Potentially Burdensome End-of-Life Transitions among Nursing Home Residents with Poor Prognosis Cancer

Abstract

BACKGROUND:

We sought to examine factors associated with potentially burdensome end of life (EOL) transitions between care settings among older adults with advanced cancer in nursing homes (NHs).

METHODS:

We conducted a retrospective analysis of deceased older NH residents with poor prognosis solid tumors, using Medicare claims and the Minimum Data Set. We defined a potentially burdensome transition as two or more hospitalizations or an intensive care unit (ICU) admission in the last 90 days of life.

RESULTS:

Among 34,670 subjects, many had moderate-to-severe cognitive impairment (53.8%), full dependence in activities of daily living (66.5%), and comorbidities like congestive heart failure (CHF) (29.3%) and chronic obstructive pulmonary disease (COPD) (34.1%). Only 55.4% of patients used hospice at any time in the 90 days before death. 36.0% of patients experienced a potentially burdensome EOL transition, which were higher among patients who did not receive hospice (45.4% vs. 28.7%, p<0.01). In multivariable analyses, full dependence in ADLs (OR 1.70, P<0.01), CHF (OR 1.48, P<0.01), and COPD (OR 1.28, P<0.01) were associated with higher risk of burdensome EOL transition. Those with do-not-resuscitate directives (OR 0.60, P<0.01) and impaired cognition (OR 0.89, p<0.01) had lower odds of burdensome EOL transition.

CONCLUSIONS:

NH residents with advanced cancer have substantial comorbidities and functional impairment, yet more than a third experienced potentially burdensome EOL transitions. These findings help identify a population at risk for poor EOL outcomes in order to target interventions, and point to the importance of advanced care planning in this population.

Precis:

Nursing home residents with advanced cancer have substantial comorbidities, cognitive deficits, and functional impairment, yet more than a third experienced a potentially burdensome EOL transition. These findings help identify a population at risk for poor EOL outcomes in order to target interventions, and point to the importance of advanced care planning in moderating the risk of burdensome EOL transitions.

INTRODUCTION

Acute hospitalizations represent a large part of health care utilization and costs at the end of life (EOL) for patients with advanced cancer.^1,2 Multiple hospitalizations and intensive care unit (ICU) utilization toward the end of life is considered burdensome by many older adults, based on qualitative analyses involving families and other observational studies.^3-6 Nonetheless, community-dwelling patients with advanced cancer and a high symptom burden or functional decline are at high risk for these transitions, whereas receipt of hospice care decreases the likelihood of burdensome transitions.⁶

Nursing home (NH) residents with cancer are a growing and particularly vulnerable group of older adults.⁷ Though this population is not well studied, some studies point to multiple comorbidities, high symptom burden, and high rates of hospitalization among NH residents with cancer.^8-10 Advanced comorbidities and frailty increase the risks and complications of cancer treatment.¹¹ Therefore, we need a more comprehensive understanding of burdensome transitions of care at the EOL for this population, which could inform supportive care measures for NH residents as well as other frail older adults with cancer.

Geriatric assessment can detect vulnerabilities and predict outcomes in older adults with cancer beyond traditional oncology performance scales, and the assessment of geriatric domains is now recommended by leading cancer societies.^12-14 However, little is known about whether geriatric domains, such as cognitive and functional impairments, are associated with important clinical outcomes in NH residents with advanced cancer when determining cancer management near the end of life, even though all NH residents receive a standardized geriatric assessment every three months. Knowledge of how these domains impact care utilization is especially needed given that conditions such as advanced dementia and functional dependency – which are more prevalent among NH residents than in community-dwelling older adults¹⁵– may represent competing risks of death and may themselves lead to poor outcomes.¹⁶

In this study, we described demographic characteristics and geriatric domains, including comorbidities, cognition, and functional status, among NH dwelling older adults with advanced cancer in the last 90 days of life. We then described the frequency of potentially burdensome EOL transitions. Finally, we determined which demographic and clinical geriatric domains were associated with potentially burdensome EOL transitions.

