Factors That Impact Family Perception of Goal-Concordant Care at the End of Life

Abstract

Background:

Goal-concordant care (GCC)—care aligned with a patient's known goals and values—is a measure of the quality of end-of-life (EOL) care that can be assessed by surveying family members after a patient's death. It is unknown whether patient characteristics affect this measure.

Objective:

The objective of the article was to examine family report of GCC and its associations with patient characteristics.

Methods:

Using the Health and Retirement Study, which is a nationally representative, longitudinal cohort of adults over age 50, we sampled decedents whose family completed the 2014 postdeath interview. Families reported frequency of GCC at the EOL. A multivariable regression model assessed the associations between family report of GCC and decedent characteristics.

Results:

Of 1175 respondents, 76% reported that the decedent “usually” or “always” received GCC. Proxy report of GCC was independently associated with age (adjusted odds ratio [AOR] 1.02, 95% confidence interval [CI] 1.01–1.03), having three or more chronic medical conditions (AOR 1.34, CI 1.02–1.77), the presence of written or verbal advance care planning (ACP) (AOR 1.38, CI 1.02–1.88), and an interaction term of race and ability to participate in EOL decision making (AOR 3.83, CI 1.02–14.40). African American race was not independently associated with GCC (AOR 0.73, CI 0.5–1.06).

Conclusion:

Family's report of GCC is associated with ACP, age, and multimorbidity. Being African American and perceived as able to participate in EOL decision making was significantly associated with report of GCC. Bringing the patient's voice into EOL care discussions through upstream ACP with likely surrogates may be particularly important to improving GCC for African Americans.

Background

Goal-concordant care (GCC) is defined as care aligned with a patient's known goals and values, and has been proposed as a marker of quality care at the end of life (EOL).¹ Measuring GCC at the end of life, however, poses several challenges.^1–3 Foremost, first-person accounts of EOL care are nearly impossible to obtain, and therefore many previous studies have relied on measuring concordance between advance care planning (ACP) documentation or patient surveys and EOL care received as a proxy for GCC. This approach cannot account for changes in patient preference over time or ACP that remains undocumented—such as verbal discussions with health care providers and family members.

An alternative approach considers the perspective of bereaved family members.^1,4–8 While also imperfect, measuring GCC using after-death surveys of family members or caregivers may help account for possible changes in goals and preferences over time, as well as account for less formal or undocumented ACP discussions. Two recent studies have used bereaved family surveys as a measure of GCC during the last month of life. These two studies focused primarily on the care received in the immediate period before death with a focus on race, care received, and location before death. To the authors' knowledge, no study to date has examined a wide variety of decedent and proxy-related demographic factors to assess GCC during the last year of life.^4,5 Therefore, we aimed to examine the association between personal characteristics, including race and health factors, as well as a surviving family member's assessment of GCC at the EOL to identify potentially modifiable factors or appropriate targets for interventions to improve quality of care.

Methods

Data sources

We accessed data from the Health and Retirement Study (HRS), funded by the National Institute on Aging and conducted by the University of Michigan.⁹ The HRS is a longitudinal cohort study of a nationally representative group of adults over the age of 50 in the United States. The cohort was created in 1992 and, with periodic sample replenishments, continues to survey ∼20,000 participants every other year.^9,10 A comprehensive overview of the HRS with information on recruitment and enrollment procedures can be found at http://hrsonline.isr.umich.edu

Every two years, HRS subjects complete “core” interviews. At each interview period, HRS staff identifies all participants—or decedents—who have died since the last interview and conducts a postdeath interview with a proxy, typically the next of kin, with knowledge about the decedent. Together, these interviews include decedent's demographic, socioeconomic, and health data, as well as proxy-reported information about the patient's EOL care experience.

Sample

One thousand two hundred forty-two HRS participants died and were included in the 2014 postdeath survey of their proxy. This number includes the overall postdeath response rate of 80.3% in 2014. On average, the postdeath proxy interview occurred 12.6 months after death. We excluded 67 subjects because the proxy either did not respond to the primary outcome measure (n = 66) or did not respond to several other questions of interest (n = 1). Overall, 95% of the sample responded to the question of interest.

