Preferences for more aggressive end-of-life pharmacologic care among racial minorities in a large population-based cohort of cancer patients

Abstract

Background:

Minority patients receive more aggressive and potentially suboptimal care at the end of life (EOL). We investigated preferences about pharmacologic interventions at the EOL and their potential variation by socio-demographic factors among recently diagnosed cancer patients.

Methods:

A population-based cross-sectional survey of cancer patients identified through the Texas Cancer registry was conducted using a multi-scale inventory between March 2018 and June 2020. Item responses to questions about potential pharmacologic interventions at the EOL were the focus of this investigation. Inverse probability weighted multivariate analysis examined associations of socio-demographic characteristics, health literacy, and trust in medical professionals with pharmacologic preferences.

Results:

Of the 1480 included responses, 13.3% stated they would take a medication that may prolong life at the cost of feeling worse. Adjusted analyses showed Black or Hispanic race/ethnicity, living with another person, and having a higher trust score were more likely to express this preference. In contrast, 41–65y (vs. 21–40y), living in a rural area, and adequate or unknown health literacy were less likely to express this preference. Overall 16% of respondents were opposed to potentially life shortening palliative drugs. In adjusted analysis Black or Hispanic respondents were more likely to be opposed to potentially life shortening drugs while age 65–79 and ≥ college education were associated with a decreased likelihood of opposition to this item.

Conclusions:

Black and Hispanic cancer patients were more likely to express preferences toward more aggressive EOL pharmacologic care. These findings were independent of other sociodemographic characteristics, health literacy and trust in the medical profession.

INTRODUCTION

Published studies show that racial/ethnic minority patients and Medicaid enrollees are more likely to receive aggressive end-of-life (EOL) care.^1–4 This is also true specifically for patients dying of cancer.^5,6 End-of-life (EOL) quality care measures commonly regard aggressive EOL care as low quality EOL care, further defined as aggressive therapy in the final 30 days of life and/or failure to enroll in hospice.^7–9 However, high quality EOL care most importantly adheres to patients preferences, while aiming to avoid aggressive health care interventions such as chemotherapy in the last month of life and includes early enrollment in hospice services.^10–13 These measures help facilitate care that is consistent with the National Academy of Medicine’s (formerly the Institute of Medicine) definition of “good death”.¹⁴ However, that definition does not incorporate patient preferences regarding EOL care aggressiveness. Eliciting and adhering to patient EOL preferences remains a challenge for medical professionals, but is a key component of effective EOL care.^1,15

Little is known about whether higher levels of EOL care intensity is commiserate with patient preferences or whether some other factor such as access health literacy^16,17 or medical mistrust may influence EOL care aggressiveness among specific groups of patients. Access to hospice care may also play a role.¹⁸ In an effort to ascertain whether differential access to hospice benefits might explain some variation in EOL care, we previously investigated whether managed care versus fee-for-service payments influenced EOL care among Texas Medicare beneficiaries and showed that hospice use did increase with elimination of differences between the two payer models over time. However, even accounting for insurance plan and other potential confounding factors, Black and Hispanic cancer patients continued to have persistent lower hospice utilization compared to non-Hispanic whites.¹⁹ Texas is comprised of a racially and socio-economically diverse population with close resemblance to future national demographic projections.^20,21 We sought to investigate patient preferences regarding pharmacologic interventions at the end of life in a large cohort of Texans recently diagnosed with cancer. We conducted a population-based survey study to determine possible associations between preferences and socio-demographic factors, health literacy, and general trust in medical professionals.

