Guardianship and End-of-Life Care for Veterans with Dementia in Nursing Homes

Abstract

Background/Objectives:

Experts have suggested that patients represented by professional guardians receive higher-intensity end-of-life treatment than other patients, but there is little corresponding empirical data.

Design:

Retrospective cohort study.

Setting and Participants:

Among Veterans aged ≥65 who died from 2011–2013, we used Minimum Data Set assessments to identify those who were nursing home residents and had moderately severe or severe dementia. We applied methods developed in prior work to determine which of these Veterans had professional guardians. Decedent Veterans with professional guardians were matched to decedent Veterans without guardians in a 1:4 ratio, according to age, gender, race, dementia severity, and nursing facility type (VA-based versus non-VA).

Measurements:

Our primary outcome was intensive care unit (ICU) admission in the last 30 days of life. Secondary outcomes included mechanical ventilation and cardiopulmonary resuscitation in the last 30 days of life, feeding tube placement in the last 90 days of life, ≥3 nursing home-to-hospital transfers in the last 90 days of life, and in-hospital death.

Results:

ICU admission was more common among patients with professional guardians than matched controls (17.5% versus 13.7%), but the difference was not statistically significant (adjusted OR, 1.33 [95% CI, 0.89–1.99]). There were no significant differences in receipt of any other treatment, nor was there a consistent pattern. Mechanical ventilation and cardiopulmonary resuscitation were more common among patients with professional guardians, while feeding tube placement, ≥3 end-of-life hospitalizations, and in-hospital death were more common among matched controls.

Conclusion:

Rates of high-intensity treatment were similar whether or not a nursing home resident with dementia was represented by a professional guardian. This is in part because high-intensity treatment occurred more frequently than expected among patients without guardians.

Introduction

Health care providers frequently encounter patients with diminished capacity who have no suitable surrogate.¹ When the impairment is expected to be prolonged, the default mechanism for identifying a medical decision-maker is to ask the court to appoint a professional guardian — a paid official who has no prior relationship with the patient.^{2, 3}

Medical decision-making under guardianship is inherently difficult. Because the professional guardian is a stranger, he or she may have little knowledge of the goals that should guide decisions unless the patient has an advance directive. In addition, while most states have statutes directing the guardian to use a substituted judgment or best interests standard, few have specific language regarding the guardian’s authority to make decisions about life-sustaining treatment.³ Experts have suggested that professional guardians react to such uncertainty in two ways. One is by choosing high-intensity treatment, regardless of the likely benefits and harms.^{4, 5} The other is by deferring decision-making to the court, with high-intensity treatment occurring by default while the medical team awaits a judge’s ruling.⁶

There has been little empirical study of these issues, however. Despite longstanding calls for research on end-of-life decision-making under guardianship,⁷ a recent systematic review bemoaned the “dismal state of the science” in this area.⁸ The primary reason is that data about guardianship are not collected in population-based surveys or by most states and courts.⁹ As a result, it has been challenging to identify large samples of patients under guardianship and to study their end-of-life outcomes.

In prior work, our team overcame this issue by developing a method to identify a population-based sample of individuals with professional guardians using data from the Department of Veterans Affairs (VA).¹⁰ We examined the end-of-life care for a small number of patients at VA facilities in a single state.¹¹ Our objective in the present study was to use this method to assemble a larger, national cohort of decedent Veterans who were represented by professional guardians and had moderately severe or severe dementia and to compare their end-of-life care to the care received by decedent Veterans without guardians who had similar characteristics. Our hypothesis was that individuals who were represented by professional guardians would be more likely to receive high-intensity treatment near the end of life.

Methods

Study Population

We obtained data from both the VA and the Centers for Medicare and Medicaid Services on all Veterans age 65 and older who died between January 1, 2011 and December 31, 2013. We identified nursing home residents with moderately severe and severe dementia using data from the Minimum Data Set (MDS), a federally mandated instrument that contains detailed information about disease diagnoses, functional status, and cognition. MDS data are collected on admission and then quarterly for residents in all licensed U.S. nursing homes, including residents in VA facilities. Although prior studies have used data from the MDS to identify patients with different levels of dementia severity, there is no consensus approach, and investigators have utilized different combinations of measures that result in cohorts with narrower and broader ranges of function and cognition.^12–15 In selecting criteria, we sought to approximate Stage 6 on the Global Deterioration Scale (GDS),¹⁶ for moderately severe dementia, and Stage 7 on the GDS, for severe dementia. At these stages, disability becomes significant enough that dementia increasingly eclipses any other medical conditions.¹⁷