Outcome measure

We defined the primary outcome, proxy-reported GCC, by the survey proxy's response to the following question: “Thinking about [his/her] experiences with the health care system over the past year, how often were [his/her] preferences for care taken into account: never, sometimes, usually, or always?”⁹ We used responses of usually or always to denote receipt of GCC.

Independent variables

We identified demographic factors that based upon the existing literature and our collective clinical experiences may influence proxy report of GCC.^{4–8,11–18} These variables were drawn from both the core and postdeath interviews, as appropriate. The primary independent variables of interest were self-reported race—defined as non-Hispanic White, non-Hispanic Black, or African American (denoted as African American throughout), or other—age at death, country of birth, language, marital status, education, net worth, religiosity, Medicare and Medicaid coverage, medical comorbidities, functional status, and EOL care factors. Proxies were asked about specific factors related to care at the EOL, which included designation of a durable power of attorney for health care or health care proxy, completion and type of ACP, ability to participate in EOL decision making, expectation of death, hospice utilization, location of death, and survey proxy's relationship to the decedent.

In the postdeath survey, proxies reported new and confirmed previously recorded chronic illnesses to create a comprehensive list of chronic diseases. Individual conditions were combined to create a single dichotomous variable representing the presence of three or more chronic conditions. Functional status was defined by the number of basic activities of daily living (ADLs)—bathing, grooming, dressing, toileting, transferring, and feeding—that proxies reported the decedent had required assistance with during the last three months of life.¹⁹ Severe impairment in functional status was defined as needing assistance with four or more ADLs. High religiosity was defined as going to religious services two or more times per month in the last year of life.

The postdeath survey asked two questions about the decedent's ACP: one asked if the decedent had written instructions about the type of treatment or care desired at the EOL; the other asked if the decedent had ever discussed preferences regarding EOL care with anyone. We combined the two variables to assess the presence of any form of written or verbal ACP.

Statistical analysis

We first examined descriptive statistics and frequency distributions for all variables. Next we used bivariate logistic regression to establish the relationships between the main outcome and independent variables for the full sample, and then stratified by race. Those variables with a statistically insignificant relationship with the outcome (p > 0.05) were excluded from multivariable analyses. Given the small sample size, we aimed to fit a parsimonious multivariable model and avoid multicollinearity. So, we examined a correlation matrix of the remaining variables, and for those highly related to each other, that is, for those pairs with correlation coefficients >0.25 or < −0.25, we excluded the superfluous variable. Finally, because prior studies have demonstrated significant relationships between race and EOL decision-making patterns, including engagement in ACP, we used the remaining variables to construct interaction terms with African American race and presence of written or verbal ACP and ability to participate in EOL decision making. By using these interaction terms, we could identify associations with race conditional on these contextual factors; relationships that may not be detected by examining the independent variables alone. The interaction terms were further tested for their independent contributions to the multivariable model using tests of goodness of fit: Akaike information criterion,²⁰ Bayesian information criteria,²¹ and likelihood ratio test. Those variables lacking statistical significance and failing to improve the fit of the model were excluded. This study was determined to be exempt by the institutional review board at the Icahn School of Medicine at Mount Sinai.

Results

We identified 1175 deceased HRS participants—or decedents—whose proxies responded to the 2014 survey question regarding GCC. Table 1 presents study sample characteristics. On average, decedents were 80 years old at the time of death. Half (51%) were female, 72% were non-Hispanic White, and 17% were African American. The most common chronic medical conditions included hypertension (71%), a heart condition (53%), and cancer (38%). Fifty-six percent of decedents had three or more chronic medical conditions, while 4% had no known chronic medical conditions.

Table 1.

