including both state plan HCBS and waiver services. Breadth was defined as the proportion of duals with ADRD who used any HCBS services in a county each year and thus reflected the scope of HCBS coverage. Intensity was defined as the average HCBS spending per Medicaid‐enrolled month per HCBS user with ADRD, which reflected the amount of services used by an eligible individual. This approach of constructing HCBS generosity measures has been validated among the general population through factor analysis in a previous study. ²⁵ HCBS breadth and intensity of the baseline date year were linked to each individual in our cohort.

We controlled for a comprehensive list of covariates, including demographics (age, gender, and race), days since first ADRD diagnosis, comorbidities, and history of prior health care utilization (i.e., any NH admissions and the number of hospitalizations in the previous year). Length of ADRD history was calculated based on the person's first ADRD diagnosis in MBSF chronic condition segment, and it was accounted for because newly diagnosed ADRD patients and those who lived with ADRD for years could be very different in terms of care needs. We accounted for prior history of NH admissions and hospitalizations because the literature suggested that risk of NH placement was greater among those who had prior events. ²⁶ We also controlled for selected comorbidities that may be related to individual care need, ²⁷ including anemia, chronic kidney disease, chronic obstructive pulmonary disease (COPD), congested heard failure (CHF), diabetes, hypertension, ischemic heart disease, peripheral vascular disease (PVD), stroke, depression, and anxiety. Lastly, we accounted for county‐level characteristics, such as percent of population aged 65 years old or above, median household income, median home value, female labor participation, unemployment rate, poverty rate, and the supply of NH beds and home health agencies, because these factors may be related to both the county‐level HCBS generosity and the likelihood of NH admission.

2.4. Statistical analysis

We first conducted descriptive analyses by describing the distribution of HCBS breadth and intensity at the county level and comparing individual characteristics between those who used NHs versus those who were not during the follow‐up period. We then estimated a logistic model with state‐fixed effects and county random effects to examine the relationship between the likelihood of NH admission and HCBS generosity among duals with ADRD, accounting for individual‐ and county‐level covariates. The unit of HCBS breadth measure was every 10 percentage‐point, and the unit of HCBS intensity measure was every $100. The state‐fixed effects account for state‐level time‐invariant characteristics that may affect NH admission. The county random effects allowed the baseline rate of NH admission to vary across counties and accounted for individual clustering effects within a county.

Among those who had NH admission during the follow‐up period, we separately estimated two logistic models to examine the relationship between HCBS generosity and physical impairment, as well as the cognitive impairment at their NH admission. These two logistic models for physical and cognitive impairment also accounted for state‐fixed effects, county random effects, and individual‐ and county‐level covariates. The average treatment effects for HCBS breadth and intensity were calculated for all models. For easier interpretation, we also calculated the predicted probability of each outcome at the 25th percentile, median, and 75th percentile of HCBS breadth and intensity, respectively.

As HCBS may have different effects on those who enter NHs for postacute care and for long‐term custodian care, we performed a sensitivity analysis by identifying NH admissions only among those who were admitted from the community (those who were admitted to an NH after an inpatient stay were coded as 0). We then repeated the analyses on the relationship between HCBS generosity and individual physical and cognitive impairment only among those who were admitted directly from the community, instead of among all that were admitted to NHs regardless of admission source. In addition, we used different cut‐off points for the binary outcome of physical impairment (severe impairment [ADL ≥24] vs. all others) ²⁸ and cognitive impairment (severe impairment [CFS = 4] vs. all others), and repeated the analysis to check the robustness of the findings.