A nursing home resident was classified as having moderately severe dementia if the following criteria were met on an MDS assessment within 120 days of death: a diagnosis of Alzheimer’s disease or non-Alzheimer’s dementia; partial or complete dependence in at least one activity of daily living (ADL);¹⁸ and a Brief Interview for Mental Status (BIMS) score from 2–7 or a Cognitive Performance Scale (CPS) score of 4. A resident was classified as having severe dementia if, in addition to a diagnosis of dementia, he or she had partial or complete dependence in at least six ADLs, including eating, as well as a BIMS score ≤1 or CPS score ≥5. The BIMS is a cognitive screening tool with questions assessing temporal orientation and recall that is scored from 0–15, with lower scores corresponding to more severe impairment. It is administered as part of MDS Version 3.0, which has been used in nursing facilities outside the VA since October 1, 2010 and in VA nursing facilities since July 1, 2012. Scores from 2–7 are consistent with significant deficits across multiple domains, while scores ≤1 reflect profound cognitive impairment (for example, the participant is not able immediately to repeat more than 1 of 3 words).^{19, 20} The CPS is a validated measure that was calculated for all nursing home residents in MDS Version 2.0 and for residents who are unable to attempt the BIMS in MDS Version 3.0. It is scored from 0–6, with higher scores indicating more severe deficits: a score of 4 reflects moderately severe impairment in decision-making, communication, and short-term memory, and scores ≥5 indicate very severe impairment.²¹

Identification of Patients with Professional Guardians and Matched Controls

Our approach to identifying patients with professional guardians has been detailed previously.^{10, 11} Briefly, using the cohort of decedent Veterans with dementia described above, we queried the VA electronic medical record and VA administrative data with SQL Server 12.0 (Microsoft Corp., Redmond, WA). We searched for forms of the words “guardian” and “conservator” in standardized VA note titles and in structured fields in VA administrative data that contain information about patients and their next of kin. We then manually reviewed the charts of patients flagged through these searches and determined whether a guardian made medical decisions. We also ascertained from the chart whether the guardian was a professional or a family member, and excluded patients who had family guardians.

To identify a control group of Veterans without professional guardians, we used a greedy nearest neighbor matching algorithm.²² Each nursing home resident with a professional guardian was matched to four nursing home residents in the study population who did not have guardians. We excluded patients as potential matches if they were missing data on the covariates below. Patients were matched according to age (+/− 3 years), race (white versus non-white), gender, dementia severity (moderately severe versus severe), and type of nursing facility (VA-based versus non-VA based). The research plan was approved by the institutional review boards at Yale University and the VA Connecticut Healthcare System.

Outcomes

Our outcomes were high-intensity medical treatments, using measures included in a recent systematic review of studies evaluating the intensity of end-of-life care.²³ The primary outcome was admission to an intensive care unit (ICU) in the last 30 days of life. ICU admission was defined as the presence of a critical care revenue code, in the Medicare data, or a critical care bed section code, in the VA data. We included general, specialty, and coronary care units and excluded psychiatric or intermediate critical care. Secondary outcomes included mechanical ventilation in the last 30 days of life and cardiopulmonary resuscitation (CPR) in the last 30 days of life, ascertained using procedure codes from the International Classification of Diseases, Ninth Revision (ICD-9)^{14, 24} in the Medicare and VA data. We ascertained whether a feeding tube was inserted in the last 90 days of life using ICD-9 codes for inpatient placements and Current Procedural Terminology, Fourth Edition codes for outpatient placements.²⁵ Finally, we examined markers of potentially burdensome care transitions, including ≥3 nursing home-to-hospital transfers in the last 90 days of life as well as death in the hospital.²⁶

Covariates

Several covariates were included a priori as potential confounding factors. Because there is substantial geographical variation in end-of-life care spending and intensity for patients with dementia,²⁷ we obtained data from the Dartmouth Atlas of Health Care on area-level spending by hospital referral region (HRR) for Medicare beneficiaries in the last six months of life and, using data from 2011, divided the 306 HRRs into quintiles. We then used zip code information from the MDS assessment closest to death to assign each participant to an HRR quintile. To account for changes in care patterns over time, we also included year of death as a covariate.