Decedent Characteristics

Characteristics	Full sample (n = 1175)	Proxy report of GCC (n = 893)
Decedent variables
Female, %	51.23	52.30
Age at death (years)	79.7	80.5
Non-Hispanic White, %	72.34	74.58
Non-Hispanic Black or African American, %	17.02	15.79
Other race, %	10.64	9.63
U.S. born, %	92.26	93.28
English speaking, %	96.94	97.42
Married, %	44.51	44.23
High school education or higher, %	72.09	72.56
Net worth, mean inflation adjusted to 2012 $US	340,061	365,623
High Religiosity, %	36.60	36.73
Medicare coverage, %	86.21	87.68
Medicaid coverage, %	26.47	25.31
3+ Chronic conditions, %	55.83	57.78
Severe functional impairment, %	40.34	41.99
End of life care factors
Hospice at the end of life, %	46.21	48.82
Health care proxy, %	62.47	65.06
Written and/or verbal ACP, %	70.64	73.35
Respondent able to participate in end of life decisions, %	14.21	14.67
Death expected, %	58.38	60.25
Death in hospital, %	31.32	30.12
Survey proxy variables
Survey proxy is spouse of decedent, %	34.55	33.15
Survey proxy is child of decedent, %	48.26	50.73

ACP, advance care planning; GCC, goal-concordant care.

Three-quarters (76%) of proxy respondents reported that the decedent “usually or always” received GCC at the end of life. Table 2 illustrates the bivariate relationships between proxy-reported GCC and the independent variables for the entire population, as well as stratified by race. Among the full sample, older age, multimorbidity, functional impairment, ACP, and other EOL care factors were positively associated with report of GCC, while African American race was inversely associated with GCC. When examining the bivariate relationships within racial subgroups, African American decedents were significantly more likely to report GCC when proxy reported that the decedent was able to participate in EOL decision making (odds ratio [OR] 3.83, 95% confidence interval [CI] 1.11–13.26), while no bivariate relationship existed for this variable when examining White patients, other races, or the population as a whole. Among African Americans, report of GCC was also more likely among those with a high school education or higher (OR 1.87, 95% CI 1.01–3.46) and a legally appointed proxy for medical decisions (OR 1.91, 95% CI 1.01–3.61). While there was a significant relationship between report of GCC and presence of ACP for the full sample, this bivariate relationship did not reach statistical significance among the subgroup of African American decedents (OR 1.83, 95% CI 0.99–3.40).

Table 2.

Odds Ratios of Factors Associated with Proxy-Reported Goal-Concordant Care among Full Cohort and Stratified by Race

Variables	Proxy report of GCC, full cohort	Proxy report of GCC, stratified by race
Variables	Proxy report of GCC, full cohort	Non-Hispanic White	African American	Other race
Decedent variables
Female	1.19	1.02	1.82	1.56
Age at death	1.02^**	1.02^**	1.00	1.04^*
Non-Hispanic White	1.56^**
African American	0.71^*
Non-Hispanic, other race	0.45
Hispanic ethnicity	0.75
U.S. born	1.71^*	1.85	1.46	1.26
English speaking	1.83	1.00	1.00	1.37
Married	0.95	1.01	0.89	0.62
High school education or higher	1.10	0.70	1.87^*	0.92
High religiosity	1.02	1.06	1.61	0.56
Medicare coverage	1.61^*	1.58	0.95	0.98
Medicaid coverage	0.79	0.70	1.18	1.34
3+ Chronic conditions	1.39^*	1.36	1.73	1.05
Severe functional impairment	1.34^*	1.26	1.61	1.27
End-of-life care factors
Hospice at the end of life	1.56^**	1.36	1.44	3.14^*
Health care proxy	1.57^**	1.32	1.91^*	1.60
Written or verbal ACP	1.68^**	1.48^*	1.83	1.53
Respondent able to participate in EOL decisions	1.17	1.00	3.83^*	0.81
Death expected	1.37^*	1.55^*	1.15	0.78
Survey proxy variable
Survey proxy is spouse of deceased	0.78	0.81	0.75	0.51
Survey proxy is child of deceased	1.52^**	1.41^*	1.98^*	1.49

p < 0.05,

p < 0.01, odds ratios without asterisks did not reach statistical significance.

EOL, end of life.

