To examine advance care planning (ACP) trends among an increasingly diverse aging population, we compared informal and formal ACP use by race/ethnicity among U.S. older adults (≤ 65 years).
We used Health and Retirement Study data (2012-2018) to assess relationships between race/ethnicity and ACP type (i.e., no ACP, informal ACP only, formal ACP only, or both ACP types). We reported adjusted risk ratios with 95% confidence intervals.
Non-Hispanic Black and Hispanic respondents were 1.77 (1.60, 1.96) and 1.76 (1.55, 1.99) times as likely, respectively, to report no ACP compared to non-Hispanic White respondents. Non-Hispanic Black and Hispanic respondents were 0.74 (0.71, 0.78) and 0.74 (0.69, 0.80) times as likely, respectively, to report using both ACP types as non-Hispanic White respondents.
Racial/ethnic differences in ACP persist after controlling for a variety of barriers to and facilitators of ACP which may contribute to disparities in end-of-life care.
Keywords: Advance care planning, healthcare disparities, advance directives, end of life, race and ethnicity
Extensive advance care planning (ACP) programs, in which individuals are supported in identifying and describing their healthcare treatment values, goals, and preferences, have been shown to increase patients’ satisfaction with care, lessen decisional conflict among patient surrogates, and reduce healthcare treatment intensity at the end-of-life (Brinkman-Stoppelenburg et al., 2014; Jimenez et al., 2018; Smith-Howell et al., 2016; Sudore et al., 2017). ACP that incorporates goals-of-care conversations between patients, their family members, and providers (i.e., informal ACP) may be especially effective at maximizing important end-of-life outcomes when compared to written documents alone (Baidoobonso, 2014; Jimenez et al., 2018). However, less is known about national trends and practices in informal ACP compared to formal ACP (i.e., advance directives) since most research efforts have concentrated on the legal documentation of patients’ preferences.
The focus on advance directives has limited our understanding of the breadth of ACP types used across an increasingly racially and ethnically diverse U.S. older adult population (Administration on Aging, 2021). Several racially and ethnically minoritized older adult groups are less likely to complete advance directives than their non-Hispanic White peers (Portanova et al., 2017; Sullivan & Klingman, 2019). The decision-making practices of these groups regarding end-of-life care may be less adequately addressed by traditional, autonomy-centered advance directives than by informal ACP (Zager & Yancy, 2011). Indeed, Black Americans are more likely to rely on trusted family members or clergy to communicate their end-of-life preferences than to express those preferences in advance directives (Collins et al., 2018; Sanders et al., 2016). Similarly, Hispanic patients’ preference to include family members in healthcare decision-making and strong religious beliefs are not adequately accommodated by formal ACP (Cervantes et al., 2017; Shen et al., 2020).
Though evidence suggests some racially and ethnically minoritized older adult groups have identified preferences for informal ACP over formal ACP, a few studies exploring trends among older adults with chronic illness or living in a nursing home suggest that racial/ethnic differences in formal ACP extend into informal ACP (Carr, 2011; Rich et al., 2009). However, while racial/ethnic differences in formal ACP have consistently been associated with sociodemographic factors like estate planning, religious preferences, and acculturation as well as health factors (Koss & Baker, 2018; Sanders et al., 2016; Yi, 2019), less is known about the factors contributing to racial/ethnic differences in older adults’ informal ACP. In addition, we were unable to identify a study that examined racial/ethnic differences in both formal and informal ACP among a national, primarily community-dwelling sample of U.S. older adults.
Prior work has consistently identified racial/ethnic differences in formal ACP across the U.S. older adult population with recent work additionally indicating differences in informal ACP among specific population groups. We build on this work by examining racial/ethnic differences and trends in formal and informal ACP using data from a nationally representative survey of older adults in the U.S and identifying potential mechanisms behind these differences. Specifically, we aimed to 1) assess the relationship between race/ethnicity and ACP among U.S. adults aged 65 and older and 2) explore sociodemographic characteristics, potential barriers to and facilitators of ACP, and health factors that may explain this relationship. Our findings offer insight into the decision-making preferences and patterns of various racial/ethnic groups, and potential explanations for racial/ethnic differences in the use of ACP, with important implications for future ACP practice.