Methods

Data Sources:

We used Medicare claims data, including Parts A and B claims, from January 2008 to December 2011 and linked them with the Minimum Data Set (MDS) version 2.0 files and the Online Survey, Certification and Reporting (OSCAR) NH dataset. The MDS is a comprehensive clinical assessment that is federally mandated for all US NH residents on admission and quarterly thereafter. We obtained Institutional Review Board approval for this study through Hebrew SeniorLife.

Study Design:

We conducted a retrospective analysis of deceased older adults (≥65 years old) with poor prognosis solid tumors who resided in US NHs and enrolled in Medicare fee-for-service (2008-2009). We used a previously described method for developing a cohort of long-stay NH residents.¹⁷ In brief, eligibility criteria included: residing in a NH over 100 days within the study eligibility period of 1/2008 – 12/2009 (i.e., long-stay resident) with no more than 10 consecutive days outside the facility during the 100 day eligibility period, continuous Medicare enrollment, and not on Medicare Advantage. Among this population, we studied those who had died by 12/2011 according to the Medicare enrollment file. We limited our sample to include residents who in the year before death, had any International Classification of Disease (ICD) claim for a poor prognosis solid tumor, defined per prior literature [Supplemental Table 1]. Prior studies of NH dwelling residents with cancer have relied on ICD-9 codes for cancer^8,18 or resident self-report,¹⁹ whereas specific ICD-9 codes for poor prognosis cancers may more accurately capture residents dying of advanced cancer.²⁰

Primary Outcome:

Our primary outcome was any potentially burdensome transition in the last 90 days of life. We defined a potentially burdensome transition as two or more hospitalizations or any intensive care unit (ICU) admission in the last 90 days of life, similar to studies on EOL health care utilization.³ We chose the 90-day window in order to increase the likelihood that a recent quarterly MDS assessment was available before death and since studies of EOL care in the general population use a similar window of time.^21,22 We were not able to determine whether EOL transitions were goal-concordant based on patient/family preferences or whether the hospitalization improved symptom management, and for that reason we classify them as potentially burdensome, meriting further study. We defined hospice utilization as any hospice claim in the last 90 days of life.²³ Hospice use could have occurred before or after potentially burdensome transitions. This approach was chosen given the variable timing of hospice use in the 90-day lookback window, with most hospice use occurring in the final weeks before death.

Covariates:

We collected resident age, race, education, and insurance status from the MDS. We took geriatric domain variables from the most recent MDS assessment before death. These measures include comorbidities, defined by provider-report on the MDS. We assessed cognition by the Cognitive Performance Scale (CPS)²⁴, which categorizes impairment as none, mild, moderate, severe, or very severe impairment. We dichotomized cognition as none or mild vs. moderate-to-severe impairment. We assessed functional status by a modified Activities of Daily Living (ADL) scale, categorizing the number of severe ADL impairments in bed mobility, dressing, eating, toileting, and personal hygiene, ranging from 0-5.²⁵ The MDS documents do-not-resuscitate (DNR) advanced directives. Chemotherapy administration in the last 90 days of life was determined, per prior literature, using ICD-9 diagnostic code V58.1 or procedure code 9,925 for inpatient chemotherapy, and using Current Procedural Terminology codes (96400, 96408, 96410, 96412, 96414, and 96545), Common Procedure Coding System (Q0083-Q0085, J7150, or J8999-J9999), and revenue center codes (0331, 0332, and 0335) for outpatient chemotherapy.²⁶

We described facility characteristics as measured by the OSCAR system in the year of long-stay qualification, including percentage of residents with black race, moderate to severe cognitive impairment, and DNR directives, and linked this with patient-level data. We also report facility all-cause hospitalization rate per resident-years. Missing data for all co-variates was <1%, except for education status which had missing data for 1.7% of residents.