3. RESULTS

During the study period, the average county‐level HCBS breadth (i.e., the percent of older duals with ADRD using HCBS) was 52.9% (standard deviation 16.4%), and HCBS intensity was $894 (standard deviation $735) per user per month for duals with ADRD. There were considerable geographic variations in both breadth and intensity (Figure 1). In 2012, for example, the county‐level HCBS breadth for duals with ADRD ranged from 15.1% (1st percentile) to 90.0% (99th percentile) with the mean of 51.9% (standard deviation 15.9%); HCBS intensity ranged from $34 (1st percentile) to $2974 (99th percentile) with the mean of $921 (standard deviation $778). Over the study period, the variations across time were smaller. Between 2010 and 2013, the average variations, that is, the difference between the highest and lowest value, in county‐level HCBS breadth and intensity were 10.7%‐point (standard deviation 9.9%) and $332 (standard deviation $376).

FIGURE 1.

County‐level variation in HCBS breadth and intensity, 2012. Note: The range for breadth is 0 to 1. The unit for intensity is one dollar. Data are not available for counties in dark gray. FFS, fee‐for‐service; HCBS, home and community‐based services [Color figure can be viewed at wileyonlinelibrary.com]

On average, 17.1% of the identified community‐residing duals with ADRD used NH during a 1‐year follow‐up period. Table 1 compares individual characteristics, including sociodemographics, prior health care utilization, and clinical diagnoses, between individuals who had an NH admission in the follow‐up year and those who had not. Duals who entered NHs were more likely to be female and non‐Hispanic white. Those who entered NHs were generally frailer than those who remained in the community: they were older, more likely to have hospitalization and NH use in the prior year, and had higher prevalence for most medical conditions (CHF, chronic kidney disease, COPD, hypertension, ischemic heart disease, PVD, and stroke) and mental conditions (depression and anxiety), although they had lower prevalence of diabetes.

TABLE 1.

Descriptive statistics for dual ADRD patients by NH admission status

Variable	NH admission	No NH admission	p‐Value for the
	n = 62,337 (17.1%)	n = 301,973 (82.9%)	difference
Sociodemographics
Age (years)	81.9	80.1	<0.01
Male	26.8%	28.0%	<0.01
Race/ethnicity			<0.01
White	72.9%	54.3%
Black	17.1%	19.0%
Asian	3.2%	10.5%
Hispanic	4.8%	12.1%
Other	2.0%	4.1%
Health conditions
Days since ADRD diagnosis	1420	1551	<0.01
Hospitalization in prior 365 days			<0.01
No hospitalization	53.5%	69.6%
One	20.2%	15.9%
Two or more	15.4%	9.8%
NH use in prior 181–365 days	10.9%	4.7%	<0.01
Clinical diagnoses
Anemia	79.6%	79.9%	0.10
Anxiety	36.5%	32.0%	<0.01
CHF	55.4%	52.0%	<0.01
Chronic kidney disease	40.4%	34.5%	<0.01
COPD	46.2%	43.9%	<0.01
Depression	63.0%	57.7%	<0.01
Diabetes	53.5%	57.9%	<0.01
Hypertension	94.3%	93.5%	<0.01
Ischemic heart disease	71.7%	70.6%	<0.01
PVD	45.5%	44.1%	<0.01
Stroke	38.0%	31.8%	<0.01

Abbreviations: ADRD, Alzheimer's disease and related dementias; CHF, congestive heart failure; COPD, chronic obstructive pulmonary disease; NH, nursing home; PVD, peripheral vascular disease.

Results from regression analyses showed a negative relationship between HCBS breadth and NH admission (Table 2, first column): the average marginal effect of a 10‐percentage‐point increase in HCBS breadth on the probability of NH admission was −1.4 percentage‐point (p < 0.01), accounting for individual‐ and county‐level covariates, state‐fixed effects, and county random effects. The association between HCBS intensity and the probability of NH admission was not significant at 0.05 level (average marginal effect: −0.05 percentage‐point, p = 0.08). The full results from regressions are available at Appendix Table A1.

TABLE 2.

