Advance Directives And Nursing Home Stays Associated With Less Aggressive End-Of-Life Care For Severe Dementia Patients

Abstract

The number of older adults with cognitive impairment is increasing, and such adults often require a surrogate to make decisions about health care. However, little is known about the aggressiveness of end-of-life care for these people, especially those who reside in the community. We found that cognitive impairment is common among older adults approaching the end of life, whether they live in the community or a nursing home, and that nearly 30 percent of patients with severe dementia remained in the community until death. Among those patients, having an advance directive in the form of a living will was associated with significantly less aggressive care at the end of life, compared to similar patients without such a directive—as measured by Medicare spending ($11,461 less per patient), likelihood of in-hospital death (17.9 percentage points lower) and use of the intensive care unit (9.4 percentage points lower). In contrast, advance directives were not associated with differences in care for people with normal cognition or mild dementia, whether they resided in the community or in a nursing home. Timely advance care planning after a diagnosis of cognitive impairment may be particularly important for older adults who reside in the community.

More than one-third of Americans ages seventy-one and older have dementia or cognitive impairment without dementia (CIND).^1,2 Dementia, usually defined as memory and other cognitive deficits that are sufficient to cause limitations in daily functioning, is often underrecognized as a terminal, progressive condition.^3,4 This may result in end-of-life care for people with dementia that can be burdensome, painful, costly, and unlikely to improve their quality of life or chance of survival.^3–8

Little research has been done on end-of-life care for patients with dementia living in the community. However, many studies have raised concerns about the provision of potentially overaggressive care for patients with dementia in nursing homes.^3,9–11 In addition, there are questions about whether and how advance directives shape the care of patients with dementia.^12–14

Advance directives might play a particularly important role in the care of patients with cognitive impairment—especially those with severe dementia—because accurately identifying and respecting patients’ preferences for care is challenging in light of their difficulty in fully comprehending complex information.^15–17 Often when a patient with dementia is dying, decisions must be made about caring for his or her infections, nutrition, and emergency care.¹⁸ Medical records for community-dwelling patients with dementia may not be available to guide such decisions when a patient urgently needs care.

Many programs across the nation aim to address this gap by using standardized forms such as the Physician Orders for Life-Sustaining Treatment (POLST). The potential impact of these types of policy interventions on patients in the community as opposed to those in a nursing home is largely unknown.

We investigated the interactions of nursing home stays, dementia, and the use of advance directives with the cost and aggressiveness of end-of-life care. We examined prospectively collected data on a nationally representative cohort of older Americans who underwent cognitive assessments and were linked to Medicare claims and postmortem interviews that were completed by proxy informants.

We compared the aggressiveness of care provided to patients with severe dementia, CIND or mild dementia, or normal cognition with and without advance directives across the nursing home and community settings. We examined total Medicare spending in the six months before death, a summary measure of use. We also examined the following three outcomes as measures of the agressiveness of end-of-life care: in-hospital death and the use of the intensive care unit and life-sustaining treatment.^6,19

Our findings suggest that engaging the patient and family members in advanced care planning when cognitive impairment is first diagnosed and documenting the patient’s preferences in writing can lessen the burden that overly aggressive care places on patients and the Medicare program.

Study Data And Methods

Study Population

We linked survey data from the nationally representative Health and Retirement Study (HRS) to Medicare claims for respondents who died in the period 1998–2007 at age sixty-five or older²⁰ (additional information about our methods, sample, and results can be found in the online Appendix).²¹ Our sample of 3,876 included survey respondents who had consented to the Medicare claims linkage (5,363 of 6,336 patients, or 84.6 percent), had fee-for-service Medicare coverage during the last six months of life (4,445 of 5,363 patients, or 82.9 percent), and had a postmortem interview completed by a proxy in 2000 or later (4,068 of 4,445 patients, or 91.5 percent), during which added questions about advance directives were asked. An additional 192 respondents who reported having coverage through the Veterans Affairs health system were excluded.

Dementia Categorization

We used HRS cognition measures from the respondent’s last survey interview.²² Details of the instruments and cutoff points we used are presented in the online Appendix²¹ and have been described elsewhere. In brief, the scales for patients’ and proxies’ responses indicating a level of impairment were based on “gold standard” CIND and dementia diagnoses validated in a previous study.^23,24

We characterized patients in our sample as having severe dementia if they were categorized as having dementia based on their HRS survey and also had limitations in at least three of the six activities of daily living (ADL)—walking, dressing, bathing, eating, getting into and out of bed, and toileting. Patients whose last HRS survey indicated CIND or dementia but had fewer than three ADL limitations were combined for this analysis in the category CIND/mild dementia.

