Hospice Improves Care Quality For Older Adults With Dementia In Their Last Month Of Life

Abstract

The Medicare hospice benefit was originally designed around a cancer disease paradigm but increasingly serves people living with dementia. At this time, almost half of all older adults receiving hospice care have dementia. Yet there is minimal evidence as to whether hospice benefits people living with dementia outside of nursing facilities. We asked whether and how perceived quality of last-month-of-life care differed between people living with dementia and persons without dementia and whether hospice use among people living with dementia was associated with higher perceived quality of care compared with the quality of care for people living with dementia who did not use hospice. We used nationally representative data from the National Health and Aging Trends Study and Medicare claims from 2011–17 to examine the impact of hospice enrollment on proxy perceptions of last-month-of-life care quality. Proxies of people living with dementia enrolled in hospice compared with proxies of those not enrolled more often reported care to be excellent (predicted probability: 52 percent versus 41 percent), more often reported having anxiety and sadness managed (67 percent versus 46 percent), and less often reported transitions in the last three days of life (10 percent versus 25 percent). There were no differences in the impact of hospice on proxy ratings of care for people living with dementia compared with ratings of care for people without dementia. Policy makers should consider these benefits when weighing changes to hospice policy and regulations that may affect people living with dementia.

Alzheimer’s disease and other dementia syndromes (hereafter “dementia”) affect one in three older adults ages eighty-five and older.¹ Dementia is one of the top six causes of death,² with the proportion of people with dementia dying in home and community-based settings increasing rapidly from 14 percent in 1999 to 28 percent in 2019.^3,4

Hospice is the dominant model of care for seriously ill individuals at the end of their life, serving 50.7 percent of Medicare decedents.⁵ Hospices support people with serious illness and their families through interprofessional care that encompasses symptom management and psychosocial, spiritual, and caregiver support with the goal of enabling individuals to remain at home at the end of life. People living with dementia now make up one of the largest growing segments of the hospice enrollees and account for almost half of all hospice enrollments.^6,7

Despite the fact that the majority of people living with dementia live and die in the community,^3,4 evidence around dementia end-of-life experiences and impacts of hospice primarily come from nursing home and hospital settings. Studies have shown that among people living with dementia in nursing facilities, use of hospice (compared with nonuse) is associated with better caregiver ratings of care and fewer unmet needs.^8–12 In nursing home comparisons of hospice enrollees with any diagnosis with nonenrollees with any diagnosis, hospice is associated with fewer hospitalizations in the last thirty days of life and better assessment and treatment of pain.^13–15 The minimal information available regarding home- and community-dwelling people living with dementia in the United States indicates that they have similar medical acuity and symptom burden as people living with dementia in nursing homes.^16–21 Moreover, they are often cared for by family caregivers who lack formal training and who experience significantly more strain caring for people living with dementia who are in the last year of their life.^22,23 The potential of hospice to improve the end-of-life experience for people living with dementia and their families is enormous,²⁴ yet we know relatively little about how hospice affects end-of-life outcomes for people living with dementia across all settings. Understanding whether hospice care benefits people living with dementia is critical as hospice undergoes ongoing reform that might affect people with dementia.^25–27

We leveraged the National Health and Aging Trends Study to examine the quality of end-of-life care at the population level. We sought to investigate whether and how perceived quality of last-month-of-life care differed between people living with dementia and persons without dementia and whether hospice use among people living with dementia was associated with higher perceived quality of care compared with perceived quality of care for people living with dementia who did not use hospice. We examined hospice enrollment for at least one day in the last month of life, as well as hospice enrollment for thirty or more days. We used the National Health and Aging Trends Study linked to Medicare claims to examine outcomes including overall quality of care; symptom management; support for personal and spiritual needs; provision of goal-consistent, respectful, patient- and family-involved coordinated care; and transitions in the last three days of life.