Statistical Analysis:

We analyzed data using SAS software, version 9.4 (SAS Institute Inc.). We calculated descriptive statistics using means with standard deviations (SDs) for continuous variables and proportions for categorical variables. We compared the proportion of potentially burdensome EOL transitions among residents using hospice versus not using hospice, using Chi-squared tests. We stratified our multivariable analysis by hospice, given its known association with burdensome EOL transitions. We used hierarchical logistic regression modeling with outcome of any burdensome transition in the last 90 days of life, controlling for both resident and facility variables. Models included age, sex, race/ethnicity, education, comorbidities, cognitive status, number of ADLs in which residents were dependent, and receipt of chemotherapy in the last 90 days of life. These covariates were defined a priori based on the prior literature linking demographic and functional characteristics with poor health outcomes in older adults with cancer.²⁷

Further, in order to address differences in facility practices, we adjusted for facility-level characteristics that are known to be associated with health outcomes, including the percentage of facility residents who are black, have moderate-to-severe cognitive impairment, or have a DNR directive. We also adjusted for facility-level hospitalization rate, since variation in hospitalization from NHs is a well-described phenomenon.²⁸ Finally, we described the primary discharge diagnosis of the hospitalization closest to death.

Results

Resident & Facility Characteristics:

We included 34,670 NH residents with poor-prognosis solid tumors. The average age was 82.7 (SD 7.9), 62.7% were female, and 74.7% had a high school education or below (Table 1). Many residents had comorbid CHF (29.3%), COPD (34.1%), and diabetes (35.0%). Over half (53.8%) had moderate-to-severe cognitive impairment. 66.5% were dependent in all ADLs. Only 55.4% of residents used hospice at any time in the 90 days before death, and 61.8% had a DNR directive. Further, 2.1% of residents received chemotherapy in the last 90 days of life. Facility characteristics are described in Table 1. Socio-demographic variables and geriatric domains were similar between hospice and non-hospice users.

Table 1:

Patient and Facility Characteristics, by Hospice Status

	All Patients	Hospice in Last 90 Days of Life	No Hospice in Last 90 Days of Life
Sociodemographic Measures	N = 34,670	N = 19,508	N = 15,162
Age at Death (Mean, SD)	82.7 (7.9)	82.6 (7.8)	82.9 (7.9)
Female Sex	21,723 (62.7%)	12,394 (63.5%)	9,329 (61.5%)
Race/Ethnicity
White Race	28,968 (83.6%)	16,580 (85.0%)	12,388 (81.7%)
Black Race	4,267 (12.3%)	2,265 (11.6%)	2,002 (13.2%)
Hispanic Ethnicity	549 (1.6%)	281 (1.4%)	268 (1.8%)
Other Race/Ethnicity or Missing	886 (2.6%)	382 (2.0%)	504 (3.3%)
Medicaid Insurance Status (Yes/No)	12,158 (35.2%)	6,747 (34.7%)	5,411 (35.8%)
Education: High School or Less	25,889 (74.7%)	14,452 (74.1%)	11,437 (75.4%)
Comorbidities
Congestive Heart Failure	10,142 (29.3%)	5,249 (26.9%)	4,893 (32.3%)
Chronic Obstructive Pulmonary Disease	11,813 (34.1%)	6,256 (32.1%)	5,557 (36.7%)
Diabetes Mellitus	12,143 (35.0%)	6,592 (33.8%)	5,551 (36.6%)
Hip Fracture	1,921 (5.5%)	1,088 (5.6%)	833 (5.5%)
Functional and Clinical Measures
Cognitive Performance Scale
Moderate-Severe Impairment	18,638 (53.8%)	10,565 (54.2%)	8,073 (53.2%)
Number of Independent ADLs
Zero (Completely Dependent)	23,049 (66.5%)	12,930 (66.3%)	10,119 (66.7%)
One	7,969 (23.0%)	4,490 (23.0%)	3,479 (22.9%)
Two	1,662 (4.8%)	960 (4.9%)	702 (4.6%)
Three	698 (2.0%)	380 (1.9%)	318 (2.1%)
Four	518 (1.5%)	297 (1.5%)	221 (1.5%)
Five (Completely Independent)	774 (2.2%)	451 (2.3%)	323 (2.1%)
Chemotherapy, last 90 days	742 (2.1%)	385 (2.0%)	357 (2.4%)
Do Not Resuscitate Order	21,418 (61.8%)	12,700 (65.1%)	8,718 (57.5%)
Any Hospice Use in last 90 days	19,508 (56.3%)
Facility Characteristics
Patients with Black Race	12.3% (18.8)	11.8% (18.3)	12.8% (19.5)
Patients with moderate cognitive impairment or worse	44.3% (14.8)	44.5% (15)	44.0% (14.6)
Patients with DNR order	53% (22.6)	53.5% (22)	52.7% (23.3)
Hospitalizations per patient-years [Number, (Standard Deviation)]	1.5 (1)	1.5 (0.7)	1.5 (1.3)