METHODS

Data Source, study sample and data collection procedure

This survey study is part (Project 4) of the larger multi-year Comparative Effectiveness Research on Cancer in Texas (CERCIT) program, funded by the Cancer Prevention and Research Institute of Texas (CPRIT) (https://www.utmb.edu/scoa/research/supported-research-programs/comparative-effectiveness-research-on-cancer-in-texas/current-projects).²² We recruited participants from the statewide population-based Texas Cancer Registry (TCR) ((https://www.dshs.state.tx.us/tcr/).²³ The TCR is the 4th largest cancer registry in the US that meets the rigorous data standards of the National Program of the Central Cancer Registries and the Centers for Disease Control. The registry is also Gold Certified by the North American Association of Central Cancer Registries. A sample of 6222 TCR registrants over 18 years of age or older with a diagnosis of a solid malignancy in the past 12 months were obtained. We excluded those individuals with hematologic malignancies because their cancer treatment is commonly medically intensive and physically burdensome. We surmised that response rate would be concomitantly low among this group and we wanted to minimize contact of individuals who may be severely ill. We chose to limit our sample to those within 12 months of diagnosis because we had limited resources and could only perform one timepoint of contact. By limiting to a timepoint early after diagnosis, we endeavored to obtain a better response rate as this sample was obtained from a population-based sample whose data were captured at diagnosis and not updated thereafter. Of these, 5335 were determined to be alive and able to receive study questionnaires by mail delivery. A $10 (US) retail gift card was mailed with the study questionnaire packet and individuals contacted were invited to keep the gift card whether they opted to return the study questionnaire or not. In total, 1566 individuals responded between March 2018 and July 2020, yielding a response rate of 27.9%. Of these respondents, 1480 (94.5%) were included for this analysis as described below. Attempts were made to follow up by mail and/or telephone for all non-responders. We performed telephone follow-up early in the data collection time period, however yield was very low (<10%) for phone contact. For the first 2300 potential participants, we sent 3 follow-up reminders and two replacement study questionnaires at t 2 weeks, 4–6 weeks, and 8–10 weeks to non-respondents. During the course of this study’s data collection efforts, the TCR/Texas Department of State Health Services changed their policy regarding the number of allowable attempted contacts to registrants and we limited contacts accordingly to a single follow-up reminder at 8–10 weeks after first mailing to those in the remaining sample that did not respond to the first mailing. The subsequent response rates were approximately 25% with either 1 versus 3 mail out attempts. Potential participants’ primary care physicians were notified of the study to ensure that there were no medical objections regarding participation; none objected. Due to resource constraints, questionnaires were available only in English. Individuals identified with Spanish surnames received an additional recruitment letter written in Spanish advising them to have an English-speaking family member to assist with translation of questionnaires. All study procedures were approved by both the Texas Department of State Health Services and MD Anderson institutional review boards.

Study measures

Study questionnaires included several validated measures. EOL care preferences were evaluated using a previously published and validated instrument that generates 7 dichotomous outcome variables related to EOL patient preferences.^24,25 For this is analysis, we focused preferences regarding potential EOL pharmacological interventions using two items whose stem presented a hypothetical scenario of having an illness with certainty of < 1 year to live. Possible response to both items were: Yes, No, Don’t know. One of these items asked: “To deal with that illness, do you think you would want drugs that would make you feel worse all the time but might prolong your life?” A “Yes” response was categorized as “for life prolonging drug”. The other item asked: “If you reached the point at which you were feeling bad all the time, would you want drugs that would make you feel better, even if they might shorten your life?” A “No” response was categorized as “Opposed to palliative drug that might shorten life”. We sought to keep the directionality of the responses similar in attempt to make it easier to identify preferences for more aggressive care vs. less aggressive care. The Trust in Medical Profession scale is a previously validated instrument that includes 11 trust related statements that are assessed via a 5-level Likert scale. Aggregate scores determine trust levels in physicians generally, not trust in a respondent’s specific individual physician, and reflect four domains of general trust (fidelity, competence, honesty, and global trust). The scale has shown high reliability (Cronbach’s alpha, 0.89) and response variability (range 11–54, mean 33.5, standard deviation 6.9) and a higher numerical score corresponds to higher levels of trust.²⁶ Per the design of the Trust in the Medical Profession Scale respondents who completed this component of the survey with 0–2 missing answers were considered for this analysis while those with 3 or missing values were excluded. Health literacy was assessed using the Newest Vital Sign (NVS), a 6-item health literacy assessment structured around reading and understanding information on a nutrition label. The NVS has been validated with high reliability and validity against the Rapid Assessment of health Literacy in Medicine (REALM) and the Short Test of Functional Literacy in Adults (STOFLHA) and has a high sensitivity for ascertainment of health literacy.^27,28 The scale classifies respondents’ high likelihood of limited health literacy (score: 0–1), possibility of limited health literacy (score: 2–3), and adequate health literacy (score: 4–6). For the purposes of this study we condensed this to two categories: adequate health literacy (score ≥4) or limited health literacy (score < 4) and those that did not fill out the scale were classified as “unknown”. Self-reported health was collected using the Medicare Health Outcomes Survey,²⁹ which asks respondents to rate their health on a 5-point scale from poor to excellent.