Study Data And Methods

Data Sources

Data are from the National Health and Aging Trends Study, an ongoing annual longitudinal population-based survey of Medicare enrollees linked to Medicare and hospice claims.²⁸ Assessments of last-month-of-life care for decedents are completed by proxies; weighted response rates ranged from 95 percent in 2012 to 97 percent in 2017.²⁹ The National Health and Aging Trends Study is sponsored by the National Institute on Aging (NIA U01AG32947) and is conducted by Johns Hopkins University.²⁸

Our sample included 2,805 National Health and Aging Trends Study participants who died between 2011 and 2017 (the last available year of Centers for Medicare and Medicaid Services claims data) and who were ages seventy and older, per National Health and Aging Trends Study guidance for maintaining national representativeness when pooling data across waves.³⁰ Of these, we excluded 657 decedents who did not have a sample person interview within two years before death and eighty-nine who did not have a last-month-of-life survey or end-of-life outcomes. Online appendix exhibit A1 provides a flow chart with additional details of sample derivation.³¹ Our final study sample was 2,059 decedents.

Measures

We identified participants’ dementia status in the National Health and Aging Trends Study round before death, based on the algorithm validated for use in surveys³²: report of physician diagnosis, proxy responses to the AD8 dementia screening interview,³³ or cognitive testing. We drew sociodemographic, health, and functional characteristics from the last National Health and Aging Trends Study interview before death (mean: 10.5 months before death; standard deviation: 10.0 months): age, gender, race and ethnicity, educational attainment, partnership status, income, number of children, living alone, census region, metropolitan area type, health conditions (self-report of whether a doctor told them they had cancer, heart disease, lung disease, stroke, or diabetes), and receiving help with or not doing activities of daily living (ADLs: bathing, eating, dressing, toileting, transferring, walking inside). Missingness was less than 5 percent and was not imputed.

We examined setting based on National Health and Aging Trends Study residence categorizations in the round before death, at hospice admission, and at death. The National Health and Aging Trends Study uses interviewer observations and questions of the sample person or facility staff person to categorize residences as nursing homes, community-based residential care settings, or private homes. This aligns with prior work in other large data sets.³⁴ Community-based residential settings include group homes, board and care homes, assisted living, or continuing care retirement communities.

We drew age at death and characteristics of hospice use from the Medicare claims data. We measured hospice use in two ways: at least one day in the last month of life and for thirty continuous days before death. Appendix exhibit A2 details the definition of hospice use.³¹

Our outcome measures for proxy ratings of last-month-of-life care quality included eleven ratings of care and one clinical indicator of care quality. To build off prior literature,^35–37 we dichotomized variables with Likert scales to compare the most positive rating with any less-positive rating for adjusted analyses. Appendix A3 provides details on outcome measure definitions and dichotmization.³¹ We added a commonly used clinical indicator of care quality: late transitions (those in the last three days of life).³⁸ Mean time to completion of proxy interview after death was about six months (mean: 188.5 days; SD: 131.6 days), and 71 percent of proxies were spouses or children.

Analyses

We described the characteristics of participants overall, comparing people living with dementia with participants without dementia, and examined differences in proxy ratings of care between people living with dementia and those without dementia, using chi-square tests to determine differences.

We then examined whether having received hospice in the last month of life affected ratings of care quality. We estimated covariate balancing propensity scores³⁵ for each decedent’s likelihood of hospice enrollment, conditional on variables known to affect hospice use: age; sex; race and ethnicity; education; marital status; number of children; income; presence of cancer, heart disease, lung disease, stroke, or diabetes; metropolitan area; census region; and year of death. We used the product of the inverse probability of treatment weights³⁹ and the National Health and Aging Trends Study survey weights in all analytic comparisons. Appendix A4 and exhibits A5–7 provide details on covariate balancing propensity score specification and standardized differences.³¹ We tested the interaction between hospice use and dementia status to assess whether the impact of hospice was greater for people living with dementia compared with the impact for those without dementia. We repeated the above analyses for thirty continuous days of hospice before death compared with no hospice.

Analyses were conducted using SAS, version 9.4, and Stata, version 17.0. All analyses account for the National Health and Aging Trends Study survey weights.

Limitations

This study has some limitations. It examines whether participants had dementia, but not whether dementia was the cause of death. It is likely that many persons with dementia died of other conditions, and we cannot assess how their experience was shaped by the presence of dementia versus other serious illnesses. Also, we were unable to adjust for unmeasured characteristics of older adults who do and do not enroll in hospice, such as preferences for end-of-life care or market-based variation in the availability of hospice care. Although imperfect, the inverse probability of treatment weight is among the most rigorous methodologic tools available outside of a randomized trial and has been used in a wide range of studies to investigate policy-relevant questions using observational survey data. Finally, this study does not represent the national population of nursing home residents ages seventy and older, as the National Health and Aging Trends Study does not capture information about demographics and dementia status for those living in nursing homes at the time of the first National Health and Aging Trends Study interview.