Main results—Average marginal effects of HCBS generosity on NH admission, physical, and cognitive impairment at NH admission, based on logistic regressions

Variables	NH admission	High physical impairment at NH admission	High cognitive impairment at NH admission
HCBS breadth (10‐pct)	−0.014***	0.015***	0.002
	(0.0012)	(0.003)	(0.003)
HCBS intensity ($100)	−0.0005*	0.002**	0.001**
	(0.0003)	(0.0008)	(0.0006)
Observations	364,310	62,194	60,774

Note: *p < 0.1, **p < 0.05, and ***p < 0.01. Average marginal effects are reported. Robust standard errors in parentheses. The model accounted for individual‐ and county‐level covariates, state‐fixed effects, and county random effects.

Abbreviations: HCBS, home‐ and community‐based services; NH, nursing home.

Table 2 also presents the findings from the logistic regressions on the relationship between HCBS generosity and individual physical and cognitive impairment at the time of NH admission. It appeared that a higher level of HCBS breadth was associated with greater physical impairments at NH admission, while higher HCBS intensity was related to greater impairments in both physical and cognitive functioning. More specifically, the average marginal effect of a 10‐percentage‐point increase in HCBS breadth on the probability of having high physical impairment was 1.5 percentage‐point (p < 0.01), relative to having low physical impairment. On the other hand, the average marginal effect of a $100 increase in HCBS intensity on the probability of having high physical impairment at NH admission was 0.2 percentage‐point (p = 0.03), and its average marginal effect on having high cognitive impairment was 0.1 percentage‐point (p = 0.047). The main results are also presented as the predicted probability of each outcome at the 25th percentile, median, and 75th percentile of HCBS breadth and intensity, respectively (Table 3).

TABLE 3.

Predicted probability of outcomes at different HCBS breadth and intensity levels

Variables	NH admission	High physical impairment at NH admission	High cognitive impairment at NH admission
HCBS breadth
25th percentile	0.198	0.439	0.479
50th percentile	0.183	0.455	0.481
75th percentile	0.170	0.469	0.482
HCBS intensity
25th percentile	0.181	0.451	0.477
50th percentile	0.180	0.454	0.479
75th percentile	0.179	0.458	0.482

Abbreviations: HCBS, home‐ and community‐based services; NH, nursing home.

Findings from the sensitivity analysis that focused on NH admissions from the community suggested that greater breadth was associated with greater impairment in both physical and cognitive status (Appendix Table A2). Findings from the sensitivity analysis with alternative cut‐off points for binary physical and cognitive impairment variables were consistent with the main analysis, with less statistically significant associations between HCBS intensity and the outcomes (Appendix Table A3).

4. DISCUSSION

In this study, we constructed HCBS generosity measures (breadth and intensity) among dual‐eligible ADRD patients and examined the association between HCBS generosity and NH admission for community‐dwelling duals with ADRD. Our findings suggest that different aspects of HCBS generosity were associated with reduced or delayed NH admission for this population. More specifically, greater HCBS breadth was related to a reduced risk of NH admission and increased likelihood of have high physical impairment at NH admission. Greater HCBS intensity was associated with a higher level of physical and cognitive impairment at NH admission.

While the relationship between HCBS investment and NH admission has been well studied in the literature, ⁹ , ¹⁰ , ¹¹ , ¹² , ¹³ few studies have attempted to disentangle the roles of broader HCBS coverage and more intensive services. We found an association between HCBS breadth and NH admission among ADRD residents: a 10‐percentage‐point increase in HCBS breadth has an average marginal effect of a 1.4‐percent‐point reduction in the probability of NH admission. It is likely that individuals with ADRD may have more access to services in counties with greater HCBS breadth, which may help them to remain in the community. In addition, in a market with more HCBS users, individuals may have better information and knowledge about HCBS, such as the availability of services and types of services. ²⁹ Therefore, in markets with greater HCBS breadth, even if an individual is not a current HCBS user, they or their family members are more likely to initiate the use of HCBS once additional supports are needed. The association between HCBS intensity and the likelihood of NH admission was relatively small and only marginally significant. Although a higher level of HCBS intensity may help individuals to remain in the community, it is more relevant to those who use HCBS, and thus the effect among all duals could be diluted, which may explain the small effect of HCBS intensity. Future studies focusing on HCBS users may wish to explore the effect of HCBS intensity on delaying NH admissions.