We conducted a retrospective, cross-sectional study using data from the Health and Retirement Study (HRS), a large, nationally representative longitudinal survey consisting of multiple birth cohorts of adults aged 51 and older (Sonnega et al., 2014). HRS data was chosen for this study because of the survey’s oversampling of key racial/ethnic groups, specifically Black and Hispanic households, and its inclusion of self-reported ACP. Respondents and their spouses/partners are interviewed approximately every two years on various topics related to aging. HRS is sponsored by the National Institute on Aging (grant number NIA U01AG009740) and conducted by the University of Michigan. Because of its use of publicly available, deidentified data, this study was deemed “not regulated” by the Health Sciences and Behavioral Sciences Institutional Review Board at the University of Michigan (HUM00196992).
We used pooled 2012 to 2018 data from HRS respondents for whom complete data was available. This range was selected because 2012 was the first year in which HRS investigators asked respondents about their ACP. We captured data from respondents’ last available interview during the study period, which was either of a) the last interview in which living respondents participated (through 2018), or b) the interview immediately preceding death for decedents. Because we were specifically interested in the ACP of older adults, we included only respondents aged 65 and older at the time of the interview.
Our primary outcome was ACP type, a nominal variable with four distinct, mutually exclusive categories; 1) No ACP, 2) Informal ACP only, 3) Formal ACP only, and 4) Both informal and formal ACP. We constructed this variable based on responses to the following three HRS questions:
“Have you discussed with anyone the care or medical treatment you would want to receive if you were to become seriously ill (i.e., goals-of-care conversations)?”
“Have you written instructions about the care or medical treatment that you would want to receive if you cannot make those decisions yourself (i.e., living will)?”
“Have you completed a legal arrangement for a specific person(s) to make decisions about your care or medical treatment if you cannot make those decisions yourself (i.e., durable power of attorney)?”
Respondents who reported no goals-of-care conversations nor possession of either advance directive (i.e., living will or durable power of attorney) were categorized as “No ACP”. Respondents who reported having a goals-of-care conversation but no advance directive were categorized as “Informal ACP only”. Respondents who did not report having a goals-of-care conversation but did report completion of at least one advance directive were categorized as “Formal ACP only”. Respondents who reported having a goals-of-care conversation and at least one advance directive were categorized as “Both ACP Types”.
To compare the use of ACP types by race/ethnicity, we identified respondents’ self-reported race/ethnicity, categorized into three groups: non-Hispanic White (i.e., White), non-Hispanic Black (i.e., Black), and Hispanic. Due to the small size (equal to less than 2.5% of the total sample) and racial/ethnic heterogeneity of the remaining sample, respondents who self-identified as American Indian, Alaskan Native, Asian, Native Hawaiian, Pacific Islander, or any other race/ethnicity were excluded from our analyses.
Models controlled for sociodemographic characteristics that have previously demonstrated an association with ACP or at least partially explained racial or ethnic differences in advance directive completion (Khosla et al., 2016; Rao et al., 2014). Sociodemographic characteristics included age (at interview), gender (i.e., female versus male), marital status, educational level, Medicaid eligibility, rural residence, country of birth (i.e., U.S. born versus elsewhere), attendance at religious services, net worth (in quartiles), and nursing home residence. Recent work indicates that older adults dually eligible for Medicare and Medicaid are less likely to have an outpatient ACP claim than those with Medicare eligibility alone (Palmer et al., 2021). We controlled for rural residence using the 2013 Beale Rural-Urban Continuum codes (i.e., rural versus urban/suburban) because end-of-life care practices may vary by geographic region, potentially impacting patient decision-making (Nicholas et al., 2011), and because older adults in rural areas may have fewer resources for engaging in ACP (Christensen et al., 2019). We included country of birth as a proxy measure for acculturation, given several studies’ findings that greater acculturation is associated with advance directive completion (Kelley et al., 2010; Yi, 2019). We also included attendance at religious services because prior work has demonstrated a positive association between religious service attendance (independent of religious affiliation) and advance directive completion (Hoe & Enguidanos, 2020; Koss, 2018). Several studies have also documented positive associations between advance directive completion and both greater financial resources (Kelly et al., 2013) and nursing home residence (Jones et al., 2011). These characteristics were divided into three categories according to their association with ACP: 1) baseline demographics (i.e., age, gender, marital status, and education), 2) potential barriers to ACP (i.e., Medicaid eligibility, rural residence, and birth outside the U.S.), and 3) facilitators of ACP (i.e., religious service attendance, net worth, and nursing home residence).
