A Decade of Studying Drivers of Disparities in End-of-Life care for Black Americans: Using the NIMHD Framework for Health Disparities Research to Map the Path Ahead

Abstract

Objective:

The purpose of this paper is to provide a review of the existing literature on racial disparities in quality of palliative and end-of-life care and to demonstrate gaps in the exploration of underlying mechanisms that produce these disparities.

Background:

Countless studies over several decades have revealed that our healthcare system in the United States consistently produces poorer quality end-of-life for Black compared with White patients. Effective interventions to reduce these disparities are sparse and hindered by a limited understanding of the root causes of these disparities.

Methods:

We searched PubMed, CINAHL and PsychInfo for research manuscripts that tested hypotheses about causal mechanisms for disparities in end-of-life care for Black patients. These studies were categorized by domains outlined in the National Institute of Minority Health and Health Disparities (NIMHD) framework, which are biological, behavioral, physical/built environment, sociocultural and health care systems domains. Within these domains, studies were further categorized as focusing on the individual, interpersonal, community or societal level of influence.

Results:

The majority of the studies focused on the Healthcare System and Sociocultural domains. Within the Health Care System domain, studies were evenly distributed among the individual, interpersonal, and community level of influence, but less attention was paid to the societal level of influence. In the Sociocultural domain, most studies focused on the individual level of influence. Those focusing on the individual level of influence tended to be of poorer quality.

Conclusions:

The sociocultural environment, physical/built environment, behavioral and biological domains remain understudied areas of potential causal mechanisms for racial disparities in end-of-life care. In the Healthcare System domain, social influences including healthcare policy and law are understudied. In the sociocultural domain, the majority of the studies still focused on the individual level of influence, missing key areas of research in interpersonal discrimination and local and societal structural discrimination. Studies that focus on individual factors should be better screened to ensure that they are of high quality and avoid stigmatizing Black communities.

Introduction

For several decades, the palliative care literature has documented disparities in end-of-life (EOL) care for Black patients in the United States. Black patients receive hospice services at a lower rate, are more likely to be hospitalized and receive care in an intensive care unit at the EOL and have lower rates of completion of advance directives.(1–20) Even more concerning is the finding that bereaved family members of Black patients are more likely to report that their loved-one received poor quality EOL care.(21–23) Early studies examining causes of EOL disparities focused on differences beliefs, behavior and characteristics of Black patients and families.(5, 7, 9, 24–30)

The Black Lives Matter movement and the COVID-19 pandemic have brought renewed attention to racial disparities in health care. This environment has opened the door for a re-examination of the focus of health disparities literature. In particular, scholars such as Dr. Rhea Boyd have clearly identified the need to focus on various forms of racism; structural, institutional and interpersonal, as explanatory factors in health disparities research.(31) Boyd and colleagues argue that research focusing on biological and behavioral characteristics of Black patients obfuscates the role of racism in producing disparities. Similarly, Dr. Rachel Hardeman examines the framing of research questions in health services research, finding that focusing on a deficit model of health disparities upholds racism within our research infrastructure.(32)

These critiques invite us to systematically examine the state of EOL health disparities research in order to identify gaps in our framing of research questions. Where might we be missing opportunities to name and examine various types of racism as primary drivers of disparities? What research practices may be inadvertently upholding racist ideologies? In 2018 the National Institute on Minority Health and Health Disparities (NIMHD) published a research framework aimed at conceptualizing the “wide array of determinants that promote or worsen minority health or cause, sustain or reduce health disparities.”(33) This framework is meant to widen the lens though which researchers ask questions about etiological factors and intervention targets. The NIMHD framework identifies domains of determinants of disparities (biological, behavioral, sociocultural and environmental) and further categorizes these domains using the socioecological model into four levels of influence; individual, interpersonal, community and societal. The authors noted that in the broader health disparities literature, there is an overrepresentation of studies examining the individual level with fewer focusing on community and societal levels of influence. This suggests that institutional and structural racism are under-investigated in the broader health disparities literature.

We aimed to use the NIMHD framework to evaluate the literature on disparities in EOL outcomes for Black patients and families. This review may reveal areas of investigation that are under-represented in the literature and suggest a path forward to a more comprehensive understanding of the etiologies of critical disparities in quality of EOL care.

Methods:

We searched the PubMed, CINAHL and PsychInfo databases for articles published between 5/17/2011 and 5/17/2021 using the search terms palliative care, advance care planning, advance directives, surrogate decision making, living will terminal care, hospice, resuscitation orders, DNR orders, end-of-life care, terminal illness African Americans and Black. An example of the search strategy for PubMed can be found in Appendix 1.