Socio-demographic and cancer clinical characteristics were obtained from a combination of TCR registry data and self-reported information. Race/ethnicity was self-reported or, in cases of missing self-reported values (69 of 1480), obtained from TCR registry. Only 69 of 1480 respondents (5%) did not self-report race/ethnicity and were thus assigned by the TCR data. Agreement between TCR race/ethnicity data for the remaining 1411 was high with 94% agreement on ethnicity and 98% agreement on race. Race and ethnicity groups are non-Hispanic white, non-Hispanic Black, and Hispanic. The “non-Hispanic” designation is dropped hereafter in the manuscript for simplicity. Race/ethnicity were one of the fundamental demographic variables included in our study in an attempt to better understand significant EOL care variation that we had previously observed among minorities dying of cancer in Texas.⁵ Additional items included: age, gender, primary language spoken in the home, income, marital status, whether the individual lived alone or with others, and highest-attained education level. Gender was assessed using an item included on our survey asking respondents the question “Are you male or female?” with response options of “male” or “female”. Rurality and Texas Health Service Region were derived by zip code at the time of diagnosis of cancer, with rurality being defined per the US Department of Agriculture 2013 Rural/Urban Continuum Codes with scores of 1–3 yielding an urban designation and 4–9 being considered as rural.³⁰ Finally, Cancer type and stage at diagnosis were obtained from the TCR database.

Statistical Analysis

Descriptive statistics were generated for demographic and socio-economic characteristics of respondents. In order to address potential selection bias from our findings observed in this convenience sample, we performed an inverse probability weighting analysis.^31,32 For each observed case (n=1480), we computed the probability of return of the survey from the pool of 5535 eligible cases. We determined the weights based upon five available factors: age, gender, race/ethnicity, cancer site, health services area in Texas. The inverse probability of response was derived to weight each observation attempting to balance the selection bias due to non-response. Finally, normalized inversed probability (inversed probability divided by the mean) was used in the final weighting analyses. Normalized inverse probability was used to reduce the weight loading because weights can increase the standard errors of estimates and introduce instability in the data. Significance of differences between responses to items with regard to socio-demographic characteristics were assessed using a Rao-Scott Chi-quare test for the weighted data.³³ Logistic regression analysis was conducted to examine the significance of demographic, socio-economic factors, health literacy and medical professional trust score on respondents’ preferences. All factors were included in the multivariable model, and odds ratios (OR), and 95% confidence intervals (CI) were reported for each covariate, a p-value of <0.05 was deemed to be statistically significant. Finally, responses to the items were assessed using a one-way analysis of variance (ANOVA) test to examine the association of EOL drug intervention preferences with medical professional trust using the Trust in the Medical Profession aggregate scores of each respondent. Trust score was included in the multivariable models as a continuous variable. Data analyses were performed using SAS (version 9.4 SAS Institute Inc., Cary, NC, USA).

RESULTS

Of the 1566 respondents, 1494 (95.4%) completed the EOL care scale, and 1480 completed the Trust in Medical the Profession Scale. Unweighted and weighted respondent characteristics are displayed in Table 1. In the unweighted sample, a majority of respondents were female (66.2%), age 41–64 (51%), married (65.7%), college-educated (72.9%), and reside within an urban location (86%). A total of 71% of respondents were white compared to 11% Black and 15% Hispanic. This is notably different from the racial make-up of Texas which is 41% non-Hispanic white, 40% Hispanic and 13% non-Hispanic Black.²⁰ A majority of respondents reported an income level of $40,000 – $99,999 (32%) which is well above the reported median per capita income for Texans which is $30,143.²⁰ Most respondents (41.7%) reported having “good” self-reported health. The most common primary disease sites were breast (39.6%) and colorectal (15.4%), with 58.5% of respondents having cancer localized to the primary site compared to 23.1% who had regional extension of their tumor and 7.3% with distant spread of their disease.