Agrressiveness Of End-Of-Life Treatment

We used Medicare claims data to identify previously validated measures of aggressive and potentially burdensome end-of-life care.^6,19 We calculated total Medicare spending in the last six months of life in 2007 US dollars. We identified the following using Medicare Provider Analysis and Review (MedPAR) inpatient and skilled nursing files: receipt of life-sustaining treatments (intubation and mechanical ventilation, tracheostomy, gastrostomy feeding tube placement, hemodialysis, and enteral and parenteral nutrition), intensive care unit use, and in-hospital death.^6,19

Advance Directives And Nursing Home Use

HRS postmortem interviews with proxy respondents indicated whether the patient had had a written advance directive in the form of a living will that expressed a desire to limit care in certain situations—a frequently used measure of advance directive status.^14,25,26 We compared to patients with this form of written advance directive to all other patients because other research had found no independent effect of simply using a durable power of attorney without an additional living will.¹⁴

To determine nursing home status, we used Medicare skilled nursing claims data and proxy reports of whether the patient was living in a nursing home at the time of death or during the two years before death (or the two years before the last interview).

Statistical Analysis

We used general linear models with a log link and gamma distribution to model Medicare spending and a logit link and binomial distribution for dichotomous treatment measures to test for differences in use of health care in the last six months of life among patients with normal cognition, CIND/mild dementia, and severe dementia.^21,27 Since patterns of care differ across sites of care, we stratified our results by nursing home use near the end of life.

In a second set of models, we interacted cognition with advance directive status to allow the effect of advance directives to vary for patients who likely had different needs for surrogate decision making. Regressions controlled for patients’ sociodemographic characteristics as measured in the survey wave prior to death and for Elixhauser comorbidities that we calculated from hospitalizations occurring six to eighteen months before death.²⁸ We also controlled for characteristics of the geographic regions where patients were treated, including nursing home capacity²⁹ and end-of-life spending levels. To ensure the robustness of our results, we performed a number of analyses, which are described in the online Appendix.²¹

Limitations

Our research design required the use of Medicare claims. Consequently, we focused on decedents ages sixty-five and older, and we report only Medicare spending, not spending by other payers such as Medicaid or patients’ out-of-pocket spending. Medicaid claims are not yet available for HRS respondents, although Medicaid may cover a significant portion of nursing home care.

We classified cognition based on survey measures with cutoff points that had been validated against a full clinical assessment from the Aging, Demographics, and Memory Study.²³ This strategy has been shown to correctly classify 79 percent of HRS respondents compared to clinical exams, so there is a potential for misclassification.²⁴

Our dementia classification was conservative, however, because we used measures from interviews before the end-of-life period, when significant further decline occurs. Posthumous proxy reports indicated cognitive and physical decline between the last HRS interview and the last six months of life. At the last HRS interview, 34.2 percent of normal cognition patients, 44.9 percent of those with CIND, and 78.9 percent of those with dementia had trouble with one or more ADL. These rates increased to 57.5 percent, 69.1 percent, and 87.7 percent, respectively, in the last three months of life. Physical declines were likely accompanied by declines in cognitive functioning.

We do not know whether respondents were able to make complex decisions at the time decisions about their treatment in the last six months of life were made. However, our normal cognition and CIND/mild dementia groups likely included patients with impaired capacity to make complex decisions. This would increase the likelihood that our results would show no difference across groups. Thus, our results probably understate the differences in end-of-life utilization among severe dementia patients compared to those with normal cognition.

Study Results

Our sample was 3,876 Medicare beneficiaries aged 65 and above (see Appendix Exhibit 2 for a full description of the sample by levels of cognitive functioning).²¹ Cognitive impairment was common among these patients at the time of their last HRS interview: 43.1 percent had CIND/mild dementia, and 21.7 percent met the criteria for severe dementia (Exhibit 1 and Appendix Exhibit 2)²¹ at a mean of 436 days before death.

EXHIBIT 1. Cognitive Functioning In 3,876 Health And Retirement Study Respondents With Linked Medicare Claims At Last Interview Before Dying In The Period 1998–2007.