We also controlled for health characteristics that may increase the perceived need for ACP or access to ACP within the healthcare system. These included self-rated health, diagnosis of several chronic conditions, and functional status. We included a binary measure of self-rated health (i.e., good, very good, or excellent versus poor or fair) as a proxy for overall health given its strong associations with a number of important health outcomes (Giltay et al., 2012; Lee, 2000; van der Linde et al., 2013). We also included an index of seven common chronic conditions (i.e., Alzheimer’s disease and related dementias, arthritis, diabetes, heart disease, lung disease, psychiatric illness, stroke history). We separately controlled for a history of any cancer (excluding skin cancer) because individuals with a cancer diagnosis are more likely to complete an advance directive than those without (Park et al., 2015). Finally, we controlled for difficulties with Activities of Daily Living (e.g., bathing or dressing; range 0-5) or Instrumental Activities of Daily Living (e.g., handling money or taking medication; range 0-5) because individuals with poorer functional status have previously been more likely to complete ACP (Hansen et al., 2019; Kim et al., 2021).
First, we reported descriptive statistics overall and by racial/ethnic group using counts and percentages for categorical variables and means with standard deviations for continuous variables. We examined differences in respondent characteristics across racial/ethnic groups using ANOVA, chi-square, and Kruskal Wallis tests and we computed a correlation matrix between all continuous and ordinal covariates (Supplemental Table 1). We then estimated a series of five, sequential multinomial logistic regression models to explore factors that may help to explain the relationship between race/ethnicity and ACP type. In the first model, we regressed the nominal outcome, ACP type, on race/ethnicity indicators (i.e., Black and Hispanic with White as the omitted reference category). Subsequent models added the groups of covariates described above, beginning with the baseline demographics, then adding factors we characterize as potential barriers to ACP, then facilitators of ACP, and finally, health factors that may increase the perceived need for ACP. We employed this sequential modeling strategy to identify the extent to which any associations between race/ethnicity and ACP type could be independently explained by known or potential drivers, including barriers to or facilitators of ACP, that may differ between racial/ethnic groups. To account for potential correlation between spouses’/partners’ ACP (Koss, 2017), we clustered standard errors by household, allowing for intragroup correlation.
Multinomial logistic regression models were estimated for each of the five models to account for the nominal outcome (i.e., ACP type). These models produce a set of coefficients for all model covariates for each level of the nominal outcome except the baseline (i.e., reference) category. The model coefficients associated with each outcome (i.e., informal ACP only, formal ACP only, or both ACP types) represent the relative risk ratios (RRR) (and their 95% confidence intervals) of reporting a specific ACP type (e.g., formal ACP only) compared to no ACP for respondents of a specific race/ethnicity (e.g., Hispanic) compared to respondents of the reference race/ethnicity (e.g., White).
Because the model coefficients’ magnitudes are difficult to interpret, we additionally obtained adjusted risk ratios (aRRs) and 95% confidence intervals for each of the nominal outcomes using Stata’s adjrr post-estimation command. Unlike RRRs, aRRs represent relative likelihoods of a specific outcome (e.g., formal ACP only) for individuals of one racial/ethnic group compared to the specified reference group (whereas RRRs are relative likelihoods of a specific outcome compared to the baseline outcome [i.e., no ACP use] for each race/ethnicity compared to White). ARRs are conditional on the outcome; for instance, if most individuals in one racial/ethnic group use both ACP types, then the remaining individuals of that group may be less likely to have just one type of ACP. For this reason, a group could have an RRR1. aRRs do not depend on the reference category in the original model, allowing us to make comparisons between any two racial/ethnic groups.