Inclusion Criteria

Articles including Black or African American adult population in the United States and patients with serious or terminal illness OR healthy adults in studies of advance care planning pertaining to future serious or terminal illness were included. Studies with quantitative methods including cross-sectional, cohort, case/control, pre/post intervention studies and randomized controlled trials were included. Articles published in the English language in peer reviewed journals were included. Finally, articles must address causes of disparities in type of EOL care (rather than chronic care) delivered, processes of care (e.g. palliative care consult), symptom management, advance care planning or patient/family ratings of quality of EOL care to be included. Studies of interventions to ameliorate these causes were included.

Exclusion Criteria

We excluded studies that documented disparities alone without exploring causal mechanisms. Qualitative studies, editorials and reviews were excluded. Case studies/case series, pilot, exploratory and feasibility studies were excluded. Studies of instrument development and validation studies were excluded. Studies focusing on populations outside of the United States and pediatric populations were excluded. Studies focused on physician aid in dying or physician assisted suicide were excluded.

Categorization and Evaluation of Quality

One investigator, EC, evaluated all articles and a second investigator (SY or AG or JN or JW) independently evaluated them for risk of bias and categorization in the twenty categories possible by combining the 5 domains × 4 levels of influence using the NIMHD framework.(33) Articles could be placed in more than one category as many evaluated multiple causal factors. We used the National Institute of Health quality assessment tools to evaluate risk of bias in each article (https://www.nhlbi.nih.gov/health-topics/study-quality-assessment-tools). Differences in categorization or quality rating were resolved with discussion.

Results

After removal of duplicates, 950 unique titles were returned from the PubMed, PsychInfo and CINAHL databases from 5/17/2011 – 5/17/2021. Of those, 507 were excluded by title and 443 abstracts were reviewed for inclusion and exclusion criteria. 391 articles were excluded either by abstract or full-text review including 179 which evaluated for disparities without examining a causal mechanism, 115 which did not include a hypothesis related to disparities in EOL outcomes, 92 which were not the accepted article type or population of interest and 5 for which we were unable to obtain the full text of the article (Figure 1). Of the studies which evaluated for disparities without examining a causal mechanism, 28 failed to detect racial disparities or demonstrated better outcomes for Black patients, including two that found increased use of inpatient palliative care services(34, 35) and two that showed increased caregiver psychological wellbeing and resilience.(36, 37) After the screening process, 52 articles were included in this review (see Appendix 2 for a table detailing each article).

Figure 1:

Flow diagram showing screening of manuscripts

Domains of Influence

Figure 2 depicts the number of times each category in the NIMHD 20-category framework was represented in the studies. There are more than 52 counts in this figure because some articles included more than one category in their investigation resulting in a total of 73 categorizations detailed below.

Figure 2:

Distribution of Studies by Domains and Levels of Influence

Biological

Five studies explored the biological domain, and all focused on the individual level of influence.(38–42) These studies generally found that factors such as age, functional disability, and comorbid conditions such as memory problems and chronic pain were all related to both preferences for types of EOL care(38, 40–42) and advance care planning.(39) For example, younger Black patients with chronic kidney disease were more likely to prefer life-prolonging care compared with whites of the same age groups, but differences were attenuated in older age groups,(38) and markers of increased disease severity such as number of hospitalizations, emergency room visits and prescription medications were associated with increased advance care planning for Black women.(42) All of the studies had a moderate or high risk of bias, including limits on generalizability due to samples from small geographic regions.

Behavioral

Five studies explored the behavioral domain.(39, 40, 43–45) Three studies with low to moderate risk of bias focused on the individual level of influence.(39, 43, 45) Some, but not all, measures of religious practice were associated with decreased advance care planning, and religious practice explained some of the racial disparity in advance care planning.(39, 43) Increased social and physical activity were related to increased advance care planning for white but not Black older individuals.(45) The two studies focusing on the interpersonal level of influence conducted with Black patients with HIV had a moderate to high level of bias.(40, 44) Social support increased the likelihood of nonaggressive EOL treatment preferences,(40) and advance care planning.(44)

Physical/Built Environment

Two studies, both with low risk of bias, examined the physical/built environment.(46, 47) Community resources such as lower quality of available housing and transportation(47) and lower median income(46) were associated with reduced hospice utilization and supply.