ADL = Activities of Daily Living; SD = Standard Deviation; DNR = Do-Not-Resuscitate

Note: Missing data across all variables <1%, except for education where missing data 1.7%

Potentially Burdensome Transitions:

Overall, 36.0% of residents experienced one or more potentially burdensome EOL transition, and potentially burdensome transitions were more common for residents who never received hospice (45.4% vs. 28.7%, p<0.01, Figure 1). Specifically, 24.6% of residents had two or more hospitalizations in the last 90 days of life, which was higher among those not using hospice (31.6% vs. 19.1%, p<0.01). 25.5% of residents had an ICU stay in the last 90 days of life, which was also higher among those not using hospice (33.2% vs. 19.6%, p<0.01). Potentially burdensome transitions varied by region, with the South having the highest proportion of burdensome transitions (38.2%) and the Mid-West with the lowest proportion (33.7%).

Figure 1:

Potentially Burdensome End of Life Transitions, by Hospice Status

ICU = Intensive Care Unit

Potentially Burdensome EOL Transitions among Residents not using Hospice:

Among 15,162 residents without hospice in the last 90 days of life, several factors were associated with higher odds of potentially burdensome EOL transition (Table 2). Black race was associated with higher odds of potentially burdensome EOL transition (OR 1.21, 95%CI 1.07-1.38, p<0.003). Residents with comorbidities had higher odds of potentially burdensome EOL transition, including those with CHF, COPD, and diabetes. Compared to those with no impairments in ADLs, those who were impaired in 5 ADLs had significantly higher odds of potentially burdensome EOL transitions (OR 1.70, 95%CI 1.32-2.18, p<0.001). Finally, residents receiving chemotherapy in the last 90 days of life had increased odds of potentially burdensome EOL transition (OR 1.47, 95%CI 1.17 to 1.85, p=0.001).

Table 2:

Factors Associated with Potentially Burdensome Transition in the last 90 days of life among those not using hospice

N = 15,162	Odds Ratio
Age at Death (Reference: 65-74)
75-84	0.89 (0.81-0.98) *
85+	0.72 (0.65-0.80) **
Sex (Reference: Male)
Female	1.07 (1.00-1.15)
Race/Ethnicity (Reference: White)
Black	1.21 (1.07-1.38) **
Hispanic	1.20 (0.92-1.58)
Other	1.34 (1.12-1.61) **
Medicaid Insurance Status	0.92 (0.86-0.99) *
Congestive Heart Failure	1.48 (1.37-1.59) **
Emphysema / COPD	1.28 (1.19-1.38) **
Diabetes Mellitus	1.26 (1.17-1.35) **
Hip Fracture	1.10 (0.95-1.28)
Cognitive Performance Scale (Ref: Mild Impairment or Better)
Moderate Impairment or Worse	0.89 (0.83-0.96) **
# of Independent ADLs
Zero (Completely Dependent)	1.70 (1.32-2.18) **
One	1.40 (1.09-1.81) **
Two	1.15 (0.86-1.54)
Three	1.00 (0.71-1.42)
Four	1.21 (0.83-1.76)
Five (Completely Independent)	Ref
Chemotherapy, last 90 days	1.47 (1.17-1.85) **
Do Not Resuscitate Order	0.60 (0.56-0.65) **
Facility-Level Variables
5% increase in % of black residents	1.00 (0.98-1.01)
5% increase in % with moderate cognitive impairment or worse	0.97 (0.96-0.99) **
5% increase in % with a DNR order	0.97 (0.96-0.98) **
Hospitalizations per patient-years	1.43 (1.34-1.52) **

^*p<0.05;