We found that higher levels of HCBS breadth and intensity were associated with higher levels of physical and cognitive impairment at NH admission, further supporting the role of HCBS in delaying NH placement. This finding is consistent with prior studies regarding the relationship between HCBS spending and NH admission among the general older population. ⁹ , ¹¹ , ¹³ The HCBS generosity appears to have a stronger relationship with physical impairment, as compared to cognitive impairment, at the time of NH admission. Medicaid HCBS offers a variety set of services, such as personal care, homemaking, and meal preparation, that support the independent living for individuals with physical disabilities, thus they may be able to remain living in the community longer when the HCBS policies are more generous.

This study had several limitations. First, although we have accounted for many individual characteristics, such as ADRD history, history of hospitalizations and NH admission, and comorbidities, in the analysis, it is possible that there are still unobserved factors that may contribute to the likelihood of NH admission. However, the main variable of interest was the generosity of HCBS, which was constructed at the county level, and we excluded counties with fewer than 100 eligible individuals. Thus, the potential unobserved factors at the individual level are not likely to lead to systematic bias in the relationship between HCBS and the likelihood of NH admission. Second, our study is only focused on the association, rather than causal relationship, between HCBS generosity and NH admission. There may be some endogeneity concerns with HCBS generosity and outcomes. For example, although we have accounted for county‐level economic characteristics, long‐term care supply, and county random effects, some uncaptured county‐level factors and other programs supporting community living (e.g., Medicare home health services) may indirectly influence HCBS generosity as well as NH admission. In addition, we measured HCBS intensity as the average spending among all eligible individuals in the county. While it is possible that an individual spending is related to both the intensity measure as well as the outcome, we do not expect such influence to be large as we only included counties with at least 100 eligible individuals. Third, this study excluded individuals who used any nursing facility services within 6 months prior to baseline. Thus, our study population did not include community‐dwelling individuals who had a short postacute care stay in nursing facilities. However, these individuals may have different care needs than the population involved in this study. Interestingly, a recent study also suggested that HCBS generosity was related to an increased likelihood of community discharge among individuals who received postacute service. ³⁰ Lastly, this study was focused on individuals enrolled in Medicare and Medicaid FFS, and the findings may not be generalized to those who were enrolled in managed care plans. Although the Medicaid‐managed care enrollment in duals with ADRD is minimal in most counties between 2010 and 2013, in counties that had large managed care penetration, the FFS enrollees included in our study may represent only a small proportion of all duals with ADRD. In spite of these limitations, this study offers new contribution to the literature. To the best of our knowledge, this is the first national study to examine the role of HCBS in reducing and/or delaying NH admissions among dual‐eligible ADRD population and that disentangles the effect of HCBS breadth and intensity among this vulnerable population.

In conclusion, duals with ADRD are at high risk of NH placement as they are socioeconomically disadvantaged, have significant physical and cognitive impairment, and largely rely on informal care to remain living in the community. This study presents the evidence that higher levels of HCBS were associated with reduced risks of NH admission among duals with ADRD. Thus improving HCBS access and increasing service intensity may delay or reduce institutionalization for this population. As Medicaid‐managed care becomes more widespread, it will be important for future research to investigate the role of HCBS breadth/intensity in reducing NH admission within managed care.

APPENDIX A.

TABLE A1.