Sociocultural Environment

Eighteen studies explored the sociocultural domain; eleven explored the individual level of influence. Several studies with varying risk of bias found that cultural religious beliefs about God’s control over the timing of death (sometimes referred to as fatalism) were more common among Black people (48) and associated with less advance care planning (2, 39, 49) and less of a preference for comfort-oriented EOL care.(50) However, one study with low risk of bias failed to show a relationship between religious affiliation, beliefs and practices with disparities in EOL care.(51) Religious affiliation was examined as an intervention to reduce disparities in a cluster randomized controlled trial in large urban Black churches, which demonstrated improved caregiver knowledge of treatment options and improved self-efficacy for EOL decision-making.(52)

Two studies with low and moderate risk of bias explored sociodemographic factors. One study used the Social Vulnerability Index (SVI), which included socioeconomic status and household composition within the composite of 15 factors, and found that higher social vulnerability was related to reduced use of hospice in Black but not white patients, however, it is unknown which factors within the SVI are driving this outcome.(47) Black people were less likely than white people to engage in advance care planning in one study, but younger age explained more of this difference than household income, household composition and education.(53) In a study of patients with serious illness with moderate risk of bias, mistrust and perceived discrimination explained the lower rates of advance care planning in Black patients compared with white patients to a greater extent than socioeconomic factors.(54) Finally, one study with low risk of bias found that sociodemographic factors such as education, net worth and urban residence attenuated but did not eliminate racial disparities in EOL care spending.(55)

Six studies examined the interpersonal level of influence. Two studies with moderate risk of bias explored the role of family caregivers. Conner et al. showed that the presence of a caregiver increased the likelihood that Black patients would be enrolled in hospice, while Zhang et al. found in a small convenience sample that Black family caregivers of patients with advanced lung cancer were more likely than white caregivers to believe that treatment was curative and to associate hospice care with hopelessness.(56, 57) One randomized controlled trial leveraged the role of family caregivers by offering an intensive advance care planning session involving communication with family and friends to reduce disparities in advance care planning.(58)

Interpersonal interactions with physicians were also explored. A randomized controlled study with low risk of bias revealed that physicians exhibit implicit bias in their nonverbal communication with Black compared with White standardized patients, which could have downstream effects on quality of EOL care.(59) This conclusion is supported by another study with moderate risk of bias, which showed that experience of racial discrimination explained much of the difference between Black and white patients in rates of advance care planning.(54) Finally, a cluster randomized controlled trial with moderate risk of bias showed that social networks in churches could be leveraged to enhance knowledge of EOL treatment options.(52)

Three studies with moderate risk of bias looked at community effects. Home ownership increased the likelihood of advance care planning among Black people but not white people.(53) When controlling for estate planning, which may be a proxy for access to legal services, disparities between Black and white patients in advance care planning were no longer significant.(60) Conner and colleagues found that religious affiliation explained a small portion of the variance in hospice use among Black patients.(57)

A single study with low risk of bias explored the societal level of influence. Fishman and colleagues found that the Black news media was less likely to report stories about cancer that included information about treatment failures, adverse effects or that focused on death and dying compared with the mainstream media.(61)

Health Care Systems

The majority of the studies (35) examined the health care system domain. Ten studies focused on the individual level of influence. Only three of these had a low risk of bias. Several examined patient preferences. One found that personal health values and preferences did not fully explain disparities in advance care planning,(43) while another found that discussion of preferences did not affect disparities in EOL health care spending.(55) Another study with moderate risk of bias found that while there were no significant racial differences in the influence of religious/spiritual beliefs, cultural beliefs and distrust of healthcare on preferences for care at EOL in patients with chronic kidney disease, Black patients who were less than 65 years-old were more likely than similar-age white patients to prefer life-extending care.(38)

Two studies on pharmacologic outcomes found that patient preferences were related to differences in EOL outcomes. One study with moderate risk of bias found that the patient’s preference to focus on cure rather than pain management predicted lower adherence to opioid analgesia regimens among Black patients,(62) and another study with moderate risk of bias showed that EOL treatment preferences explained differences in willingness to use life-prolonging pharmacologic care at EOL even at the cost of feeling worse.(63)

Two studies with moderate risk of bias explored individuals’ health literacy. One showed that trust in healthcare and health literacy did not explain disparities in preferences for life-prolonging pharmacologic care,(63) while another showed that health literacy partially explained the relationship between race and knowledge of cardiopulmonary resuscitation (CPR), which may affect EOL care decisions.(64) Three studies with moderate to high risk of bias explored patient and caregiver knowledge. One study with moderate risk of bias found increased beliefs that hospice must be received in a facility and that it is reserved for those who will only live a few days among Black patients, which could lead to lower uptake of hospice services.(48) One found that Black family caregivers had a lower level of knowledge about dementia compared with white family caregivers of patients with dementia, which could influence outcomes,(49) while another found that awareness that death was imminent decreased in-hospital death in white but not Black patients, suggesting that increased knowledge of prognosis alone will not reduce disparities in location of death.(65)