NOTES Authors’ analysis of Health and Retirement Study Core and Exit interviews linked to Medicare claims. On average, last interview took place 436 days before death. Patients in the cognitive impairment without dementia (CIND)/mild dementia category (n = 1,670) either had cognitive functioning consistent with mild cognitive impairment regardless of their physical functioning or had dementia symptoms and limitations in fewer than three activities of daily living (ADL). Patients in the severe dementia category (n = 839) had cognitive functioning consistent with dementia and limitations in three or more ADLs. Community-dwelling patients had no self-report of nursing home use or Medicare claim for skilled nursing home use in the last six months of life.

Relative to patients with higher levels of cognitive functioning, patients with severe dementia were older at death (86.6 years versus 83.9 years for patients with CIND/mild dementia and 79.3 years for those with normal cognition), more likely to be nonwhite (21.6 percent versus 19.8 percent for patients with CIND/mild dementia and 8.6 percent for those with normal cognition), and less educated (56.7 percent had less than a high school education, compared to 51.6 percent for patients with CIND/mild dementia and 27.4 percent for those with normal cognition) (Appendix Exhibit 2).²¹

The majority of patients with severe dementia received nursing home care near the end of life. However, 28.7 percent (241 of 839 patients) remained in the community (Appendix Exhibit 2),²¹ as did 51.4 percent (858 of 1,670) of patients with CIND/mild dementia. Only 36.4 percent of patients with severe dementia had treatment-limiting advance directives, and more of these patients living in nursing homes had such documents, compared to patients remaining in the community (40.0 percent versus 27.4 percent; Appendix Exhibit 2).²¹

Severe Dementia And End-Of-Life Care

Compared to patients with severe dementia living in the community, those who had a stay in a nursing home in the last six months of life had lower rates of use of the intensive care unit (ICU) (10.3 percent versus 17.1 percent; 95% confidence interval: −11.6 percent, −1.6 percent) and of in-hospital death (18.6 percent versus 26.5 percent; 95% CI: −13.6 percent, −2.0 percent) (Exhibit 2). However, there was no difference in Medicare spending across settings of care after we adjusted for patient characteristics.

Exhibit 2.

Medicare Utilization In The Last Six Months Of Life Among Community-Based And Nursing Home Patients With Normal Cognition, Cognitive Impairment Without Dementia (CIND)/Mild Dementia, And Severe Dementia Who Died In The Period 1998–2007

	Patients in:
	Community (n = 2,064)	Nursing home (n = 1,812)	Difference (community - nursing home)
Per capita Medicare spending ($1,000s)a
Normal cognition	26.2	40.1	−13.9****
CIND/mild dementia	22.9	35.0	−12.0****
Severe dementia	28.0	25.7	2.2
In-hospital death
Normal cognition	35.2%	23.8%	11.3****
CIND/mild dementia	30.5	22.5	7.9****
Severe dementia	26.5	18.6	7.8***
Life-sustaining treatment
Normal cognition	17.1	15.9	1.2
CIND/mild dementia	15.8	14.3	1.5
Severe dementia	16.2	12.4	3.8
ICU use
Normal cognition	25.2	20.4	4.7**
CIND/mild dementia	20.2	17.3	2.8
Severe dementia	17.1	10.3	6.8***

SOURCE [See Exhibit 1].

NOTES Reported values are predicted spending (probabilities) calculated using average marginal effects from general linear model regressions of Medicare use in the last six months of life on cognitive functioning and patient characteristics. Patients in the cognitive impairment without dementia (CIND)/mild dementia category (n = 1,670) either had cognitive functioning consistent with mild cognitive impairment regardless of their physical functioning or had dementia symptoms and limitations in fewer than three activities of daily living (ADL). Patients in the severe dementia category (n = 839) had cognitive functioning consistent with dementia and limitations in three or more ADLs. ICU is intensive care unit.

^a2007 dollars.

^**p < 0.05

^***p < 0.01,

^****p < 0.001

Compared to patients with CIND/mild dementia living in the community, those who had a stay in a nursing home had higher Medicare spending ($34,966 versus $22,943; 95% CI: $9,163, $14,885) and were less likely to die in the hospital (22.5 percent versus 30.5 percent; 95% CI: −11.9%, −3.9%) (Exhibit 2). Patients with CIND/mild dementia experienced similar rates of ICU use and life-sustaining treatment regardless of their nursing home use.