Table 1 –

Patient Demographics

	Number	Raw %	Weighted %
Total cases	1480	100.0	100.0
Gender
Female	979	66.2	60.4
Male	501	33.9	39.6
Age
21–40	101	6.8	7.4
41–64	754	51.0	50.5
65–79	556	37.6	36.7
80+	69	4.7	5.5
Race/Ethnicity
White non-Hispanic	1054	71.2	61.1
Black non-Hispanic	148	10.0	13.4
Hispanic	233	15.7	20.2
Others	45	3.0	5.3
Education
Some high school or less	117	7.9	9.9
High school	257	17.4	19.4
Some college and above	1079	72.9	68.5
UNK	27	1.8	2.3
Income
Less than $19,999	193	13.0	15.0
$20,000 – $39,999	184	12.4	13.4
$40,000 – $99,999	475	32.1	31.2
$100,000 or more	393	26.6	24.0
UNK	235	15.9	16.4
Self-reported health
Excellent	484	32.7	30.9
Good	617	41.7	41.1
Fair	295	19.9	21.7
Poor	78	5.3	5.9
Marital status
Married	972	65.7	64.8
Not Married	484	32.7	33.1
UNK	24	1.6	2.1
Living arrangements
Alone	256	17.3	16.8
With someone	1207	81.6	81.8
UNK	17	1.2	1.4
Language at home
English	1327	89.7	86.0
Others	101	6.8	9.6
UNK	52	3.5	4.5
Rural/Urban -
Urban	1274	86.1	85.9
Rural	206	13.9	14.1
Health literacy
Limited literacy(<=4)	449	30.3	33.0
Adequate literacy(>4)	895	60.5	56.9
UNK	136	9.2	10.1

NH = Non-Hispanic, HS = High School, UNK = Unknown

Among respondents, 13.3% noted that they would take a medication that may prolong life at the cost of feeling worse. Distribution of respondent answers by socio-demographic factors are shown in Table 2. On univariate analysis ethnicity, educational attainment, income level, living situation, language spoken at home, rurality, and health literacy were statistically significant. A weighted analysis of trust in the medical profession is displayed in Table 3 and shows that individuals who answered yes to this item had a significantly higher mean trust score 40.2 vs those who responded no 37.3 (p <0.001). On multivariable analysis (Table 4) respondents who were Black OR= 3.60 (95% CI; 2.16–5.98, p<0.001) or Hispanic OR =1.98 (95% CI; 1.17–3.36, p=0.011), living with another person (vs. alone) OR=1.99 (95% CI: 1.01–3.91, p=0.045), and having a higher trust score OR 1.03 (95% CI; 1.01–1.06, p=0.019) were independently associated with an increased likelihood of preference for taking a life prolonging medication. In contrast, age 41–65y (vs. 21–40y) OR=0.44 (95%CI; 0.24–0.84, p=0.012), age 65–79y (vs. 21–40y) OR=0.44 (95%CI; 0.22–0.89, p=0.022), rurality OR 0.36 (95% CI; 0.17–0.74, p=0.005) and adequate health literacy OR=0.57 (95% CI; 0.37–0.88, p=0.012) or unknown health literacy OR=0.51 (95% CI; 0.27–1.00, p=0.049) (versus limited health literacy) were significantly associated with a decreased likelihood of preference for life prolonging drug at the cost of feeling worse.

