State Regulations and Hospice Utilization in Assisted Living during the Last Month of Life

Abstract

Objectives:

To examine the association between hospice/staffing regulations in residential care / assisted living (RC/AL) and hospice utilization among a national cohort of Medicare decedents residing in RC/AL at least one day during the last month of life, and to describe patterns of hospice utilization.

Design:

Retrospective cohort study of fee-for-service Medicare beneficiaries who died in 2018 and resided in an RC/AL community with 25+ beds at least one day during the last month of life.

Setting/Participants:

23,285 decedents who spent time in 6,274 RC/AL communities with 146 state license classifications.

Methods:

Descriptive statistics about hospice use; logistic regression models to test the association between regulations supportive of hospice care / registered nurse (RN) staffing requirements and the odds of hospice use in RC/AL in the last month of life.

Results:

56.4% of the study cohort received hospice care in RC/AL at some point during the last 30 days of life, including 5.7% who received more intensive continuous home care (CHC). A larger proportion of decedents who resided in RC/ALs with supportive hospice policies received hospice (57.3% vs 52.6%), with this difference driven by more CHC hospice programs. This association remained significant after controlling for sociodemographic characteristics, comorbidities, time spent in RC/AL and Hospital Referral Region fixed effects. Decedents in RC/ALs with explicit RN staffing requirements had significantly less CHC use (2.0% vs 6.8%).

Conclusions and Implications:

A large proportion of RC/AL decedents received hospice care in RC/AL regardless of differing regulations. Those in licensed settings with explicitly supportive hospice regulations were significantly more likely to receive hospice care in RC/AL during the last month of life, especially CHC level of hospice care. Regulatory change in states that do not yet explicitly allow hospice care in RC/AL may potentially increase hospice utilization in this setting, although the implications for quality of care remain unclear.

Brief summary:

Hospice utilization was higher during the last month of life among assisted living decedents in licensed settings with supportive hospice regulations, than those residing in settings without these supportive regulations.

Introduction

The growing number of seriously-ill older adults remaining in residential care or assisted living (RC/AL) until the end of life in the United States presents unique challenges for policy-makers trying to balance patient autonomy and safety. By 2015, over 90,000 fee-for-service Medicare beneficiaries were receiving hospice services in RC/AL at the time of death, representing 18% of hospice recipients dying in the community.¹ A study of a representative sample of US Medicare decedents derived from the Medicare Current Beneficiary Survey (MCBS) between 2002-2017 confirmed this increase in the number of beneficiaries residing in community-based residential care settings at end of life, from 8% to 11% of national decedents over the study period.² The prevalence of Alzheimer’s disease and related dementia (ADRD) is also increasing among RC/AL residents,^3-5 alongside national trends that make ADRD the fifth leading cause of death among those age 65 and older.⁶ Determining the six-month prognosis required to be eligible for hospice is particularly challenging among persons with dementia, as evidenced by their higher likelihood of disenrolling from hospice care after a long stay,^7,8 and the lower likelihood of being referred to hospice at all.⁹

The level of state and federal oversight required to protect vulnerable RC/AL residents is a matter of enduring debate.^10,11 RC/AL regulations vary greatly between and even within states across different license types,^12,13 and these policies are likely to influence the provision of hospice and end-of-life care. In a recent inventory of RC/AL regulations pertaining to dementia care, all states were found to have at least one regulation on the topic, but many required a separate license or certificate.¹³ Other states required a dementia-specific state agency review, and still others exerted less regulatory control and only included some dementia care provisions within their respective RC/AL regulations.¹³ Previous studies have shown that state-level policies are significantly associated with change in hospice use over time in other settings, for example as they relate to Medicaid reimbursement rates for nursing home residents.¹⁴

Yet, we do not have a good understanding of what share of RC/AL residents receive hospice before death and what policies may be driving these trends. Prior work has demonstrated important state-level variability in end-of-life care trajectories among a national cohort of RC/AL residents, with variation not only in the number of days decedents spent in RC/AL during the last month of life, but also in the number of days receiving hospice services.¹⁵ For the most part, our knowledge about end-of-life care provided in RC/AL remains limited either to those who receive hospice care,¹⁶ which we can detect from place of service codes in hospice claims, or to those participating in primary data collection efforts often in a small number of states.¹⁷ To date no national study has examined whether hospice utilization in RC/AL varies by state regulations.