Study Results

Sample Characteristics

Among 2,059 decedents ages seventy and older, participants had a mean age at death of 84.6 (SD: 7.1), 54 percent of participants were women, the majority identified as White non-Hispanic, and a third had less than high school education attainment and annual incomes in the bottom tercile (exhibit 1). At the last interview before death, 80 percent of participants lived in home settings (data not shown); 52 percent of those who used hospice in the last month of life began receiving it at home. For a table that includes additional data, see appendix exhibit A8.³¹ Overall, 36 percent died at home and 32 percent in the hospital.

Exhibit 1:

Characteristics of decedent study sample among Medicare beneficiaries ages 70 and older by dementia status, 2011–17

	Overall (N = 2,059)	Without dementia (N = 1,108)	With dementiaa (N = 951)
Demographics b
Age at death (median, SD), yearsc	84.5 (7.1)	83.0 (6.3)	86.8 (7.5)
Women (versus men), %	54	52	56
Black, %	9	8	11
Hispanic or Latino/a/e/x, %	6	5	8
White, %	80	82	76
Less than high school education versus high school or more, %	31	28	37
Single, divorced, widowed (versus married or partnered), %	61	59	64
Annual income, %
$0–14,600	29	25	35
$14,601–30,000	35	34	37
$30,001–1,000,000	36	40	29
Family and residence b , %
4+ children (versus fewer or none)	32	32	32
Metropolitan area type (versus nonmetro)	81	82	80
Northeast census region	21	22	18
Midwest census region	23	25	20
South census region	36	34	39
West census region	21	19	23
Health conditions b , %
Cancer	19	20	16
Heart disease	40	39	41
Lung disease	28	30	25
Stroke	10	6	15
Diabetes	31	32	29
Functional impairment b,d , %
Bathing	42	22	72
Eating	26	10	50
Dressing	43	24	72
Toileting	23	8	45
Transferring	27	11	52
Walking inside	35	16	64

SOURCE Authors’ analysis of the data from participants in the National Health and Aging Trends Study ages 70 and older who had a sample person interview and who died between 2011 and 2017. Weighted percentages (or medians) are shown. For a table that includes additional data, see online appendix exhibit A8 (see note 31 in text).

^aDementia status was determined at cohort entry based on the National Health and Aging Trends Study algorithm using: report of diagnosis, proxy responses to the AD8 dementia screening, or cognitive testing.

^bAssessed at cohort entry, which was a mean 10.5 (standard deviation [SD]: 10.0) months before death.

^cAssessed at time of death via National Health and Aging Trends Study last-month-of-life survey or via claims data.

^dImpairments indicates the percentage of participants who received help or did not do this activity of daily living.

Comparing the 40 percent (n = 951) of people living with dementia within the study population with the 60 percent (n = 1,108) with no dementia, people living with dementia had greater prevalence of social and functional vulnerabilities: older mean age at death (87 versus 83 years; p < 0.001), lower annual incomes (35 percent versus 25 percent had incomes of $0–$15,000; p < 0.001), higher prevalence of stroke (15 percent versus 6 percent; p < 0.001), and significantly higher levels of all measures of ADL impairment. A higher proportion of people living with dementia used hospice (59 percent versus 43 percent; p < 0.001) and had a longer median length of stay (median, fifteen days for people living with dementia for the enrollment ending in death versus ten days for those with no dementia). Among people living with dementia, the most common location of death was at home (37 percent), followed by 23 percent at hospital and 24 percent in nursing homes; in comparison, 39 percent of those with no dementia died in the hospital and 35 percent died at home, with only 13 percent dying in a nursing home (p < 0.001) (appendix exhibit A8).

Last-Month-Of-Care Quality Is Similar For People Living With Dementia And People Without Dementia

In unadjusted ratings of last-month-of-life care in the overall study sample, 48 percent of proxies reported excellent care quality and more than three-quarters of proxies rated personal care always managed, care goal-consistent, and care coordinated; other measures left greater room for improvement (exhibit 2). For a table that includes additional data, see appendix exhibit A9.³¹ In unadjusted comparative analyses, meaningful differences in proxy ratings of last-month-of-life care included higher prevalence of personal care always managed, family kept informed, and care coordination among people living with dementia compared with care coordination among those without dementia. People living with dementia also had fewer late transitions. Only two differences were observed in adjusted analyses: people living with dementia (compared with those with no dementia) had anxiety more often managed, and they had less goal-consistent care (not shown).