Statistical Analysis

We summarized categorical variables with counts and percentages and continuous variables with means and standard deviations. We then examined differences between the two groups (professional guardian versus no guardian) using chi-square and student’s t-tests, as appropriate. Associations between professional guardianship and the primary and secondary outcomes described above were assessed using a logistic regression model that accounted for each matched set as a cluster. Odds ratios and 95% confidence intervals were derived using a robust variance estimator, first for professional guardianship alone and then adjusting for the covariates above. Analyses were performed using SAS, version 9.4 (SAS Institute, Cary, NC), and a two-sided p-value <0.05 was considered statistically significant.

Results

Among 790,876 Veterans aged 65 and older who died from 2011–2013, there were 73,361 in nursing homes who had moderately severe or severe dementia (Figure 1). Using text searches and chart review, we determined that 217 patients had professional guardians. We identified 868 matched controls without guardians.

Figure 1.

Identification of patients with professional guardians and matched controls. Abbreviations: VA = Department of Veterans Affairs.

The characteristics of patients with professional guardians and matched controls are shown in Table 1. Their mean age was similar (for patients with professional guardians, 82.9 years [SD, 7.3 years]; for matched controls, 82.4 years [SD, 7.3 years]). In both groups, 96.8% of patients were male, 83.9% were white, and 11.1% were residing in a VA-based skilled nursing facility. The largest percentage of patients in each group lived in the highest-spending HRR quintile.

Table.

Characteristics of patients with professional guardians and matched controls

Characteristic	Patients with Professional Guardians (n = 217)	Matched Controls (n = 868)	p-valuea
Age, mean (SD)	82.9 (7.3)	82.4 (7.3)	0.99
Male, n (%)	210 (96.8)	840 (96.8)	1.0
White, n (%)	182 (83.9)	728 (83.9)	1.0
VA-based facility, n (%)	24 (11.1)	96 (11.1)	1.0
Dementia severity, n (%)			1.0
Moderately severe	88 (40.6)	352 (40.6)
Severe	129 (59.4)	516 (56.4)
Year of death, n (%)			0.28
2011	79 (36.4)	283 (32.6)
2012	65 (30.0)	309 (35.6)
2013	73 (33.6)	276 (31.8)
HRR expenditure quintile, n (%)			0.22
1 (lowest)	25 (11.5)	101 (11.6)
2	30 (13.8)	136 (15.7)
3	63 (29.0)	212 (24.4)
4	28 (12.9)	162 (18.7)
5 (highest)	71 (32.7)	257 (29.6)

^aStudent’s t-test for continuous variables and chi-square test for categorical variables.

Abbreviations: HRR = hospital referral region

Rates of high-intensity end-of-life treatment are shown in Figure 2. There were no statistically significant differences between the two groups, and there was not a consistent trend in association. ICU transfer in the last 30 days of life was more common among patients with professional guardians than matched controls (17.5% versus 13.7%; adjusted OR [aOR], 1.33 [95% CI, 0.89–1.99]), as were mechanical ventilation (7.4% versus 6.1%; aOR, 1.21 [95% CI, 0.67–2.19]), and cardiopulmonary resuscitation (1.8% versus 1.0%; aOR, 1.77 [95% CI, 0.55–5.88]). Feeding tube placement in the last 90 days of life was less common among patients with professional guardians than among matched controls (6.9% versus 8.1%; aOR 0.82 [95% CI, 0.46–1.46]), as were repeated hospital transfers in the last 90 days of life (12.9% versus 15.9%; aOR 0.77 [95% CI, 0.50–1.20]) and in-hospital death (25.8% versus 29.3%; aOR 0.83 [95% CI, 0.59–1.17]).

Figure 2.

Association between professional guardianship and intensity of end-of-life care. Intensive care unit transfer, mechanical intubation, and cardiopulmonary resuscitation were measured in the last 30 days of life. Feeding tube placement and ≥3 hospitalizations were measured in the last 90 days of life. Models were adjusted for year of death and hospital referral region quintile.

Discussion

In this retrospective cohort study, we examined end-of-life care among nursing home residents with moderately severe or severe dementia and compared the treatment received by patients with professional guardians to the treatment received by patients did not have a guardian. We hypothesized that high-intensity treatment would be much more common for persons under guardianship, given case reports and expert opinion suggesting that these patients are particularly vulnerable to receiving potentially burdensome end-of-life care. Our findings did not confirm this hypothesis. Among the six high-intensity treatments that we examined, three occurred more often for patients with professional guardians and three occurred more often for patients in the matched control group. These differences were small and were not statistically significant.