^**p<0.01

Several factors were associated with lower odds of potentially burdensome EOL transitions, including Medicaid insurance status (OR 0.92, 95%CI 0.86 to 0.99, p=0.03) and age (OR 0.72, 95%CI 0.65 to 0.80, p<0.001, for residents 85+ years, compared to those aged 65 to 74 years). Compared to residents with mild or no impairment in cognition, those with moderate- to-severe impairment had lower odds of burdensome EOL transition (OR 0.89, 95%CI 0.83 to 0.96, p=0.003). Finally, residents with a DNR directive had significantly decreased odds of potentially burdensome EOL transition (OR 0.60, 95%CI 0.56 to 0.65, p<0.001).

At the facility level, those in facilities with more moderate-to-severe cognitively impaired residents had fewer potentially burdensome EOL transitions (OR per 5% increase in cognitively impaired residents: 0.97, 95%CI 0.96 to 0.99, p<0.001), as did facilities with a greater percentage of residents with DNR directives (OR per 5% increase: 0.97, 95%CI 0.96 to 0.98, p<0.001). There was no association between the percentage of black residents and burdensome EOL transitions. Residents residing in facilities with higher all-cause hospitalization rates (measured in hospitalizations per patient-years) had increased odds of potentially burdensome EOL transition (OR 1.43, 95%CI 1.34 to 1.52, p<0.001).

Potentially Burdensome EOL Transitions in Hospice Population:

Among 19,508 residents who used hospice in the last 90 days of life, median days spent in hospice was 22, with 13.6% of hospice users spending three or fewer days on hospice and 28% spending seven or fewer days on hospice. In multivariable logistic regression models with the outcome of any potentially burdensome EOL transition, similar associations were found as in the non-hospice population except that female residents in this population had higher odds of potentially burdensome EOL transition (OR 1.13, 95%CI 1.05 to 1.21, p<0.001), and there was no association between cognitive impairment and potentially burdensome EOL transition.

Diagnoses of last hospitalization before death:

Among the 23,061 residents who experienced a hospitalization in the last 90 days of life, we identified the primary discharge diagnosis code of the admission closest to the end of life. The most common reasons for admission included renal failure or dehydration (17.3%) followed by pneumonia (9.8%), urinary tract infections (4.9%), sepsis (3.4%), and aspiration (3.0%). Other admission diagnoses each comprised less than 3.0% of all hospitalizations.

Discussion:

In a large cohort of long-stay NH residents with advanced cancer, we found that more than a third of residents experienced a potentially burdensome EOL transition. More than a third had comorbidities and moderate-to-severe cognitive impairment, and over two-thirds were severely functionally impaired. Common causes for hospitalization included kidney failure, dehydration, and infection. Importantly, only about half of residents ever used hospice in the last 90 days of life, and most utilized hospice only in the final weeks of life. These findings describe a vulnerable older adult population with poor-prognosis cancer and the potential to improve these patients’ EOL care.

To our knowledge, we are the first to describe potentially burdensome EOL transitions for NH-dwelling older adults with poor prognosis cancers. Our observation that a third of NH residents with poor-prognosis cancer experience a burdensome EOL transition is higher than reports for other NH populations. Using a related definition of burdensome EOL transition, Gozalo et al. found that 19% of NH residents with dementia experience a burdensome EOL transition, compared to 36% in our study.³ The incidence of ICU-level care in our study (26%) is more than twice the incidence in a study of NH residents with severe cognitive and functional impairment,²⁹ although the time before death in the prior study was shorter (30 versus 90 days). Such high rates of acute care use in NH residents with cancer near death call into question the utility of hospital transfers and intensive treatments, especially since many of these hospitalizations may be avoidable.³⁰ Our data suggests that admissions for the top diagnoses, including dehydration and common infections, may be avoidable with proper advanced care planning.⁴² Moreover, these interventions may be inconsistent with the goals of care of residents and their families, as has been demonstrated in other NH populations³¹ as well as in populations with advanced cancer.^32,33