Exhibit 2:

Outcomes of last-month-of-life care quality, overall and comparing people living with dementia and those without dementia among Medicare beneficiaries ages 70 and older by dementia status, 2011–17

Measure	Overall	Without dementia	With dementia	P-valuea
Quality of care (no.)	1,987b	1,051	936	0.832
Excellent (%)	48.3	48.5	48.0
Very good to poor (%)	51.7	51.5	52.0
Pain (no.)	1,857	1,012	845	0.0527
No pain (%)	30.3	30.8	29.4
Pain not managed (%)	16.0	17.8	13.2
Pain managed (%)	53.8	51.4	57.3
Trouble breathing (no.)	1,901	1,033	868	0.054
No trouble breathing	44.0	42.0	46.8
Breathing not managed (%)	11.1	12.6	8.9
Breathing managed (%)	44.9	45.3	44.3
Anxiety or sadness (no.)	1,782	988	794	0.0903
No anxiety (%)	41.7	43.2	39.5
Anxiety not managed (%)	27.1	27.7	26.2
Anxiety managed (%)	31.1	29.1	34.3
Personal care managed (no.)	1,973	1,036	937	0.0173
Always (%)	81.0	79.0	83.7
Usually to never (%)	19.0	21.0	16.3
Enough spiritual support (no.)	1,543	806	737	0.714
No (%)	57	56.5	57.6
Yes (%)	43	43.5	42.4
Goal-consistent care (no.)	1,936	1,026	910	0.2786
No (%)	11.7	11.0	12.8
Yes (%)	88.3	89.1	87.3
Treated with respect (no.)	1,972	1,033	939	0.717
Always (%)	87.2	87.3	87.0
Usually to never (%)	12.8	12.7	13.0
Patient or family input (no.)	1,934	1,024	910	0.4052
No (%)	9.3	9.8	8.5
Yes (%)	90.7	90.2	91.5
Family informed (no.)	1,973	1,040	933	0.032
Always (%)	80.6	78.9	83.1
Usually to never (%)	19.4	21.2	16.9
Care coordinated (no.)	1,879	986	893	<0.001
No (%)	13.4	16.9	8.6
Yes (%)	86.6	83.1	91.4
Late transitionc (no.)	1,957	1,056	901	0.0197
No (%)	81.5	79.7	84.2
Yes (%)	18.5	20.3	15.8

SOURCE Authors’ analysis of the data from participants in the National Health and Aging Trends Study ages 70 and older who had a sample person interview and who died between 2011 and 2017. Outcome measures are presented as weighted numbers and percents and are drawn from the last-month-of-life survey were completed by a proxy (71 percent of respondents were spouses or children); mean time to completion after decedent death was approximately 6 months (mean: 188.5 days; standard deviation: 131.6 days). For a table that includes additional data, see online appendix exhibit A9 (see note 31 in text).

^aP values are the result of unadjusted bivariate analyses including chi-square tests for categorical variables to determine statistically significant differences between those with and without dementia.

^bAs response rates varied by outcome, the first row of each outcome includes the N for each population.

^cLate transitions were defined as care setting changes in the last 3 days of life.

Hospice Improves Last-Month-Of-Life Care For People Living With Dementia

Among people living with dementia, 59 percent (n = 555) used hospice for at least one day in the last month of life (appendix exhibit A8). Hospice enrollment was associated with better proxy-reported perceptions of last-month-of-life care on some domains compared with the reported perceptions of people living with dementia with no hospice (exhibit 3).