Table 3 illustrates the results of the multivariate analysis. Decedent characteristics found to be independently associated with proxy report of GCC included the following: older age at death (adjusted odds ratio [AOR] 1.02, 95% CI 1.01–1.03), having three or more chronic medical conditions (AOR 1.34, 95% CI 1.02–1.77), and the presence of written or verbal ACP (AOR 1.38, 95% CI 1.02–1.88). The adjusted model revealed no significant relationships between GCC and race or one's ability to participate in decision making at the end of life; yet, the interaction term of African American race and ability to participate in EOL decision making was associated with higher odds of proxy report of GCC (AOR 3.83, 95% CI 1.02–14.40), indicating that conditional upon being able to participate or being perceived to be able to participate, African American decedents had proxies report higher rates of GCC.

Table 3.

Multivariate Analysis of Proxy-Reported Goal-Concordant Care

Variables	Adjusted odds ratio	95% Confidence interval
Age at death	1.02^**	1.01–1.03
African American	0.73	0.50–1.06
U.S. born	1.48	0.92–2.37
3+ Chronic conditions	1.34^*	1.02–1.77
Severe functional impairment	1.12	0.84–1.51
Hospice at end of life	1.33	0.99–1.77
Written or verbal ACP	1.38^*	1.02–1.88
Respondent able to participate in EOL decisions	0.90	0.58–1.39
African American + able to participate in EOL decision making	3.83^*	1.02–14.40

Multivariate logit:

p < 0.05,

p < 0.01.

Discussion

This survey of bereaved next-of-kin proxy respondents found that decedents' older age, multimorbidity, and ACP, defined by reported ACP discussions or presence of an advance directive, were independently associated with greater proxy report of GCC at the EOL. While African American race was not independently associated with GCC, proxies of the subset of African Americans who had proxy-reported ability to participate in their own EOL decision making were more likely to report GCC.

Advanced age and multimorbidity correlate positively in this study with proxy report of GCC. While this study was unable to assess why those who possess these characteristics receive more GCC, we hypothesize that older people and those with multiple chronic illnesses have more health care encounters, and therefore more opportunities to discuss their prognosis and preferences for care in the setting of serious illness. Alternatively, proxies' perceptions and assessment of GCC near the EOL may be systematically different, and potentially considered less shocking in the context of multimorbidity and advanced age. Further research is needed to understand the driving forces behind this finding and test this hypothesis.

The lack of association in our data between race and proxy-reported GCC at the EOL is consistent with the findings of two recent studies examining proxy report of GCC and other aspects of care quality through postdeath surveys about Medicare decedents.^4,5 Our results conflict, however, with the previous literature that found that, even with ACP, African American patients are less likely to receive GCC.^14,22,23 While some evidence suggests that this disparity is decreasing with time, it still exists.²⁴ One reason for the difference may be that these studies compare written ACP documentation or patient report of preferences in the months prior to death to the medical care received in the last week(s) of life. Other research has demonstrated that African American patients are more likely to change their goals and preferences over time^16,25 and to have verbal ACP conversations rather than written documentation of preferences.^15,26 The method of assessing GCC in this study, unlike others, accounts for possible changes in preferences over time and informal discussions of preferences that may ultimately influence care.

While neither race nor proxy-reported ability to participate in EOL decision making was an independent predictor of GCC, African Americans—but not other races—who were reported by proxies to be able to participate in their own EOL decision making were significantly more likely to have proxy-reported GCC. This observational study cannot evaluate causality and could not measure whether patient participation in medical decisions actually occurred. While statistically significant, the wide CI reflects imprecision and highlights the limitation of the small sample size; only 15% of patients in the sample were reported as being able to participate in EOL medical decisions. Nevertheless, we hypothesize that when African American patients were able to participate or were perceived to be able to participate in their own EOL decision making and directly communicated their preferences, proxies were more confident that GCC was provided. Whereas among those who could not participate, the proxy was less confident that preferences informed the care provided, therefore less likely to report GCC. This shift in confidence is perhaps experienced differentially across racial groups. Together, these findings support the overarching importance of bringing the patient's voice into the EOL medical decision-making process. The most effective mode of doing so may vary across racial or cultural groups, and efforts to train providers to discuss these topics with diverse populations without bias are needed.^27–29