Multiple reasons could explain the high rates of potentially burdensome EOL transitions in our study . At the resident level, we found that burdensome EOL transitions were more common in non-whites and those with multiple comorbidities. Racial disparities in EOL care are well described.^34-36,37. While the association between functional impairment and health care utilization is a well-described phenomenon in community-dwellers with and without cancer ^38,39 and in NH residents with curable cancers,⁴⁰ we found that coexisting functional impairment and advanced cancer was associated with increased odds of burdensome EOL transitions. NH residents with functional impairment and advanced cancer may have complex symptoms and care needs that are challenging to manage. In contrast, we found no association between impaired cognition and potentially burdensome EOL transitions. This likely reflects the focused clinical approach to dementia care that is more central to NHs, as well as improvements in palliative care for this population over time.^3,41

Our finding that DNR directives were associated with fewer potentially burdensome transitions raises the possibility that increased goals of care conversations about likely health trajectories in advanced cancer could reduce burdensome transitions. Fewer residents utilizing hospice experienced a potentially burdensome EOL transition, yet hospice was infrequently used despite the fact that all residents in this study were hospice-eligible. Additionally, our work suggests regional variation in potentially burdensome EOL transitions, with the highest prevalence in the South. Further work is needed to explore and overcome patient-level and facility-level barriers to advance care planning and hospice access for this population.

Our study’s findings that potentially modifiable facility-level variables are associated with increased burdensome EOL transitions points to the role of facility- and health system-level interventions in improving EOL care for this population. In particular, the finding that the facility-level hospitalization rate is associated with more burdensome transitions calls for a greater focus on avoiding preventable hospitalizations from NHs,²⁸ including interventions for patients with advanced cancer. Future studies should explore regional or health-system factors affecting care of NH residents with cancer as well as explore the role of nurse staffing ratios and NH-based palliative care and symptom-management interventions on EOL outcomes. Policy and financing incentives to reduce potentially burdensome EOL transitions and facilitate better palliative care in NHs could also be beneficial to this population.

Finally, our findings provide data to help educate oncology teams and clinicians about NH care, and has important implications for the integration of oncology care into the NH setting. Coordinating care between NH providers, front-line staff, and oncologists is challenging given financial, transportation, and communication barriers. These patients are likely too ill to travel to oncology clinic and too ill for chemotherapy, and so it likely falls on NH clinicians to have discussions about transitions to hospice and symptom management. Further integration of oncology and palliative care into the NH setting, as well as education on the unique needs of this population, have the potential to improve care coordination, better manage symptoms, and facilitate transitions to hospice care when disease-directed therapy is considered no longer advisable.

Our study has limitations. We could not differentiate how many burdensome transitions were due to cancer versus underlying comorbidities. We were unable to assess the exact stage or timing of cancer diagnosis, the extent to which cancer contributed to death, or the extent to which patients were receiving active oncology care, including oncologist visits, surgery, or radiation. However, the claims-based definition that we used is more likely to reflect active, poor-prognosis malignancy, compared to methods used in prior studies. It is possible that we underreported the number of residents with advanced cancer since many NH-dwelling older adults may elect not to pursue cancer diagnostics upon discovery of a potentially serious malignancy and do not typically undergo cancer screening. The data available to us was from 2008-2011, and while hospice use may have increased more recently, the associations we describe are likely still relevant given that our similar findings in a hospice and non-hospice population. Finally, most hospice enrollment occurred in the final weeks of life, and so we did not have a sufficient time window to determine if potentially burdensome transitions were reduced and EOL care improved after receipt of hospice. Future efforts to study EOL care should prospectively capture goal concordance and symptom burden in order to capture domains relevant to older adults with cancer and inform geriatric oncology practice.

Conclusion:

In summary, our study demonstrates that over a third of NH residents with advanced cancer experience a potentially burdensome EOL transition. These NH residents have significant comorbidities and functional impairment, which increase the likelihood of burdensome transitions. As the population with advanced cancer and comorbidities ages we can expect that many will require NH level of care. It is therefore essential to dedicate efforts to preventing burdensome transitions in this unique population through better palliative care and facility-based initiatives. Further study of the association between improved communication between oncologists and NH providers about EOL goals and values as well as policy initiatives that discourage avoidable transitions, are needed to improve care for this vulnerable, growing population.