Exhibit 3:

Comparing associations of last-month-of-life care quality outcomes among Medicare beneficiaries ages 70 and older, 2011–17, by use of hospice for at least 1 day in the last month of life

Source/Notes: SOURCE Authors’ analysis of the data from participants in the National Health and Aging Trends Study ages 70 and older who had a sample person interview and who died between 2011 and 2017. NOTES Analyses used covariate balancing propensity score methods to estimate the propensity for enrolling in hospice at least 1 day in the last month of life, estimated with variables known to affect hospice use: age; sex; race and ethnicity; education; marital status; number of children; income; comorbidities of cancer, heart disease, stroke, or diabetes; metropolitan area; census region; and year of death. From the propensity score model, we estimated normalized inverse probability of treatment weights. For each outcome measure we used postestimation commands to present the adjusted predicted probabilities. All analyses account for National Health and Aging Trends Study survey weights. Late transitions were defined as care setting changes in the last 3 days of life.

Findings from the adjusted propensity score models indicate that proxies of hospice-enrolled people living with dementia more often rated care as excellent compared with people living with dementia who were not enrolled in hospice (predicted probability: 52.0 percent versus 41.4 percent; p = 0.012) (exhibit 3). A higher percentage of proxies of people living with dementia in hospice versus not in hospice rated sadness and anxiety managed (predicted probability: 66.9 percent versus 45.7 percent; p = 0.009) and rated enough spiritual support (predicted probability: 51.1 percent versus 30 percent; p < 0.001). A smaller percentage of people living with dementia in hospice had late transitions compared with people living with dementia who were not in hospice (predicted probability: 9.8 percent versus 24.6 percent; p < 0.001) (exhibit 3).

We observed no differences in the impact of hospice on people living with dementia compared with its impact on people without dementia (interaction terms not shown). In sensitivity analyses, we found no major differences in outcome measures by proxy type or by proxy self-reported familiarity with the daily routine of the sample person.

Longer Hospice May Improve Last-Month-Of-Life Care Even More For People Living With Dementia

Finally, we observed a potential dose response of hospice care. Even though all confidence intervals overlapped, the effect of people living with dementia receiving thirty continuous days of hospice in the last month of life compared with receiving no hospice tended to be higher than the effect of people living with dementia receiving hospice of one or more days in the last month of life compared with receiving no hospice (exhibit 4).

Exhibit 4:

Comparing associations of last-month-of-life care quality outcomes among Medicare beneficiaries ages 70 and older, 2011–17, by use of hospice for 30 continuous days in the last month of life (versus no hospice) and with hospice use of at least 1 day in the last month of life (versus no hospice)

Source/Notes: SOURCE Authors’ analysis of the data from participants in the National Health and Aging Trends Study ages 70 and older who had a sample person interview and who died between 2011 and 2017. NOTES Analyses used covariate balancing propensity score methods to estimate the propensity for enrolling in hospice 30 continuous days in the last month of life and propensity for enrolling at least 1 day in the last month of life, estimated with variables known to affect hospice use: age; sex; race and ethnicity; education; marital status; number of children; income; comorbidities of cancer, heart disease, stroke, or diabetes; metropolitan area; census region; and year of death. From the propensity score model, we estimated normalized inverse probability of treatment weights. All analyses account for National Health and Aging Trends Study survey weights. We present the odds ratio of quality outcomes for use of hospice for 30 continuous days in the last month of life (versus no hospice) and at least 1 day in the last month of life (versus no hospice). Late transitions were defined as care setting changes in the last 3 days of life.

Discussion

Our novel study uses data from a national sample of Medicare beneficiaries to address the questions of whether hospice benefits people living with dementia, and whether it affects people living with dementia differently than it affects those without dementia. The vast majority of our study sample lived in the community a year before death, and more than half enrolled with hospice, including almost two-thirds of those with dementia. We found that proxies of hospice-enrolled people living with dementia were more likely to rate last-month-of life care as excellent compared with proxies of people living with dementia who did not enroll with hospice. We observed no differences in the impact of hospice on ratings of last-month-of-life care for persons with versus without dementia. These findings suggest that hospice benefits people living with dementia just as well as it benefits people without dementia.