Table 2 –

Respondent Answers (weighted) to concerns about potential amount of EOL of care by socio-economic factors

	Too Little Care (%)	Rao-Scott p-value	Too Much Care (%)	Rao-Scott p-value	Any Worry (%)	Rao-Scott p-value
Total cases	32.5		20.0		52.5
Gender		0.134		0.133		0.011
Female	34.2		21.3		55.6
Male	29.8		17.9		47.7
Age		0.021		0.735		0.126
21–40	39.9		18.9		58.8
41–64	35.5		19.4		55.0
65+	26.8		21.4		48.2
80+	32.9		16.6		49.5
Race/Ethnicity		0.131		0.001		0.027
White NH	30.7		24.0		54.8
Black NH	35.8		10.3		46.2
Hispanic	37.9		16.6		54.2
Other	23.6		11.5		35.1
Education		0.174		0.005		0.744
≤ Some HS	41.3		13.1		54.4
HS	36.0		13.2		49.2
College or >	30.4		22.8		53.2
Income		0.016		0.003		0.105
≤ $19.9K	41.2		13.2		54.4
$20K– $39.9K	38.0		14.9		52.9
$40K-$99.9K	29.5		21.7		51.2
≥ $100,000	31.6		26.0		57.6
UNK	26.9		18.1		45.0
Marital status		0.227		0.105		0.744
Married	30.4		21.6		52.0
Not Married	36.4		16,4		52.9
Living Situation		0.519		0.366		0.920
Alone	33.8		17.4		51.2
With someone	32.4		20.3		52.7
Home Language		0.558		0.571		0.215
English	33.1		20.1		53.2
Others	30.4		21.4		51.8
Rural/Urban		0.406		0.136		0.819
Urban	32.0		20.6		52.6
Rural	35.5		16.1		51.6
Self-Reported Health		0.544		0.988		0.550
Excellent	29.4		20.2		49.7
Good	33.4		19.9		53.2
Fair	33.4		19.5		52.9
Poor	38.7		21.4		60.1
Health Literacy		0.005		0.002		<0.001
Limited (≤4)	34.7		16.1		50.7
Adequate (>4)	33.5		23.3		56.8
UNK	19.4		14.1		33.5

N= Number, R= Raw Percentage, W= Weight Percentage, NH = Non-Hispanic, HS = High School, K = Thousand Dollars UNK = Unknown

Table 3 –

Weighted Multivariable Analyses of Patient Perception of Appropriateness of End of Life Care