One study with moderate risk of bias considered individuals’ health care resources and showed that in addition to functional status, having health insurance was associated with increased willingness to consider hospice services.(41)

Ten studies focused on the interpersonal level of influence. Several examined the role of palliative care. One study with moderate risk of bias found that inpatient palliative care consultation decreased racial disparities in advance directive completion.(12) However, two studies revealed shortcomings of communication by palliative care teams. One with moderate risk of bias showed fewer conversations with Black patients included prognostic information compared with those with white patients,(66) while another with low risk of bias showed that palliative care clinicians overestimate survival for Black patients, leading to decreased hospice use.(67)

Communication between clinicians and patients and families were the focus of several studies. Some highlighted the importance of communication from healthcare providers. Even without palliative specialists, one randomized controlled trial with low risk of bias found that having an advance care planning session with a healthcare provider increased advance directive completion among Black patients and reduced disparities.(58) The importance of initiation of advance care planning by health care providers was supported by a study with low risk of bias which showed that perceived norms among family, friends and physicians that set expectations for advance care planning explained variance in intention to engage in advance care planning.(68) In another study with moderate risk of bias, higher quality of communication with health care providers decreased decisional conflict reported by bereaved Black family members.(69)

Other studies did not show positive effects of communication. When looking at health care utilization outcomes, one study with low risk of bias found that quality of health care provider communication did not explain racial differences in hospice or emergency room use at EOL.(70) A study with high risk of bias found that having a health care proxy or EOL discussion increased patient emotional distress as reported by caregiver among Black patients.(71) Health care provider behavior may explain this finding. One randomized controlled trial with low risk of bias showed that physicians exhibit more closed nonverbal communication behaviors when discussing goals of care for a critically and terminally ill Black compared with white standardized patient, which could increase the emotional distress of such communication.(59) On the other hand, one study with low risk of bias found that perceived discrimination and mistrust in healthcare did not explain differences between Black and white patients in advance directive completion.(72)

Sixteen studies examined the community level of influence. Embedding an advance care planning session with a healthcare provider in the community setting increased advance directive completion among Black patients.(58) Other studies looked at the role of community availability of services as a driver of disparities in EOL care. One study with low risk of bias found that counties with increased hospital beds and fewer general practitioners had greater racial disparities in hospice use.(73) Two studies with low risk of bias showed that receipt of care in minority-serving hospitals decreased the odds of receiving palliative care.(74)(75) Palliative care availability was linked to outcomes in a study with high risk of bias which found that palliative care availability in the hospital increased the rate of do-not-resuscitate orders among Black patients.(76)

Two of the sixteen raised issues of equity in quality of hospice care. One with moderate risk of bias showed that the greater the proportion of Black patients served by hospice agencies, the more frequently families reported concerns about quality and coordination of care.(77) Another study with low risk of bias found that Black patients received fewer visits from hospice staff in the last 2 days of life compared with whites.(78)

Several studies looked at practice patterns at the community level. There is some evidence that segregation of minority patients into particular institutions and networks contributes to disparities. One study with low risk of bias found that a measure of healthcare racial segregation correlated with more aggressive EOL care and less hospice use in Black patients, while another study with moderate risk of bias found that larger proportions of Black residents in nursing homes were associated with increased odds of in-hospital death and decreased hospice use.(79)(80) The effect of community practice patterns was inconsistent. A study with low risk of bias did not find a correlation between regional practice patterns and intensity of EOL care for Black patients.(81) However, a study with moderate risk of bias found that receipt of care specifically in a provider-based research network practice increased hospice enrollment among Black patients compared with receipt of care in community practices or academic medical centers.(82) Finally, prevalence of business models within a community may be important. One study with moderate risk of bias found that greater prevalence of for-profit hospices in a region was associated with increased use of hospice among racial minorities.(83)

Four of the sixteen studies, all with low risk of bias, examined both the community and societal level of influence. They found that geographic variation and practice patterns in minority-serving health care institutions contributed to disparities.(84)(85) Findings related to hospice-level variation were mixed. One study found that hospice-level variation in intensity of EOL care did not explain racial differences in hospitalization at EOL,(86) while another study with moderate risk of bias found that Black patients reported higher quality of care within the same hospice compared with whites, but across hospices Black patients reported lower quality suggesting that lower quality hospice organizations explained the overall disparity.(87)

Types of studies

The majority of studies were cross-sectional. There were three prospective cohort studies, six retrospective cohort studies and three randomized controlled trials.