Our final analytic sample included 13,117 unique older adults. Of these, 72% self-identified as White, 17% as Black, and 11% as Hispanic. The mean age of the White respondents was three years higher than that of the Black and Hispanic respondents. Fewer than 50% of the Hispanic respondents were born in the United States compared to over 90% of White and Black respondents. Additionally, a higher proportion of those in the Black and Hispanic groups rated their own health as poor or fair compared to the White group ( Table 1 ).
Characteristics of Respondents (n=13,117)
Total | NH White (n=9,489) | NH Black (n=2,168) | Hispanic (n=1,460) | |
---|---|---|---|---|
Characteristic | No. (%) | No. (%) | No. (%) | No. (%) |
Age *** , mean ± SD, years | 77.2 ± 8.2 | 78.1 ± 8.2 | 75.1 ± 7.9 | 74.8 ± 8.0 |
Female gender *** | 7,609 (58.0) | 5,425 (57.2) | 1,349 (62.2) | 835 (57.2) |
Marital status *** | ||||
Married/partnered | 6,971 (53.1) | 5,302 (55.9) | 853 (39.4) | 816 (55.9) |
Separated/divorced | 1,564 (11.9) | 908 (9.6) | 461 (21.3) | 195 (13.4) |
Widowed | 4,137 (31.5) | 3,043 (32.1) | 708 (32.7) | 386 (26.4) |
Never married | 445 (3.4) | 236 (2.5) | 146 (6.7) | 63 (4.3) |
Education *** | ||||
Less than high school | 2,638 (20.1) | 1,187 (12.5) | 659 (30.4) | 792 (54.3) |
High school graduate/GED | 4,606 (35.1) | 3,578 (37.7) | 687 (31.7) | 341 (23.4) |
Some college | 3,021 (23.0) | 2,270 (23.9) | 518 (23.9) | 233 (16.0) |
College graduate | 2,852 (21.7) | 2,454 (25.9) | 304 (14.0) | 94 (6.4) |
Medicaid beneficiary *** | 1,560 (11.9) | 642 (6.8) | 481 (22.2) | 437 (29.9) |
Rural residence *** | 3,650 (27.8) | 3,106 (32.7) | 362 (16.7) | 182 (12.5) |
Born in the US *** | 11,711 (89.3) | 9,070 (95.6) | 2,025 (93.4) | 616 (42.2) |
Attends religious services *** | 9,064 (69.1) | 6,260 (66.0) | 1,732 (79.9) | 1,072 (73.4) |
Net worth *** , quartile | ||||
1 (Low) | 2,547 (19.4) | 1,211 (12.8) | 800 (36.9) | 536 (36.7) |
2 | 3,095 (23.6) | 1,926 (20.3) | 742 (34.2) | 427 (29.3) |
3 | 3,493 (26.6) | 2,730 (28.8) | 433 (20.0) | 330 (22.6) |
4 (High) | 3,982 (30.4) | 3,622 (38.2) | 193 (8.9) | 167 (11.4) |
Nursing home resident *** | 859 (6.6) | 726 (7.7) | 90 (4.2) | 43 (3.0) |
Self-reported health poor/fair *** | 4,819 (36.7) | 3,145 (33.1) | 927 (42.8) | 747 (51.2) |
Noncancer chronic illnesses, mean ± SD | 1.9 ± 1.3 | 1.9 ± 1.3 | 2.0 ± 1.4 | 1.9 ± 1.3 |
Cancer history *** | 2,998 (22.9) | 2,388 (25.2) | 395 (18.2) | 215 (14.7) |
Functional status | ||||
No. of ADL difficulties *** , median (IQR) | 0 (1) | 0 (1) | 0 (1) | 0 (1) |
No. of IADL difficulties *** , median (IQR) | 0 (1) | 0 (1) | 0 (1) | 0 (1) |
Note. Not all categories total 100% due to rounding. Data represents respondents with complete data for final analytical model. Net worth quartiles defined as follows: 25%=$13,000; 50%=$130,000; 75%=$431,000. Counted chronic illnesses include Alzheimer’s disease and related dementias, arthritis, diabetes, heart disease, lung disease, psychiatric illness, and history of stroke (range 0-7). Rural-Urban status derived from the 2013 Beale Rural-Urban Continuum Codes.