Multivariable Analysis
Factor	Reference	Too Little			Too Much			Any Worry
		OR	(95% CI)	p-value	OR	(95% CI)	p-value	OR	(95% CI)	p-value
Gender
Male	Female	0.83	(0.63–1.09)	0.187	0.83	(0.61–1.11)	0.209	0.76	(0.59–0.98)	0.031
Age
41–64	21–40	0.85	(0.52–1.39)	0.512	0.95	(0.54–1.68)	0.858	0.82	(0.50–1.35)	0.430
65–79		0.62	(0.37–1.05)	0.076	1.19	(0.66–2.16)	0.557	0.73	(0.43–1.22)	0.226
80+		0.86	(0.41–1.79)	0.677	0.97	(0.40–2.37)	0.948	0.82	(0.38–1.75)	0.609
Race/Ethnicity
Black NH	White NH	1.19	(0.79–1.79)	0.404	0.45	(0.24–0.84)	0.012	0.76	(0.51–1.12)	0.165
Hispanic		1.45	(0.98–2.16)	0.065	0.67	(0.39–1.14)	0.136	1.08	(0.71–1.62)	0.727
Other		0.90	(0.41–1.99)	0.796	0.32	(0.11–0.89)	0.030	0.47	(0.23–0.97)	0.042
Education
HS	< Some HS	0.69	(0.39–1.20)	0.185	0.94	(0.44–2.02)	0.872	0.66	(0.38–1.15)	0.139
College or >		0.52	(0.31–0.88)	0.015	1.46	(0.72–2.92)	0.292	0.67	(0.39–1.13)	0.130
UNK		0.63	(0.16–2.49)	0.507	1.01	(0.17–5.90)	0.991	0.60	(0.17–2.15)	0.431
Income
$20K– $39.9K	< $19.9K	0.85	(0.52–1.38)	0.505	1.02	(0.51–2.05)	0.946	0.84	(0.52–1.36)	0.483
$40K-$99.9K		0.62	(0.40–0.95)	0.030	1.31	(0.73–2.36)	0.373	0.74	(0.48–1.13)	0.158
≥ $100,000		0.72	(0.44–1.18)	0.197	1.64	(0.88–3.06)	0.117	1.02	(0.62–1.66)	0.950
UNK		0.57	(0.35–0.92)	0.022	1.11	(0.57–2.14)	0.762	0.62	(0.39–0.99)	0.043
Marital status
Not Married	Married	1.20	(0.85–1.68)	0.301	0.81	(0.52–1.27)	0.353	1.03	(0.74–1.44)	0.848
UNK		1.88	(0.51–6.96)	0.345	1.34	(0.34–5.30)	0.681	2.36	(0.60–9.37)	0.221
Living Situation
With someone	Alone	1.15	(0.76–1.75)	0.514	0.98	(0.58–1.66)	0.936	1.15	(0.77–1.70)	0.500
UNK		0.80	(0.13–5.19)	0.818	3.07	(0.58–16.3)	0.198	1.78	(0.42–7.49)	0.431
Home Language
Others	English	0.56	(0.30–1.04)	0.065	2.13	(1.06–4.32)	0.035	0.94	(0.53–1.67)	0.837
UNK		0.59	(0.25–1.40)	0.231	0.66	(0.21–2.04)	0.470	0.51	(0.23–1.13)	0.096
Rural/Urban
Rural	Urban	1.20	(0.83–1.74)	0.333	0.73	(0.48–1.13)	0.155	0.97	(0.69–1.36)	0.855
Self-Reported Health
Good	Excellent	1.09	(0.81–1.46)	0.582	1.10	(0.80–1.53)	0.554	1.15	(0.87–1.52)	0.329
Fair		0.48	(0.68–1.43)	0.934	1.17	(0.78–1.75)	0.452	1.10	(0.77–1.56)	0.613
Poor		1.01	(0.52–1.96)	0.967	1.45	(0.74–2.83)	0.276	1.32	(0.72–2.43)	0.374
Health Literacy	Limited (≤4)
Adequate (>4)		1.09	(0.80–1.49)	0.595	1.20	(0.84–1.70)	0.320	1.19	(0.89–1.60)	0.238
UNK		0.57	(0.29–0.80)	0.002	0.97	(0.55–1.72)	0.928	0.56	(0.36–0.88)	0.012
Trust Score	Continuous	0.97	(0.96–0.99)	0.002	0.98	(0.96–1.00)	0.021	0.96	(0.95–0.98)	<.001

OR = Odds Ratio, CI = Confidence Interval, NH = Non-Hispanic, HS = High School, K= Thousand Dollars, UNK = Unknown

Table 4 –

Weighted analyses of Trust score vs Primary Outcome Measures

	Weighted Trust score
	Mean	SE	95% CI	p-value
Total cases	37.7	0.21	(37.3 – 38.1)	-
Too little
Difference (No-Yes)	1.5	0.48	(0.6 – 2.5)	0.002
No	38.2	0.28	(37.7 – 38.7)
Yes	36.7	0.39	(35.9 – 37.4)
Too much
Difference (No-Yes)	1.8	0.58	(0.7 – 3.0)	0.002
No	38.1	0.25	(37.6 – 38.5)
Yes	36.2	0.52	(35.3 – 37.1)
Any worry
Difference (No-Yes)	2.5	0.45	(1.6 – 3.4)	<0.001
No	39.0	0.33	(38.4 – 39.6)
Yes	36.5	0.31	(35.9 – 37.1)

SE = Standard Error, CI = Confidence Interval

Only 16% of respondents were opposed to potentially life shortening palliative drug intervention at the end of life. As shown in Table 2, race/ethnicity, education level, income, marital status, language spoken at home, and health literacy level were all significantly associated with response to this item on univariate analyses. On multivariable analysis, only Black OR=3.07 (95% CI: 1.96–4.81, p<0.001) or Hispanic OR=1.74 (95% CI:1.07–2.83, p=0.025) race/ethnicity displayed an independent significant increased likelihood of being opposed to palliative drugs that might shorten life. Age 65–79 years OR=0.5 (95% CI:0.26–0.98, p=0.043) and higher education level of college and above OR=0.55 (95% CI: 0.31–0.97, p=0.038) were associated with decreased likelihood of being opposed to palliative, potentially life shortening drugs. Trust in the medical profession was not significantly associated with this item on either unadjusted (Table 3) or adjusted analyses (Table 4).