Full results of main analysis—Associations between HCBS generosity and community‐originated NH admission, physical, and cognitive impairment at NH admission for duals with ADRD

Variables	NH admission	Physical impairment at admission (highly impaired)	Cognitive impairment at admission (highly impaired)
Breadth	0.903***	1.067***	1.008
	(0.00769)	(0.0154)	(0.0128)
Intensity	0.997*	1.008**	1.006**
	(0.00201)	(0.00352)	(0.00279)
Age	1.036***	1.022***	1.040***
	(0.000621)	(0.00118)	(0.00122)
Male	1.124***	0.916***	1.057***
	(0.0121)	(0.0184)	(0.0213)
Race: Asian	0.364***	1.480***	1.402***
	(0.00936)	(0.0793)	(0.0739)
Race: black	0.682***	1.329***	1.281***
	(0.00952)	(0.0338)	(0.0327)
Race: Hispanic	0.454***	1.291***	1.147***
	(0.00988)	(0.0588)	(0.0501)
Race: other	0.509***	1.180***	1.314***
	(0.0162)	(0.0751)	(0.0829)
Days since ADRD diagnosis	1.000***	1.000***	1.000***
	(4.15e‐06)	(7.76e‐06)	(7.84e‐06)
Prior hospitalization, one	1.185***	1.049**	0.940***
	(0.0146)	(0.0236)	(0.0211)
Prior hospitalization, two or more	1.397***	1.075***	0.871***
	(0.0206)	(0.0286)	(0.0236)
Prior NH use	1.635***	1.078***	0.797***
	(0.0277)	(0.0307)	(0.0235)
Chronic kidney disease	1.112***	1.033*	0.877***
	(0.0115)	(0.0196)	(0.0167)
COPD	1.004	0.885***	0.720***
	(0.0101)	(0.0164)	(0.0132)
CHF	1.009	1.082***	0.759***
	(0.0108)	(0.0215)	(0.0150)
Diabetes	0.968***	1.015	0.815***
	(0.00969)	(0.0188)	(0.0151)
Ischemic heart disease	1.008	0.944***	0.890***
	(0.0118)	(0.0205)	(0.0193)
Depression	1.152***	0.995	0.877***
	(0.0121)	(0.0194)	(0.0171)
Stroke	1.185***	1.191***	0.970
	(0.0118)	(0.0217)	(0.0178)
Anemia	0.953***	1.067***	0.848***
	(0.0117)	(0.0244)	(0.0194)
Hypertension	1.049**	0.971	0.721***
	(0.0216)	(0.0379)	(0.0288)
Anxiety	1.037***	0.844***	0.768***
	(0.0109)	(0.0164)	(0.0149)
PVD	1.029***	1.054***	0.853***
	(0.0104)	(0.0196)	(0.0159)
Percent of population who are of age 65+	2.371***	0.736	1.328
	(0.757)	(0.366)	(0.496)
Median household income 2010 (in thousand $)	1.008***	1.008***	1.006***
	(0.00184)	(0.00287)	(0.00205)
Female labor participation	1.002	0.972***	1.000
	(0.00478)	(0.00735)	(0.00616)
Unemployment rate	0.973	0.919	0.924
	(0.0639)	(0.104)	(0.0901)
Median home value (in thousand $)	0.998***	1.001***	0.999***
	(0.000182)	(0.000264)	(0.000155)
Poverty rate	0.989***	1.016***	1.005
	(0.00310)	(0.00515)	(0.00410)
HHA per 1000 population	0.664*	1.476	0.998
	(0.161)	(0.572)	(0.275)
NH beds per 1000 population	1.002	0.985	1.022*
	(0.00899)	(0.0153)	(0.0131)
Constant	0.0211***	0.438	0.0702***
	(0.0101)	(0.335)	(0.0440)
Observations	364,310	62,194	60,774

*p < 0.1, **p < 0.05, and ***p < 0.01. Odds ratios are reported. Robust standard errors in parentheses.

Abbreviations: ADRD, Alzheimer's disease or related dementias; CHF, congestive heart failure; COPD, chronic obstructive pulmonary disease; HCBS, home and community‐based services; HHA, home health agency; NH, nursing home; PVD, peripheral vascular disease.