Our findings must be interpreted in the context of certain study limitations. The relatively small sample size of this study limited the ability to detect significant relationships, particularly among racial and ethnic subgroups. The use of proxy report of GCC may inaccurately represent decedent's own assessment of whether care aligned with personal preferences during the course of illness. Also, chronic medical conditions were measured through prior patient report with additional diagnoses added by the proxy who may not have had full knowledge or understanding of the decedent's medical history before death potentially affecting the accuracy of this measure. Although survey proxies consisted predominantly of spouses or children, only 33% reported that they were personally “consulted” in the context of EOL medical decision making. It is unknown whether these family members were also the patient's health care surrogates. Respondents in our study may have not been aware of the decedent's care preferences or have predicted them inaccurately. Finally, proxy perception of GCC at the EOL may shift over time after death. The average length of time from death to survey completion of 12.6 months could systematically bias the proxies' recollections or interpretations of the care received. Further research regarding family member perception of EOL care over time is important to the interpretation of postdeath surveys as a measure of quality.

Despite these limitations, proxy report of GCC in and of itself is an important outcome. A person's feeling that appropriate care was given to his or her loved one near death may help in the grieving process. ACP has been shown to improve GCC and patient and family satisfaction, as well as reduce burdensome care and surviving loved ones' stress, anxiety, and depression after a patient's death.^11,13

Extrapolating from the results presented here, we hypothesize that this may be because ACP leads to more GCC, which in turn improves satisfaction. Furthermore, when assessing overall family satisfaction with care delivered to critically ill patients, families stress the importance of providers' provision of understandable explanations, and the inclusion of families in clinical discussions, shared decision making, and goal setting.^6–8 All these factors are associated with clear communication, which also helps providers understand a patient's preferences and goals for care, and thereby helps them tailor treatment to align with these. Little research thus far has directly linked GCC to patient satisfaction.¹

In this study, ACP—whether verbal or written—was associated with increased likelihood of proxy-reported GCC at the end of life regardless of race. These findings support efforts to broaden the medical community's approach to ACP as well as their understanding of what types of ACP have the greatest effect.^1,2,30,31 To ensure high-quality care for all patients, we need further research to assess the effectiveness of various means of promoting ACP throughout the course of serious illness and to prepare potential surrogate decision makers. By doing so, the patient's voice may be incorporated into care decisions, and the patient's evolving goals may guide the plan of care.

Footnotes

Acknowledgments

The Health and Retirement Study is funded by the National Institute on Aging (NIA) (U01 AG009740) and the Social Security Administration, and is performed at the Institute for Social Research, University of Michigan. Dr. Kelley receives support from the NIA R01AG054540.

Author Disclosure Statement

No competing financial interests exist.

References

Sanders

, Curtis

, Tulsky

: Achieving goal-concordant care: A conceptual model and approach to measuring serious illness communication and its impact. J Palliat Med, 2018; 21:S17–S27.

Johnson

, Butow

, Kerridge

, et al.: How well do current measures assess the impact of advance care planning on concordance between patient preferences for end-of-life care and the care received: A methodological review. J Pain Symptom Manage, 2018; 55:480–495.

Unroe

, Hickman

, Torke

; AAHPM Research Committee Writing Group: Care consistency with documented care preferences: Methodologic considerations for implementing the “measuring what matters” quality indicator. J Pain Symptom Manage, 2016; 52:453–458.

Sharma

, Freedman

, Mor

, et al.: Association of racial differences with end-of-life care quality in the United States. JAMA Intern Med, 2017; 177:1858–1860.

Khandelwal

, Curtis

, Freedman

, et al.: How often is end-of-life care in the united states inconsistent with patients' goals of care?. J Palliat Med, 2017; 20:1400–1404.

Clark

, Milner

, Beck

, Mason

: Measuring family satisfaction with care delivered in the intensive care unit. Crit Care Nurse, 2016; 36:e8–e14.

DeSanto-Madeya

, Safizadeh

: Family satisfaction with end-of-life care in the intensive care unit: A systematicreview of the literature. Dimens Crit Care Nurs, 2017; 36:278–283.

Pfaff

, Markaki

: Compassionate collaborative care: An integrative review of quality indicators in end-of-life care. BMC Palliat Care, 2017; 16:65.