ACP indicates advance care planning; ADL, activities of daily living; GED, General Educational Diploma; IADL, instrumental activities of daily living; IQR, interquartile range; NH, non-Hispanic; SD, standard deviation; US, United States.
Figure 1 shows the unadjusted distribution of ACP use across racial/ethnic groups. Overall, 82% of all respondents reported at least one ACP type and 59% reported using both formal and informal ACP. By group, 81% of White, 59% of Black, and 53% of Hispanic respondents reported using informal ACP, while 74% of White, 51% of Black, and 37% of Hispanic respondents reported using formal ACP. “Both ACP Types” was the most common outcome among White and Black respondents, while “No ACP” was the most common outcome among Hispanic respondents. Additionally, the respective unadjusted proportions of Black and Hispanic respondents reporting no ACP were approximately three and four times greater than those of the White group.
Distribution of Advance Care Planning Types within Racial and Ethnic GroupsACP indicates advance care planning; NH, non-Hispanic.
In adjusted results, across all five models, Black respondents were significantly less likely to use informal ACP only (Model 5 [M5] RRR 0.57 [95% CI 0.48, 0.68]) or both ACP types (M5 RRR 0.36 [95% CI 0.31, 0.42]) relative to no ACP compared to White respondents (Supplemental Table 2). However, while Black respondents were significantly less likely to use formal ACP relative to no ACP compared to White respondents in the first three models, this difference was no longer statistically significant with the addition of the facilitators of ACP (i.e., religious service attendance, net worth, and nursing home residence) in the fourth model (RRR 0.81 [95% CI 0.66, 1.00]). Across all models, Hispanic respondents were significantly less likely to use each ACP type relative to no ACP compared to White respondents (M5 RRR’s informal ACP only 0.78 [95% CI 0.63, 0.97], formal ACP only 0.42 [0.31, 0.57], both ACP types 0.35 [0.29, 0.43]).
Figure 2 depicts the adjusted relative risks derived from the final model (including all covariates) with White respondents as the reference group (aRRs for each of the five models are available in Supplemental Table 3). As shown, Black respondents were 77% more likely and Hispanic respondents 76% more likely to report no ACP compared to White respondents. Hispanic respondents were also 45% more likely to report informal ACP only and Black respondents were 58% more likely to report formal ACP only compared to White respondents. Black and Hispanic respondents were both 26% less likely to report using both ACP types than White respondents. Additionally, Black respondents were 89% more likely to report formal ACP only than Hispanic respondents and Hispanic respondents were 35% more likely to report informal ACP only compared to Black respondents.
Adjusted Relative Risks from Multinomial Logit Model of Advance Care Planning Type by Race/EthnicityNote. Model specification uses complete-case analysis (n=13,117). The reference category is “non-Hispanic White.” Adjusted relative risks were calculated using Stata’s adjrr post-estimation command after running the multinomial logit model. Diamonds indicate point estimates and lines indicate 95% CIs.
ACP indicates advance care planning; CI, confidence interval; NH; non-Hispanic.
ACP types varied significantly by race and ethnicity in this study of U.S. older adults. Goals-of-care conversations (i.e., informal ACP) were common. However, Black and Hispanic individuals were more likely to report no ACP and less likely to report both informal and formal ACP compared to White respondents, even after accounting for various barriers and facilitators previously associated with ACP. Collectively, these findings illustrate substantial gaps in national use of both informal and formal ACP among older adults, suggesting that current ACP practices may be inadequate for the needs of an increasingly racially and ethnically diverse U.S. older adult population.