This study had two objectives. The first objective was to examine the association between regulations of licensed RC/AL settings and hospice utilization, especially among vulnerable residents with ADRD. We hypothesized that decedents in licensed settings with regulations explicitly supportive of hospice would be more likely to receive hospice care in AL during the last month of life. We also hypothesized that regulations requiring registered nurses’ (RN) presence in RC/AL would be related to decreased hospice use in the last month of life because RNs may substitute some of the need for hospice care. The second objective was to describe detailed patterns of hospice utilization among RC/AL residents receiving hospice. There has been limited research examining the type of hospice care provided to dying RC/AL residents, particularly the use of routine home care (RHC), which is the most common type of hospice care provided in the location a beneficiary calls home, as compared with the higher intensity continuous home care (CHC), which entails the provision of at least 8 hours, and up to 24 hours of care per day, mostly by a registered nurse to maintain a terminally-ill patient at home during a medical crisis.¹⁸

Methods

We conducted a retrospective cohort study of fee-for-service Medicare beneficiaries who died in 2018 and resided in an RC/AL community with 25 beds or more at some point during the last month of life. We described their use of hospice and the association between regulations supportive of hospice care and about minimum RN staffing requirements and the odds of hospice use in the last month of life. This study was deemed exempt from review by the Institutional Review Board at Brown University on the basis of using secondary administrative claims data.

Data Sources and Study Population

Data pertaining to the list of RC/AL communities in operation during 2017, such as license type, ZIP code, and capacity were retrieved from state licensing agencies, which keep a public record of the RC/ALs licensed, registered, certified or otherwise regulated by these states. This information was supplemented with data collected through a national inventory of the regulations in effect in 2017-2018 for licensed RC/AL care settings with 25+ beds.¹² Regulatory data were linked to individual RC/ALs based on the license classification of each individual RC/AL. The licenses refer to unique combinations of regulations in effect across the RC/AL communities in our sample. For example, a given state may have a primary license type for “assisted living", which sometimes occurs by itself and sometimes co-occurs with a designation for a dementia care unit. Both of these are considered distinct license types in our analyses, whereby there may be a different set of regulations guiding end-of-life care in each setting.

Data on hospice use came from Medicare hospice claims that were linked to individual RC/AL residents using the Medicare Beneficiary ID. Date of death was obtained from the Medicare Beneficiary Summary File. We relied on a previously published methodology using beneficiaries’ 9-digit ZIP codes from the Master Beneficiary Summary File (MBSF) to identify RC/AL residents.¹⁹ A finder file of validated ZIP codes for licensed RC/AL communities with a capacity of 25 beds or more in 2017 was used to identify Medicare beneficiaries with ZIP codes that were shared with an RC/AL. This process was restricted to beneficiaries with a date of death in 2018. Because of our interest in the relationship between regulations and hospice utilization at the end of life in RC/AL, only decedents with at least one day of presence in RC/AL during the last month of life were retained in these analyses (n= 23,285). We also required continuous enrollment in fee-for-service Medicare during the last year of life to obtain an accurate record of decedents’ comorbidities, and therefore excluded any beneficiaries enrolled in Medicare Advantage.

Study Measures

For this study, we focus on two policies particularly relevant to end-of-life care delivery in RC/AL, which are coded as being present or not present in regulations, specifically 1) explicit policies allowing the delivery of hospice care and/or explicit mention that the admission of residents on hospice is permitted, and 2) any registered nurse (RN) staffing requirements (see Appendix for complete list of policy items). These are policies that we expect may affect the ability to ramp up the intensity of care during the last month of life to meet the increasingly complex care needs among dying residents. If these regulations applied to any license type in effect in a given RC/AL, then residents in that RC/AL were assigned the value of “present” (=1), else “not present”(=0). Licensed settings without these regulations can be best described as silent rather than restrictive.

The primary outcome measure is a binary variable for receipt of any hospice care in RC/AL during the last month of life. This indicator is derived from hospice claims and from the Resident History File developed at Brown University, which provides information about daily healthcare utilization of Medicare beneficiaries.²⁰ In order of certainty, we considered that the presence of an inpatient hospitalization, nursing home claims and/or Minimum Data Set (MDS) assessments, or general inpatient hospice/inpatient respite claims were indicative that a given resident in our decedent cohort was no longer in RC/AL on that day. We finally explored different hospice utilization patterns on the basis that these may shape end-of-life care delivery in this setting, particularly the recourse to CHC.