Health and Retirement

Study

, public use dataset: Produced and distributed by the University of Michigan with funding from the National Institute on Aging (grant number NIA U01AG009740). Ann Arbor, MI, 2017.

10.

Sonnega

, Faul

, Ofstedal

, et al.: Cohort profile: The Health and Retirement Study (HRS). Int J Epidemiol, 2014; 43:576–585.

11.

Wright

, Zhang

, Ray

, et al.: Associations between end-of-life discussions, patient mental health, medical care near death, and caregiver bereavement adjustment. JAMA, 2008; 300:1665–1673.

12.

Bischoff

, Sudore

, Miao

, et al.: Advance care planning and the quality of end-of-life care in older adults. J Am Geriatr Soc, 2013; 61:209–214.

13.

Detering

, Hancock

, Reade

, Silvester

: The impact of advance care planning on end of life care in elderly patients: Randomised controlled trial. BMJ, 2010; 340:c1345.

14.

Mack

, Paulk

, Viswanath

, Prigerson

: Racial disparities in the outcomes of communication on medical care received near death. Arch Intern Med, 2010; 170:1533–1540.

15.

Sanders

, Robinson

, Block

: Factors impacting advance care planning among African Americans: Results of a systematic integrated review. J Palliat Med, 2016; 19:202–227.

16.

Mukamel

, Ladd

, Temkin-Greener

: Stability of cardiopulmonary resuscitation and do-not-resuscitate orders among long-term nursing home residents. Med Care, 2013; 51:666–672.

17.

Kurella Tamura

, Goldstein

, Pérez-Stable

: Preferences for dialysis withdrawal and engagement in advance care planning within a diverse sample of dialysis patients. Nephrol Dial Transplant, 2010; 25:237–242.

18.

Danis

, Garrett

, Harris

, Patrick

: Stability of choices about life-sustaining treatments. Ann Intern Med, 1994; 120:567–573.

19.

Noelker

, Browdie

: Sidney Katz, MD: A new paradigm for chronic illness and long-term care. Gerontologist, 2014; 54:13–20.

20.

Akaike

: Akaike's information criterion. In: Lovric

(ed): International Encyclopedia of Statistical Science. Berlin, Heidelberg: Springer, Berlin Heidelberg, 2011, pp. 25–25.

21.

Schwarz

: Estimating the dimension of a model. Ann Stat, 1978; 6:461–464.

22.

Loggers

, Maciejewski

, Paulk

, et al.: Racial differences in predictors of intensive end-of-life care in patients with advanced cancer. J Clin Oncol, 2009; 27:5559–5564.

23.

Johnson

: Racial and ethnic disparities in palliative care. J Palliat Med, 2013; 16:1329–1334.

24.

Koss

, Baker

: Race differences in advance directive completion. J Aging Health, 2017; 29:324–342.

25.

Johnson

, Kuchibhatla

, Tanis

, Tulsky

: Racial differences in hospice revocation to pursue aggressive care. Arch Intern Med, 2008; 168:218–224.

26.

Gerst

, Burr

: Planning for end-of-life care: Black-White differences in the completion of advance directives. Res Aging, 2008; 30:428–449.

27.

Periyakoil

, Neri

, Kraemer

: No easy talk: A mixed methods study of doctor reported barriers to conducting effective end-of-life conversations with diverse patients. PLoS One, 2015; 10:e0122321.

28.

Chapman

, Kaatz

, Carnes

: Physicians and implicit bias: How doctors may unwittingly perpetuate health care disparities. J Gen Intern Med, 2013; 28:1504–1510.

29.

Smith

, Davis

, Krakauer

: Differences in the quality of the patient-physician relationship among terminally ill African-American and white patients: Impact on advance care planning and treatment preferences. J Gen Intern Med, 2007; 22:1579–1582.

30.

Sudore

, Fried

: Redefining the “planning” in advance care planning: Preparing for end-of-life decision making. Ann Intern Med, 2010; 153:256–261.

31.

Sudore

, Lum

, You

, et al.: Defining advance care planning for adults: A consensus definition from a multidisciplinary delphi panel. J Pain Symptom Manage, 2017; 53:821–832.e1.