We observed a higher overall prevalence of ACP compared to previous estimates, perhaps due to our incorporation of goals-of-care conversations in assessments of ACP use. Previous work shows that, despite recent growth, there is low uptake of formal ACP in both the older adult and general population (Silveira et al., 2014; Yadav et al., 2017). In a recent systematic review, only 46% of older adults had completed an advance directive (Yadav et al., 2017). In comparison, four out of five respondents in our study reported at least one ACP type and three out of five reported both. Including informal ACP with more traditionally assessed formal ACP methods increased observed prevalence of ACP, suggesting that informal ACP complements or, in some cases, is used in place of formal ACP among many older Americans.
We built on less representative prior work that separately examined formal or informal ACP by simultaneously identifying racial/ethnic differences in both ACP types among older adults. Prior studies consistently demonstrated lower rates of advance directive completion among racially and ethnically minoritized samples compared to White samples (Huang et al., 2016; Jimenez et al., 2018; Portanova et al., 2017), with some additional evidence indicating that these differences extend into informal ACP (Carr, 2012; Clark et al., 2018). We found that, while more than half of respondents reported informal ACP, there were substantial racial/ethnic differences in informal ACP, and these differences persisted after adjustment for factors that typically explain ACP use. These gaps were similar in nature to those related to formal ACP use. Together, these findings suggest substantial variation by race in perceptions of or access to ACP, whether informal or formal, meaning efforts to broaden access and use of ACP will require more tailored efforts.
Though we found that, compared to White respondents, Black respondents were more likely to use formal ACP only and Hispanic respondents were more likely to use informal ACP only, this potential discrepancy (less likelihood of any ACP use compared to no ACP, but greater likelihood of specific ACP use) is explained by the distribution of individuals by race/ethnicity in each outcome category. Because more White respondents than other individuals used both types of ACP, more Black and Hispanic respondents were in the other ACP outcome categories. Smaller proportions of Black and Hispanic respondents used either ACP type compared to White respondents, and adjusted results indicated that Black and Hispanic respondents were significantly more likely to report no ACP compared to White respondents. In all, these findings suggest that, even when accounting for goals-of-care conversations, Black and Hispanic older Americans are substantially less likely to complete any ACP compared to White older Americans.
There are several potential explanations for these persistent gaps in ACP across racial/ethnic groups. In our sequential models, modest attenuation in the relationship between race/ethnicity and ACP type was primarily explained by sociodemographic characteristics (e.g., age and education) and not by factors representing potential barriers to or facilitators of ACP. However, country of birth, nursing home residence, number of noncancer chronic conditions, and history of cancer were all significantly associated with each ACP type in the final model. In the case of Black respondents using formal ACP only relative to no ACP, the addition of potential facilitators of ACP muted observed differences in the relative odds of ACP versus no ACP use by race/ethnicity. These findings corroborate a large body of prior work which suggests that greater social and economic resources enhance the use of advance directives and more recent work suggesting that socioeconomic disadvantage reduces the odds of having end-of-life conversations (Boerner et al., 2021; Hong & Kim, 2020; Khosla et al., 2016).
Fewer social and economic resources among racially and ethnically minoritized older adults due to structural racism may impact ACP (Bailey et al., 2017). For example, our study and others identified an association between nursing home residence and completion of ACP (Jones et al., 2011). However, racial segregation persists in nursing home settings across the United States with Black residents often receiving poorer quality care (Estrada et al., 2021; Mack et al., 2020). Similarly, though greater educational attainment is also consistently associated with ACP (Huang et al., 2016; Rao et al., 2014), Black and Hispanic respondents in our study reported lower overall educational attainment than White respondents, potentially reflecting reduced access to ACP resources.