Severe Dementia And Advanced Directives

Whether or not a patient with normal cognition or CIND/mild dementia had an advance directive had no statistically significant association with Medicare spending or with use of the ICU or life-sustaining treatments in the last six months of life (Appendix Exhibit 3).²¹ Written advance directives were associated with lower Medicare spending and lower rates of in-hospital death, life-sustaining treatments and ICU use for severe dementia patients. For these patients, there was an important difference in the relationship between having an advance directive and end-of-life care for patients in the community, in contrast to nursing home patients.

Compared to patients with normal cognition, those with severe dementia had lower Medicare and ICU use and were less likely to die in the hospital when they had a stay in a nursing home—whether or not they had treatment-limiting advance directives.³⁰ Compared to patients in the nursing home with normal cognition, those with CIND/mild dementia had lower rates of in-hospital death and use of the ICU and life-sustaining treatment.

Based on Medicare spending, community-dwelling patients with no advance directives and severe dementia received care similar to those with normal cognition ($32,169 versus $26,957) (Exhibit 3). However, community-dwelling patients who had advance directives and severe dementia received markedly less aggressive care than those with normal cognition ($20,708 versus $29,777). Community-dwelling patients with severe dementia and written advance directives had $11,461 less Medicare spending in the last six months of life, compared to severe dementia patients who remained in the community but did not have written advance directives.

Exhibit 3.

Advance Directives (ADs) And Medicare Utilization In The Last Six Months Of Life Among Patients With Normal Cognition, Cognitive Impairment Without Dementia (CIND)/Mild Dementia, And Severe Dementia Patients Who Died In The Period 1998–2007

	Patients in:
	Community (n = 2,064)			Nursing home (n = 1,812)
	Normal cognition	CIND/mild dementia	Severe dementia	Normal cognition	CIND/mild dementia	Severe dementia
Per capita Medicare spending ($1,000s)a
No AD	27.0	24.3	32.2	39.9	34.7	24.7
AD	30.0	25.6	20.7	35.8	30.6	22.5
Differenceb	2.8	1.3	−11.5***	−3.9	−4.1**	−2.2
In-hospital death
No AD	35.8%	32.7%	31.8%	27.2%	23.8%	20.6%
AD	35.2%	27.5%	13.9%	17.9%	19.3%	14.6%
Difference (percentage points)b	0.54	−5.2	−17.9****	−9.3***	−4.5	−6.0**
Life-sustaining treatment
No AD	18.8%	17.2%	19.8%	16.5%	14.2%	11.6%
AD	18.2%	17.3%	10.6%	11.6%	10.6%	9.8%
Difference (percentage points)b	0.65	0.18	−9.2	−4.9	−3.6	−1.8
ICU use
No AD	26.2%	21.9%	19.6%	19.9%	16.8%	10.8%
AD	28.2%	20.5%	10.2%	17.9%	14.7%	7.1%
Difference (percentage points)b	2.0	−1.4	−9.4***	−2.0	−2.2	−3.7

SOURCE [See Exhibit 1].

NOTES AD indicates having a written document indicating a preference for limited treatment in some circumstances. Predicted spending and probabilities of use in the last six months of life are based on average marginal effects from general linear model and logistic regressions adjusting for patient characteristics, including cognitive functioning; whether the patient had a written advance directive specifying limits in treatment; and their interaction, stratified by nursing home use in the last six months of life.

^a2007 dollars.

^bAD - no AD.

^**p < 0.05

^***p < 0.01,

^****p < 0.001

We found similar patterns for in-hospital death (patients with severe dementia and advance directives were 17.9 percentage points less likely to die in the hospital than those without directives), ICU stays (−9.4 percentage points) and life-sustaining treatment (−9.2 percentage points) (Exhibit 3). In contrast, the treatment of patients with CIND/mild dementia resembled that of patients with normal cognition.

Discussion

Using nationally representative data, we found that cognitive impairment was very common among older adults residing either in nursing homes or in the community near the end of life: 43.1 percent of our sample population had CIND/mild dementia, and 21.7 percent had severe dementia (Exhibit 1). To our knowledge, this is the first study to estimate the prevalence of cognitive impairment and dementia at the end of life and to examine the associated use and costs of health care for patients living in the community. As was the case with previous studies of nursing home residents, we found higher rates of in-hospital death and use of life-sustaining treatment and the ICU among elderly patients with severe dementia or CIND/mild dementia than might be expected, given their prognosis; our study extends these results to show that aggressive care is common whether patients reside in nursing homes or the community.^{3,6,7,9,31–35}

In addition, we found that more than half of older adults remaining in the community in the last year of life were cognitively impaired: 41.6 percent had CIND/mild dementia, and 11.7 had severe dementia (Exhibit 1). Compared to similar patients who lived in nursing homes, the cognitively impaired patients in the community received much more aggressive care. This may have been because their cognitive impairment was not known to emergency health care providers, their preferences for care were less apparent to providers, or both.