These findings are consistent with a prior study in which we examined end-of-life treatment under guardianship by reviewing the charts of patients with professional guardians who received care at VA facilities in Connecticut.¹¹ This work is, to our knowledge, the only previous empirical investigation of this issue. In the prior study, we found that rates of both ICU utilization and hospice care among patients with professional guardians were similar to what has been reported for older adults overall.²⁶ The study did not include a contemporaneous control group, however, and was further limited by its small sample size, use of data from a single state, and inclusion of patients with a variety of reasons for incapacity and in multiple different residential settings. The current study extends these findings by including a control group and examining the end-of-life treatment received by a larger, national, and more homogeneous sample of nursing home residents with dementia and very substantial cognitive and functional impairment. It again fails to support reports that patients with professional guardians have a strikingly increased rate of high-intensity end-of-life care.

Patients under guardianship had slightly higher rates of several treatments, including ICU admission, which was our primary outcome measure. They had slightly lower rates of several others. We lacked sufficient power to detect differences of this magnitude because guardianship is relatively uncommon and because we based our sample size calculations on the assumption that effect sizes would need to be more than twice as large to be clinically relevant. The small differences between the groups reflect unexpected findings both for patients with and without professional guardians. The rates of high-intensity treatment among those under guardianship in this study stand in contrast to the prevailing view that professional guardians invariably choose high-intensity care. From this perspective, it is encouraging that more than 80% of nursing home residents with guardians were not transferred to the ICU near the end of life, and that 3 out of 4 did not die in the hospital. While high-intensity care was less common than we expected among patients with guardians, it was more common than we anticipated among the matched controls. Based on the most recent published data, we had expected approximately 10% of nursing home residents with moderately severe or severe dementia to be transferred to the ICU in the last month of life.²⁸ In our study, the rate of ICU transfer in the control group was 13.7% — more than a third higher.

What our findings suggest is that the experiences near the end of life for patients with dementia who are represented by professional guardians have more in common with the experiences of all patients with dementia than has been supposed. There are two potential reasons for this. First, a growing number of reports have called attention to the role of societal and health system-level factors on the care received by patients with dementia. These include beliefs among clinicians about the most appropriate default treatment options to offer,²⁹ the culture of practice in the institution at which the patient receives care,³⁰ racial and ethnic disparities,³¹ and structural issues, such as the availability of ICU beds.¹⁴ The legal system itself is one such factor and had led us to our hypothesis that the presence of a professional guardian would have a significant association with high-intensity medical treatment. It is possible that other system-level factors are more important drivers of high-intensity care and that such factors were evenly distributed between the groups, which were balanced through matching and by adjustment for regional differences in end-of-life health care spending. Second, it may be that the challenges faced by a professional guardian are not as different as we had thought from those encountered by any surrogate who must make decisions for a patient with significant cognitive and functional impairment. The usual instruction for a surrogate is to exercise substituted judgment – to make the decisions that the person would make if he or she were able to comprehend the situation.³² Although this presents particular difficulties for a professional guardian, who is a stranger, it is a strenuous intellectual maneuver even for a family member who has an intimate relationship with a loved one, since it involves understanding what the loved one would have thought about a state that is exceedingly difficult to imagine ahead of time and involves, by some accounts, a different person or self.^33–35 For this reason, there is a rich and ongoing debate about the need for dementia-specific advance directives^{36, 37} and about recommending that surrogates use alternative decision-making standards, such as making decisions according to the best interests or authentic values of the patient.^{38, 39} It may be that all surrogates — professional guardian or not — struggle equally to navigate these complex issues.

Our study has several limitations. We identified our sample from a population of Veterans, who are older than the general population and are overwhelmingly male.⁴⁰ We did not have access to information about patients’ causes of death, and, for this reason, limited our analysis to those in nursing homes who had moderately severe or severe dementia, with very substantial cognitive and functional impairment. Future studies on non-Veterans and persons with other life-limiting diseases are needed. In addition, the process we used to identify individuals under guardianship relies on information recorded in the medical chart. There is likely to be measurement error, with some patients misclassified or overlooked. Our approach remains the only available method for studying health care decision-making and outcomes in this population, however. Finally, because information about advance care planning and advance directives is not available in administrative data, we are not able to determine when high-intensity care was provided that was concordant with a patient’s wishes, or whether there were differences in advance care planning between the groups. This, too, should be the subject of future investigation.

In summary, among Veterans who were nursing home residents and had moderately severe or severe dementia, we did not find that those who were represented by professional guardians were more likely to receive high-intensity end-of-life treatment, in part because such care occurred more often than we expected in our control group. Future work should focus on addressing the challenges in end-of-life decision-making for all patients with dementia.