However, large differences in ACP by race/ethnicity remained even after accounting for known barriers and facilitators to ACP, implying a need for alternative explanations for such gaps. Though we controlled for several factors previously associated with healthcare treatment preferences, including socioeconomic status, chronic illness diagnoses, and functional status (Chang et al., 2014; Higginson et al., 2017), we were unable to measure respondents’ specific healthcare treatment preferences. In several prior studies, Black Americans were more likely to prefer life-sustaining treatments at the end-of-life than their peers of other racial/ethnic groups, reducing the perceived need for ACP (Barnato et al., 2009; Johnson et al., 2008; Rahemi & Williams, 2016). Recent work has suggested an effect of subjective life expectancy on informal ACP use and identified more optimistic life expectancy estimates among Black older adults and more pessimistic estimates among Hispanic older adults compared to White older adults. However, racial/ethnic differences in ACP persisted after controlling for subjective life expectancy, suggesting that other factors may play a larger role (Lou & Carr, 2022),
Another alternative explanation involves unmeasured patient-provider factors. Clinicians are less likely to initiate ACP conversations with individuals with whom there are perceived or real barriers to communication (e.g., language barriers, differences in communication style, etc.) or if they have preconceived views about their patients’ preferences (Ashana et al., 2022; Bazargan et al., 2021; Ladin et al., 2021). Implicit bias may therefore perniciously influence how informal and formal ACP is offered. Clinicians’ lack of awareness about racial/ethnic differences in subjective life expectancy and other culturally varying factors might inhibit efforts to offer nuanced and beneficial support to patients and family members. Healthcare system factors, including limited access to high-quality care, may also contribute to lack of awareness of healthcare treatment options, which has been observed among Hispanic older adults (Carr, 2012; Maldonado et al., 2019; Shen et al., 2020).
More broadly, ACP centers around traditionally western values (i.e., patient autonomy, informed decision-making, truth-telling, and discussing the dying process), excluding groups with different values who must adapt in order to complete traditional ACP (Blackhall et al., 1995; Sabatino, 2010; Zager & Yancy, 2011). Lower ACP rates among racially and ethnically minoritized groups may also be a consequence of a healthcare system that has historically failed to invest in culturally competent care models (Braveman et al., 2011; McCleskey & Cain, 2019; Washington, 2008). In our study, the relative risks of using each ACP type relative to no ACP were significantly higher for respondents born in the United States compared to those born elsewhere. It has been proposed that, in addition to improving overall access to healthcare, policymakers and funding organizations should direct resources towards developing and improving access to culturally responsive palliative care interventions that support patients as they encounter significant health changes, rather than focusing on planning for unknown future crises in advance (Morrison, 2020; Rutz Voumard et al., 2021). Indeed, ACP may occur more frequently and function more effectively within the context of clinically relevant palliative care.
Our findings must be considered within the context of two important limitations. First, the lack of differentiation between other racial/ethnic groups in the larger HRS dataset (e.g., Asian Americans, Native Americans, and Pacific Islanders) precluded our ability to identify patterns among them, an important limitation given known differences in end-of-life care planning and preferences among these groups (McDermott & Selman, 2018; Zager & Yancy, 2011). Second, our measures of ACP were dependent on respondents’ or their proxies’ self-report, which were subject to recall and other self-report biases. Future, prospective research may better elicit the persons (i.e., family members, healthcare providers, etc.) with whom the respondents’ completed ACP, particularly during goals-of-care conversations. Despite these limitations, this project fills an important gap in the ACP research literature. The use of HRS, a large, nationally representative longitudinal survey that oversamples several key demographic groups, strengthened our ability to examine ACP trends among populations (e.g., elderly, Black, and Hispanic households) which have been particularly vulnerable to poor health outcomes during the ongoing pandemic. Additionally, our inclusion of informal ACP revealed sizable differences in ACP utilization between racial/ethnic groups in a national sample of older adults, which may reflect important deficiencies in current ACP structures in U.S. healthcare.
Our study revealed important differences in ACP among older adults of different racial/ethnic groups in the United States. Racially and ethnically minoritized older individuals were less likely to use a combination of informal and formal ACP and more likely to be without any ACP. Though these differences may be explained, in part, by racial/ethnic differences in sociodemographic characteristics, barriers to ACP, facilitators of ACP, or health factors commonly associated with ACP, further work is needed to identify patient preferences or healthcare system factors that may impact older adults’ ability to complete advance directives or willingness to discuss their healthcare treatment goals. Our study identified racial and ethnic differences in both formal and informal ACP as well as potential mechanisms behind those differences using a national sample of older adults. Future work must identify additional strategies to equitably improve end-of-life care for older adults in the United States.