Statistical Analyses

We first provide descriptive statistics about this cohort of Medicare decedents with RC/AL utilization at the end of life and their overall hospice utilization. We also report the distribution of regulations of interest among the RC/AL communities, along with the percentage of decedents who received hospice care before death. Unadjusted and adjusted logistic regressions were estimated for each independent regulation variable on the likelihood of receiving hospice care in RC/AL. Important covariates in our models include socio-demographic characteristics such as a categorical age variable (65-74 / 75-84 / 85+), sex (male/female), and race/ethnicity (White, non-Hispanic / Black, non-Hispanic / White, Hispanic / Other), as well as being dually enrolled in Medicare and Medicaid at any point in 2017-2018 (yes/no), and number of years spent in RC/AL. We included decedents’ total number of chronic conditions in the Chronic Condition Warehouse (CCW) (0-25). We estimated multivariable logistic regression models adjusted for all covariates and a hospital referral region (HRR) fixed effect to control for regional practice patterns. We finally examined whether there were significant differences in the likelihood of receiving hospice care among residents with a diagnosis of ADRD using the CCW end-of-year flags in 2017 and 2018.

Results

Description of Study Cohort and Overall Hospice Utilization

Our final study cohort consisted of 23,285 decedents who spent time in 6,274 RC/AL communities with a total of 146 unique license classifications during the last month of life. Overall, 56.4% of our sample of decedents received hospice care in RC/AL at some point during the last 30 days of life, including 5.7% who received more intensive CHC. A majority (66.5%) of our decedent cohort had a diagnosis of ADRD. Table 1 highlights socio-demographic characteristics: specifically, 80.3% were 85 years or older, 64.0% were female, and the majority of the cohort was non-Hispanic White (95.0%). Slightly more residents with a diagnosis of ADRD were dually enrolled in Medicaid and Medicare during the study period, (13.6% vs 10.1% among those without ADRD). On average, decedents had been diagnosed with 7.7 chronic conditions and had been in RC/AL for 5.0 years, although time spent in RC/AL had wide variation with a 4.7-year standard deviation.

Table 1:

Descriptive Characteristics of 2018 Medicare Decedents with Presence in RC/AL during the Last Month of Life

	All	Alzheimer’s Disease and related Dementia	Non-Alzheimer’s Disease and related Dementia
Total
N (%)	23,285	15,494 (66.5)	7,791 (33.5)
Age- N (%)
65-74	775 (3.3)	388 (2.5)	387 (5.0)
75-84	3,808 (16.4)	2,339 (15.1)	1,469 (18.9)
85+	18,702 (80.3)	12,767 (82.4)	5,935 (76.2)
Sex- N (%)
Men	8,381 (36.0)	5,397 (34.8)	2,984 (38.3)
Women	14,904 (64.0)	10,097 (65.2)	4,807 (61.7)
Race / Ethnicity- N (%)
Non-Hispanic, Caucasian	22,126 (95.0)	14,665 (94.7)	7,461 (95.8)
Non-Hispanic, African American	513 (2.2)	394 (2.5)	119 (1.5)
Hispanic, Caucasian	301 (1.3)	223 (1.4)	78 (1.0)
Other	345 (1.5)	212 (1.4)	133 (1.7)
Dual Eligible in 2017/2018-N (%)
Yes	2,897 (12.4)	2,113 (13.6)	784 (10.1)
No	20,388 (87.6)	13,381 (86.4)	7,007 (89.9)
Chronic Conditions in 2017/2018 (0-25)
Mean (SD)	7.69 (3.1)	7.72 (3.1)	7.63 (3.1)
Years Spent in RC/AL
Mean (SD)	5.04 (4.7)	4.91 (4.6)	5.31 (4.8)

RC/AL= Residential care or assisted living, SD= standard deviation

Hospice Utilization as Factor of RC/AL Regulations

The majority of decedents (82.2%) resided in a licensed care setting with at least one applicable license having supportive hospice policies and 23.2% with at least some requirements for RN staffing. In 2018, there were 9 states and 45 (30.8%) license types in effect across 1,304 (20.8%) RC/AL communities where hospice care was not explicitly allowed, while 10 states allow hospice explicitly in some license types but not others. As Figure 1 illustrates, a significantly larger proportion of decedents who resided in RC/ALs with supportive hospice policies received hospice in RC/AL (57.3% vs 52.6%), with this difference being driven by the use of more intensive CHC hospice programs (6.2% vs 3.2% for CHC). Decedents who resided in RC/ALs with explicit RN staffing requirements had similar hospice use (56.0% vs 56.5%), but there was a significant difference in CHC use (2.0% vs 6.8%).

Figure 1:

Percentage of decedents receiving hospice care in RC/AL during the last month of life by regulation. *P value < .001. Note that these numbers represent hospice services received in RC/AL; individual decedents may have received different types of hospice services in other care settings.