Disease-specific differences in time lines for disease progression, need for help, and caregiving needs may have implications for measuring care quality. Although the last-month-of-life care quality measures are the best available, there is need for improvement.^40,41 These measures do not tell the full picture of the experience of end of life for people with dementia and their care communities. The duration of dementia is long and unpredictable and is often not recognized as terminal.⁴² The last month of life for people living with dementia may look similar to the last three or even six months of life because people living with dementia often die from conditions that deteriorate in the last forty-eight hours, such as infection. The responsibilities and burden of caregiving for people living with dementia⁴³ occurs over a much longer period of time and with a different intensity than, say, caregiving for persons dying of cancer, and may differentially affect proxy appreciation for the involvement of hospice. A year before death, people living with dementia have higher symptom burden and functional impairments than those without dementia,²¹ and people living with dementia at home need more support for functional and medical needs than those in facility settings.²⁰ Future work should examine the impact of care settings on proxy perceptions of last month of life among people living with dementia. The Consumer Assessment of Healthcare Providers and Systems Hospice Survey⁴⁴ and its predecessor, the Family Evaluation of Hospice Care,⁴⁵ indicate that proxies rate home-based hospice care more highly than hospice care in nursing facilities.

Although we found that care quality for hospice-enrolled people living with dementia was as good as it was for hospice-enrolled individuals without dementia, there is room for improvement: only half of proxies rated overall quality at the highest level. This finding, coupled with the high prevalence of dementia among hospice enrollees,^6,46 underscores the need for dementia care to be a core competency of hospice teams.⁴⁷ Other potential recommendations include adding more support for caregivers, including anticipatory guidance^48,49; providing additional home health aides for functional needs; and removing medications from standard hospice medication order sets that could exacerbate behavioral symptoms of dementia (for example, psychotropics, anticholinergics, and aggressive use of opioids that are renally metabolized).^50–52

Our findings inform an ongoing debate about whether hospice in the United States adequately serves people living with dementia and their families, given that the Medicare hospice benefit evolved to primarily serve end-stage cancer.⁵³ Prior research indicates that hospice employees perceive demand for hospice care among people living with dementia to be increasing at the same time that changes to hospice policy and regulations make it more difficult to meet this need.²⁵ Concerns about lack of access to care for people living with dementia come from two sources: difficulties people have in being admitted to hospice and the experiences of hospice enrollees who are disenrolled alive.

First, eligibility for the Medicare hospice benefit requires that physicians attest to a six-month prognosis if the disease runs its expected course, but our ability to identify a six-month prognosis for most dementia syndromes is no better than a coin flip.⁵⁴ This means some people living with dementia may not receive the full benefit of hospice care because they are enrolled too close to death or not at all.

Second, people with dementia are known to have long duration of hospice use, which has been found to be profitable to hospice organizations; for-profit hospices are more likely to care for those with dementia.^46,55,56 Given concerns about long duration of hospice use and the influx of for-profit hospices and private equity funding,^27,57 Medicare has responded by implementing policies to curb long hospice enrollment.²⁶ People living with dementia are disproportionately discharged alive from hospice^46,56,58 because their decline has stabilized or because hospices are concerned that they will not be reimbursed for care if they cannot adequately document decline. Being disenrolled for extended prognosis or being expelled from hospice negatively affects both seriously ill people and families because there is no simple replacement for hospice care.^59–63

Our finding that hospice significantly benefits those with dementia and their families underscores the need to ensure access to high-quality end-of-life care for this growing population. Future policies and regulations should focus on hospice agencies with unusually high rates of long stays and live discharge rather than current population-based approaches that may have unintended consequences of broadly reducing access to care for people living with dementia. Demonstration studies can test how removing prognostic requirements from hospice eligibility for people living with dementia affects timely access, care quality, and costs. In addition, we need robust, accessible, and affordable systems of home and community-based care tailored to the needs of people living with dementia and their care communities, which facilitate the transition to hospice.

Conclusion

In this national sample of Medicare beneficiaries, hospice-enrolled people living with dementia had higher-quality last-month-of-life care compared with people living with dementia who were not enrolled in hospice. Hospice-enrolled people living with dementia had similar ratings of last-month-of-life care as hospice enrollees with no dementia. Given the benefits of hospice to people living with dementia, efforts to enable access are warranted.

Biographies

BIOS for 2021–01985 (Harrison)

Bio1: Krista L. Harrison (krista.harrison@ucsf.edu), University of California San Francisco.

Bio2: Irena Cenzer, University of California San Francisco.

Bio3: Claire K. Ankuda, Icahn School of Medicine at Mount Sinai, New York City.

Bio4: Lauren J. Hunt, University of California San Francisco.

Bio5: Melissa D. Aldridge, Icahn School of Medicine at Mount Sinai; and the James J. Peters Bronx Veterans Affairs Medical Center, Bronx, New York.