Outcome measures

We searched for studies with outcomes including type of EOL care delivered, processes of care (e.g. palliative care consult), symptom management, advance care planning or patient/family ratings of quality of EOL care. The majority of the studies we found measured advance care planning and type of care delivered.

Study Quality

Most of the studies had some elements that raise concerns about risk of bias. Very few reported blinding procedures, many were underpowered or did not report power analyses, and a substantial number did not adequately control for confounders (Appendix 3 & 4).

Discussion

At this junction it is critical for the palliative care research community to consider how we can move forward to address structural, institutional and interpersonal racism to reduce EOL disparities for Black patients. This review found that researchers have made inroads into addressing structural inequities such as distribution and quality of medical services in communities. In fact, in the domain of Health Care Systems, the attention of researchers has been fairly equally distributed among the individual, interpersonal and community level of influence. This finding is encouraging and contrasts with the finding from the general health disparities literature that much research remains at the individual level of influence. More work remains to be done, however. In particular, the sociocultural environment, physical/built environment, behavioral and biological domains remain understudied.(88) In the healthcare domain, the societal level of influence (e.g. health policy and law) remain understudied. In the sociocultural domain, the majority of the studies still focused on the individual level of influence, missing key areas of research in interpersonal discrimination and local and societal structural discrimination.

In qualitative studies, patients have identified financial issues, health insurance, health systems barriers and communication with providers are key drivers of racial disparities in EOL care. Although beyond the scope of this paper, using the NIMHD framework to systematically review the qualitative literature could highlight overlooked areas in quantitative analyses and identify promising areas for future research. In the sociocultural environment, interpersonal discrimination and individuals’ response to discrimination both within and outside of the health care environment may profoundly affect both the type and quality of EOL care. In particular, the role of racism towards family caregivers and its effect on both EOL decision-making and family functioning during the EOL period should be investigated. In the physical or built environment, factors such as limited transportation options, cramped living spaces or unsafe neighborhoods may inhibit access to palliative care and limit hospice options. In the behavioral domain, national and local policies such as concurrent hospice could be evaluated for their effects on reducing or exacerbating disparities. Finally, most studies of EOL disparities control for biological factors such as age, type of terminal illness and comorbid conditions. However, the enormous racial disparities in COVID-19 illustrated how a community’s experience with a particular disease state may profoundly affect the lived experience at EOL. Furthermore, the burden of illness across the lifespan and earlier age of illness may be important considerations for how communities conceptualize the EOL period.

Studies exploring the individual level of influence tended to be of poorer quality with higher risk of bias, making conclusions about the role of Black patients’ characteristics and beliefs in disparities premature. The majority of the studies included in this review are cross-sectional. This research strategy for examining causal mechanisms is inherently limited in that the Bradford Hill criteria of temporality cannot be established.(89) Future studies should evaluate the effects of interventions in controlled trials on reducing or exacerbating EOL disparities.

Many of these studies focused on intermediate outcomes, such as advance care planning, and proxy outcomes for quality of EOL care such as ICU use, hospitalization and hospice enrollment. Future studies should focus on patient-centered outcomes such as goal-concordant care and patient- and family-reported quality of EOL care and satisfaction.

Fully 179 of the studies retrieved in our search (including 3 by the first author of this paper) were excluded because they merely documented disparities without exploring or addressing causal mechanisms. While it is temptingly easy to analyze large datasets by race and ethnicity, several decades of documentation of disparities has not improved outcomes for Black patients. It is time to move forward and build a research program that can help clinicians and systems reduce disparities.

There are several limitations to this study. First, we limited this search to CINAHL, PubMed and PsychInfo. Other databases, including social work databases might yield more studies in different domains given the different focus of other disciplines. Categorization of studies was sometimes difficult and subjective. We used two independent reviewers to minimize bias in categorization and evaluation of bias.

In sum, this review provides important information regarding the scope of the literature on disparities in EOL care for Black patients and their families. The NIMHD framework shows a spectrum of domains and levels of influence to explore mechanisms for disparities. Researchers can use this framework to paint a more complete picture of causes and address interpersonal, institutional and systemic racism in the future.