As can be seen in Table 2, unadjusted bivariate logistic regression models confirm that residents in licensed care settings with supportive hospice regulations were significantly more likely (OR 1.21; 95% CI: 1.13-1.29) to receive hospice care in AL during the last 30 days of life. The relationship remained significant despite adjusting for sociodemographic characteristics, comorbidities, time spent in AL, and an HRR fixed effect. The majority (95.6%) of the variance in hospice utilization was found to be within HRRs, indicating that most of the variance in this outcome occurs within these regional units rather than between them. As expected, residents in settings with minimum RN staffing requirements were slightly less likely (OR 0.98; 95% CI: 0.92-1.04) to receive hospice care in AL, although this effect was not significant. As part of sensitivity analyses, we examined the same models with the likelihood of receiving hospice in AL during the last 7 days of life for the same cohort, as well as among those with a diagnosis of ADRD and the results were very similar (see Appendix).

Table 2:

Odds Ratios of Decedents Receiving Hospice Care in RC/AL during the Last 30 Days of Life (95% confidence intervals)

	Unadjusted N=23,285	Adjusted* N=23,285	Adjusted* with HRR Fixed Effects N= 23,268
Supportive Hospice Regulations	1.21 (1.13-1.29)	1.16 (1.08-1.24)	1.20 (1.02-1.42)
Any RN Staffing Requirements	0.98 (0.92-1.04)	0.95 (0.89-1.01)	1.05 (0.90-1.22)

^*Adjusted for age, sex, race / ethnicity, dual eligibility in 2017 and/or 2018, ADRD diagnosis in 2017 and/or 2018, number of chronic conditions in 2017 and/or 2018, number of years spent in AL.

RC/AL= Residential care or assisted living, ADRD= Alzheimer’s disease and related dementia, HRR: Hospital Referral Region

Hospice Utilization Patterns and Place of Death

Figure 2 represents the proportional flow of RC/AL residents by place of death between those who received CHC in RC/AL, those who received RHC only in RC/AL, and those who received no hospice care in RC/AL. The vast majority of decedents who received any hospice in RC/AL during the last month of life remained there on the date of death, especially those receiving CHC. Those who did not receive hospice care in RC/AL for the most part transferred to a different care setting without ever receiving hospice there either, although 11.2% of the study cohort died in RC/AL without ever receiving hospice services. We found that there was only a small difference in the proportion of residents transferring to hospice in another setting between states with (12.5%) and without (12.3%) explicitly supportive hospice regulations in RC/AL. Of those 2,904 residents who did not receive hospice in RC/AL but died somewhere else with hospice: 68.8% (1,997) died in the hospital or in observation/ER with hospice, and 31.1% (904) died in the nursing home with hospice.

Figure 2:

Flowchart of decedents by hospice utilization in RC/AL and place of death.

Discussion

Our results demonstrate that over half of decedents with presence in RC/AL during the last month of life receive hospice care in this setting, regardless of differing regulations. We also found a significant association between the likelihood of receiving hospice care in RC/AL during the last month of life, especially more intensive CHC services, and license-specific regulations explicitly allowing hospice care and/or the admission of residents already enrolled in hospice. Despite a growing number of seriously-ill Medicare beneficiaries residing in RC/AL,¹ we found that nine states do not explicitly allow hospice care, and that 10 do so in only some licenses. A large proportion of decedents also remain in RC/AL without this important benefit, and yet many others transfer to receive hospice care elsewhere or experience further transfers without ever receiving hospice care. These results were consistent among decedents with a diagnosis of ADRD, which represents 66.5% of our national sample.

Prior work has shown major differences across states and RC/AL communities in terms of philosophies about end-of-life care and hospice utilization.¹⁵ Our results begin to unravel some of the underlying regulatory landscape associated with this variation. We cannot establish a causal relationship between hospice regulations and hospice utilization in RC/AL, because it is hard to determine whether supportive regulations prompted increased hospice use by removing barriers, or if a stronger hospice industry and advocates prompted the regulations to go into place. However, this relationship remained significant for supportive hospice policies even after controlling for different hospice markets using HRR fixed effects. This HRR fixed effect is likely to capture a variety of additional factors such as regional variation in aggressiveness of end-of-life care and local practice patterns;²¹ and we acknowledge that the best geographic unit to control for different hospice markets remains to be determined.²²

The characteristics of our RC/AL decedent cohort and level of hospice utilization are consistent with other data sources. Previous work examining hospice place of service codes found that a large proportion of Medicare beneficiaries who receive hospice in RC/AL are 85 years and older and have a dementia diagnosis.⁴ Data from the MCBS between 2002-2017 show that hospice utilization was higher among decedents in community-based residential care settings than in private residences, with the subtype of AL residents reaching the highest level of use at 61.1%.² We also make novel contributions about patterns of hospice utilization in RC/AL. For example, a study of intensity of hospice services among AL residents from the medical records of a large for-profit national hospice provider focused only on RHC.¹⁶ In line with a previous study of Medicare decedents concluding that beneficiaries receiving CHC were less likely to disenroll from hospice and be hospitalized,²³ our finding that nearly all residents who received CHC remained in place until death suggests that the higher intensity of care provided with CHC may play a particularly important role in managing breakthrough symptoms in RC/AL.