TABLE A2.

Full results of sensitivity analysis—Associations between HCBS generosity and community‐originated NH admission, physical, and cognitive impairment at NH admission for individuals admitted from community

Variables	NH admission	Physical impairment at admission (highly impaired)	Cognitive impairment at admission (highly impaired)
Breadth	0.899***	1.072***	1.043**
	(0.0103)	(0.0220)	(0.0194)
Intensity	0.993**	1.005	1.003
	(0.00301)	(0.00507)	(0.00445)
Age	1.037***	1.013***	1.034***
	(0.000886)	(0.00184)	(0.00187)
Male	1.237***	0.888***	0.928**
	(0.0189)	(0.0281)	(0.0287)
Race: Asian	0.389***	1.758***	1.493***
	(0.0154)	(0.145)	(0.129)
Race: black	0.640***	1.431***	1.266***
	(0.0134)	(0.0587)	(0.0527)
Race: Hispanic	0.470***	1.444***	1.159**
	(0.0163)	(0.105)	(0.0846)
Race: other	0.524***	1.275**	1.193*
	(0.0255)	(0.130)	(0.121)
Days since ADRD diagnosis	1.000***	1.000***	1.000***
	(5.98e‐06)	(1.23e‐05)	(1.25e‐05)
Prior hospitalization, one	0.939***	1.103***	0.917**
	(0.0172)	(0.0409)	(0.0333)
Prior hospitalization, two or more	0.957*	1.165***	0.840***
	(0.0219)	(0.0534)	(0.0383)
Prior NH use	1.570***	1.073	0.742***
	(0.0375)	(0.0494)	(0.0343)
Chronic kidney disease	0.972*	1.059*	0.877***
	(0.0146)	(0.0324)	(0.0264)
COPD	0.916***	0.934**	0.694***
	(0.0132)	(0.0276)	(0.0199)
CHF	0.922***	1.118***	0.766***
	(0.0141)	(0.0348)	(0.0233)
Diabetes	0.907***	1.026	0.846***
	(0.0129)	(0.0300)	(0.0242)
Ischemic heart disease	0.956***	0.964	0.898***
	(0.0155)	(0.0319)	(0.0292)
Depression	1.123***	1.022	0.854***
	(0.0168)	(0.0312)	(0.0256)
Stroke	1.133***	1.244***	0.920***
	(0.0163)	(0.0362)	(0.0265)
Anemia	0.921***	1.100***	0.859***
	(0.0154)	(0.0380)	(0.0289)
Hypertension	0.977	0.983	0.739***
	(0.0263)	(0.0553)	(0.0416)
Anxiety	1.006	0.846***	0.749***
	(0.0153)	(0.0263)	(0.0226)
PVD	1.007	1.097***	0.850***
	(0.0146)	(0.0323)	(0.0247)
Percent of population who are of age 65+	7.470***	1.113	2.455
	(3.307)	(0.750)	(1.406)
Median household income 2010 (in thousand $)	1.009***	1.004	1.011***
	(0.00258)	(0.00383)	(0.00320)
Female labor participation	1.008	0.977**	1.013
	(0.00666)	(0.00994)	(0.00908)
Unemployment rate	1.156*	1.088	0.929
	(0.0984)	(0.162)	(0.123)
Median home value (in thousand $)	0.998***	1.001***	0.999***
	(0.000259)	(0.000330)	(0.000247)
Poverty rate	0.994	1.020***	1.007
	(0.00433)	(0.00707)	(0.00611)
HHA per 1000 population	0.676	2.330	0.746
	(0.223)	(1.208)	(0.304)
NH beds per 1000 population	1.010	1.002	1.045***
	(0.0116)	(0.0199)	(0.0174)
Constant	0.00391***	0.403	0.0261***
	(0.00260)	(0.420)	(0.0242)
Observations	364,310	26,338	25,585

Note: *p < 0.1, **p < 0.05, and ***p < 0.01. Odds ratios are reported. Robust standard errors in parentheses.