The lack of available data has restricted most previous studies of patients with dementia to those in the nursing home. Our findings highlight the need for additional research to determine whether this aggressive care is driven by patients’ preferences or other factors.

Evidence for the overall effectiveness of advance directives is mixed.^14,17,36 Nonetheless, we hypothesized that written advance directives could be particularly important in guiding the treatment of patients with severe dementia, who probably lack the capacity to participate in medical decision making.

We found that being in a nursing home was associated with less aggressive end-of-life care, whether or not the patient had an advance directive. This result is consistent with findings in other studies of no differences among patients with and without advance directives.^17,36

However, we found a strong association between having an advance directive and receiving less aggressive treatment among community-dwelling patients with severe dementia. More than half of these patients experienced at least one hospitalization in the last six months of life in which surrogate decision making might have been needed, but only 27.4 percent had a written advance directive. Community-dwelling patients with severe dementia who had no advance directive received end-of-life care that was as aggressive as the care provided to patients with normal cognitive functioning.

Advance directives were unrelated to total spending, use of life-sustaining treatment and the ICU, and numbers of hospitalizations among people with normal cognition and CIND/mild dementia. That migh be because these patients remained able to participate to a substantial degree in decision making.

Our study was retrospective, and we were unable to establish a causal relationship between having a written advance directive and end-of-life outcomes. However, if the presence of a directive was merely an indication of preferences for less-aggressive care, we should have found differences in use across all levels of cognitive functioning.

Additional research is needed to determine whether having an advance directive causes differences in end-of-life care for older adults with severe dementia. But given the ethical challenges associated with clinical trials in this population, observational information is also important.

As people’s circumstances change over time and acute situations arise, patient preferences are known to change.^13,37 Strict adherence to written advance directives could fail to take these changes into account in the case of patients who are capable of participating in decisions about treatment, such as those with CIND/mild dementia.

We did not have detailed information about the individual decisions made for each patient. However, our findings suggest that having a written advance directive was related to moderation in the level of care for patients with severe dementia, but not for those with less impaired cognitive functioning. This result may be reassuring in light of concerns that the presence of an advance directive could cause physicians or surrogate decision makers to withhold treatments that patients at earlier stages of disease might prefer to receive.^12,13,38,39

Conclusion

Cognitive impairment is remarkably common among older adults near the end of life and highlights the value of advance care planning, especially for patients who reside in the community. Both written advance directives and nursing home residence are associated with lower rates of aggressive treatments that are often expensive but that are unlikely to extend the lives of patients with dementia and may decrease the quality of life. Efforts to engage patients and family members in advance care planning when cognitive impairment is first diagnosed and patients are still living in the community might help ensure that a surrogate decision maker is identified and informed about the patient’s preferences while these conversations are still possible.⁴⁰

We found that patients with CIND/mild dementia appeared to receive care that was more similar to care for people with normal cognition than for those with severe dementia. It is not clear from our results whether the similarity of care was because the preferences of patients in the two groups were similar or because cognitive impairment was not recognized as a comorbidity associated with high mortality. For example, physicians responding to the acute cause of hospitalization for patients in the community might be unable to distinguish long-standing cognitive impairment or dementia from acute changes in cognition, such as delirium, related to illness or the hospitalization itself. As a result, decisions about treatment might not fully take into account the terminal nature of dementia.

Interventions to improve clinicians’ knowledge of a patient’s diagnosis and preferences might help reduce the provision of potentially burdensome treatment to patients with severe dementia near the end of life. Such interventions include greater use of advance care planning in the community and of medic alert bracelets or other identifying diagnostic information, the implementation of Physician or Medical Orders for Life-Sustaining Treatment (POLST or MOLST), and the integration of living wills and information about treatment preferences into electronic health records.^41,42

Rates of advance care planning are increasing.⁴³ However, many patients are naming surrogate decision makers (using a durable power of attorney for health care) without giving them additional written instructions. Merely naming a proxy decision maker has not been found to be associated with less aggressive care in a national sample of Medicare beneficiaries.¹⁴ Thus, efforts to educate both patients and their families at the initial diagnosis about dementia and its progression may be necessary to prompt surrogates to learn about a patient’s preferences and prognosis in a timely manner.