Our results indicate that RN staffing requirements are not significantly associated with the odds of receiving hospice care, but there was a lower use of CHC suggesting some substitution in more intensive periods of care. While we expected some substitution effect, whereby the presence of more nursing staff may reduce the need for hospice services, there are many contextual factors affecting staffing levels beyond staffing requirements. Data from the 2010 National Survey of Residential Care Facilities confirm that RNs are quite prevalent across RC/AL communities, despite them being required in only 37.0% of licenses in our 2018 policy inventory. Previous work confirmed that RN presence was associated with a higher likelihood of AL administrators reporting that they admit and/or retain residents in need of end-of-life care.²⁴ Staffing levels are also likely determined by residents’ care needs and their families’ preferences, as well as competitiveness of local long-term care markets.¹³

There remain concerns about how adequately terminal symptoms can be managed in RC/AL even with access to hospice care, which we were not able to address with our study design. Dying long-term care residents are known to have a clinically important burden of physical and psychosocial symptoms that call for increased palliation;^25,26 yet residents with dementia were less likely to receive hospice care in a study of 127 RC/AL decedents across four states.⁹ Using electronic health records from a not-for-profit hospice coalition, Dougherty and colleagues also found that RC/AL residents on hospice were less likely to receive opioids for pain than the home hospice population.²⁷ Analyses of the 2016 Consumer Assessment of Healthcare Providers Survey (CAHPS) Hospice Survey also demonstrated that even among those receiving hospice services, bereaved family members report having received overall lower quality of end-of-life care in RC/AL compared to persons dying at home, with hospice care provision in RC/AL becoming increasingly similar to that provided in nursing homes.^28,29 A survey of 7,510 bereaved next of kin by one national hospice provider also revealed worse perceptions of hospice quality of care among decedents in nursing homes, when compared to private residences or RC/AL communities.³⁰ Remaining in home-like environments for as long as possible is increasingly promoted as part of respecting patients’ end-of-life care preferences, raising novel challenges about providing end-of-life care in residential care settings. The quality of hospice care provided in RC/AL requires additional research in light of these national trends.

This study has several strengths, including the use of a national cohort of Medicare decedents residing at validated RC/AL ZIP codes during the last month of life, providing one of the first national overview of hospice utilization patterns in this setting. We also rely on an extensive policy inventory that takes into account the complexity of RC/AL licensing across states.¹² However, this work has some limitations, including limited information about the smaller, home-like residential care settings. Research suggests that there may be even more variation in the ability of small RC/ALs to collaborate with hospice providers.³¹ We are also at times potentially capturing beneficiaries living at the same ZIP code but residing in independent living, although this is less of a concern when focusing on a decedent cohort with higher care needs, overall. We make the assumption that, in the absence of a change in ZIP code in the MBSF or evidence of transfer to other care settings in administrative claims data, beneficiaries have remained in RC/AL. We may have missed some transfers of dying residents from RC/AL communities to the private residence of family members, which we cannot clearly ascertain from hospice place of service. Finally, we can only adjust for demographic characteristics and comorbidities documented in administrative claims data.

Conclusions and Implications

In conclusion, RC/AL residents were significantly more likely to receive hospice care in place when residing in a licensed setting with supportive hospice regulations during the last month of life. In light of growing national trends in older adults remaining in residential care settings toward the end of life, pushing for regulatory change in states that do not yet explicitly allow hospice care could contribute to improve hospice utilization in this setting.

Appendix

Attachment 1: Questions from the regulation inventory used for hospice and staffing

Presence of these either or both of these supportive hospice regulations:

• "Does the state allow persons on hospice to be admitted?"

• "Does the state allow residents to use hospice services?"

Presence of any registered nurse (RN) staffing requirement:

• “Is general AL RN staff level specified to be on site as needed?”

• “Is general AL RN staff level specified to be on site less than 20 hrs/week?”

• “Is general AL RN staff level specified to be on site more than 20 hrs/week?”

• “Is dementia-specific RN staff level specified to be on site as needed?”

• “Is dementia-specific RN staff level specified to be on site less than 20 hrs/day?”

• “Is dementia-specific RN staff level specified to be on site more than 20 hrs/week?”