Abbreviations: ADRD, Alzheimer's disease or related dementias; CHF, congestive heart failure; COPD, chronic obstructive pulmonary disease; HCBS, home and community‐based services; HHA, home health agency; NH, nursing home; PVD, peripheral vascular disease.

TABLE A3.

Full results of sensitivity analysis—Associations between HCBS generosity and community‐originated NH admission, physical, and cognitive impairment at NH admission with alternative ADL and CFS cut‐offs

Variables	Severe physical impairment (ADL > =24)	Severe cognitive impairment (CFS = 4)
Breadth	1.102***	1.008
	(0.0293)	(0.0229)
Intensity	1.009*	1.005
	(0.00516)	(0.00478)
Age	1.009***	0.998
	(0.00193)	(0.00206)
Male	0.918**	0.911**
	(0.0306)	(0.0333)
Race: Asian	1.945***	1.742***
	(0.131)	(0.140)
Race: black	1.728***	1.375***
	(0.0656)	(0.0591)
Race: Hispanic	1.275***	1.067
	(0.0758)	(0.0819)
Race: other	1.422***	1.542***
	(0.129)	(0.152)
Days since ADRD diagnosis	1.000***	1.000***
	(1.22e‐05)	(1.30e‐05)
Prior hospitalization, one	1.072*	1.005
	(0.0400)	(0.0421)
Prior hospitalization, two or more	1.212***	1.060
	(0.0519)	(0.0552)
Prior NH use	1.010	0.826***
	(0.0492)	(0.0497)
Chronic kidney disease	0.952	0.826***
	(0.0300)	(0.0295)
COPD	0.812***	0.702***
	(0.0253)	(0.0243)
CHF	0.977	0.782***
	(0.0323)	(0.0279)
Diabetes	0.949*	0.830***
	(0.0297)	(0.0280)
Ischemic heart disease	0.881***	0.849***
	(0.0323)	(0.0315)
Depression	0.972	0.911***
	(0.0312)	(0.0313)
Stroke	1.305***	1.080**
	(0.0390)	(0.0362)
Anemia	0.998	0.886***
	(0.0400)	(0.0345)
Hypertension	0.843***	0.718***
	(0.0554)	(0.0412)
Anxiety	0.807***	0.786***
	(0.0269)	(0.0289)
PVD	0.985	0.879***
	(0.0308)	(0.0302)
Percent of population who are of age 65+	0.112***	0.665
	(0.0853)	(0.439)
Median household income 2010 (in thousand $)	1.002	1.002
	(0.00410)	(0.00357)
Female labor participation	0.962***	0.962***
	(0.0108)	(0.0104)
Unemployment rate	0.921	1.010
	(0.176)	(0.184)
Median home value (in thousand $)	1.001***	1.000
	(0.000340)	(0.000271)
Poverty rate	1.016**	0.996
	(0.00777)	(0.00730)
HHA per 1000 population	2.348	4.324***
	(1.373)	(2.140)
NH beds per 1000 population	1.032	0.987
	(0.0267)	(0.0284)
Constant	0.854	7.482*
	(0.971)	(8.216)
Observations	62,045	60,755

*p < 0.1, **p < 0.05, and ***p < 0.01. Odds ratios are reported. Robust standard errors in parentheses.

Abbreviations: ADLs, activities of daily living; ADRD, Alzheimer's disease or related dementias; CFS, chronic fatigue syndrome; CHF, congestive heart failure; COPD, chronic obstructive pulmonary disease; HCBS, home and community‐based services; HHA, home health agency; NH, nursing home; PVD, peripheral vascular disease.

Wang S, Yan D, Temkin‐Greener H, Cai S. Nursing home admissions for persons with dementia: Role of home‐ and community‐based services. Health Serv Res. 2021;56(6):1168‐1178. 10.1111/1475-6773.13715