DISCUSSION

We observed that Black and Hispanic individuals recently diagnosed with cancer expressed more favorable attitudes regarding more aggressive pharmacologic interventions at the EOL and that these preferences were independent of health literacy and trust in medical professionals. Correspondingly, Black and Hispanic patients were also more likely to be opposed to potentially life-shortening palliative pharmacologic interventions despite adjusting for sociodemographic characteristics, trust in medical professionals, and health literacy. Our findings are important and novel in that they show that preferences for more aggressive care were not driven by issues of medical mistrust or health literacy. In fact, higher trust in medical professionals score was independently associated with higher likelihood of preference toward life-prolonging medications as was residing with others (versus alone). Additionally, we identified some characteristics that were independently associated with preferences for less aggressive EOL drug interventions such as some age subgroups, residence in a rural area, and having adequate health literacy. Also, more educated individuals were less likely to be opposed to palliative, potentially life shortening drug interventions at the end of life.

Previously published studies have also shown that racial/ethnic minority patients more commonly express preferences towards aggressive EOL care in general as well as with respect to drug interventions. Barnato and colleagues,²⁴ who developed the EOL preferences instrument utilized in our study, showed similar results among a large national sample of Medicare beneficiaries where they also found a higher proportion of Black (28%) and Hispanic (21.2%) patients preferring life-prolonging drugs that might make them feel worse compared to white patients (15%). Additionally, in their study both Black and Hispanic patients displayed a statistically significant decreased preference to take life-shortening palliative medications.²⁴ Other investigators have reported that Black and Hispanic patients prefer more life sustaining treatments in general at the end of life when compared to whites.^34–36 Furthermore, Martin and colleagues³⁷ showed that Black patients were significantly more likely than other racial/ethnic groups to be willing to exhaust personal financial resources for life-prolonging care. Our study confirmed some of these findings but adds additional important insight by demonstrating that these preferences remain consistent even when adjusting for trust in medical professionals and health literacy.

Initially, we had hypothesized that medical mistrust, well-earned among minority populations due to decades of structural racism in health care, may play a role in EOL care preferences as other published studies had shown that preferences toward life sustaining treatment among minority patients were associated with distrust of the health care system or fear that health care was being given based on ability to pay.⁴ However, not only did our study show that more aggressive preferences were positively associated with higher trust in medical professionals, but, a recent analyses completed by our group found that it was non-Hispanic white cancer patients in Texas that displayed lower trust in medical professionals.³⁸ Several studies have identified that socio-economic factors particularly education level, income, and health literacy have been implicated as contributing to receipt of disparately aggressive EOL care potentially related to decreased participation in advanced care planning.^16,17,39 These structural and health system barriers contribute to an inherent institutional racism that facilitates inequitable EOL care for minority patients. Other research shows that minority patients are more likely to not understand that chemotherapy was unlikely to cure their metastatic cancer. However, studies reveal not only that health literacy impacts EOL care preferences and planning, but also that EOL care clinician communication was less likely to occur or was less effective for minority patients.^34,40,41 These phenomena are undoubtedly related and deserve further effort to understand and remedy. Our attempts to ascertain whether health literacy influenced EOL preferences in our cohort did reveal that adequate health literacy was independently associated with lower likelihood of preference for life prolonging drugs at the EOL, but the finding of more aggressive preferences among minority cancer patients persisted when adjusting for health literacy.

Improving EOL care for all cancer patients remains a matter of some urgency as the population ages, both from a quality of care standpoint and importance of caring for patients toward “a good death”,¹⁴ but also due to the need to mitigate the unsustainably high health care expenditures associated with suboptimal quality EOL care.¹¹ A recent analysis of over 1250,000 Medicare beneficiaries compared the rates of life-sustaining measures completed at cancer centers and reported as the concentration of minority patients increased for respective centers, the EOL care quality decreased as measured by increased emergency room visits and ICU admissions, lack of hospice referrals and more life-sustaining treatments.⁴² Other studies have shown similar patterns of aggressive EOL care among minority Medicaid beneficiaries.^5,6 Understanding EOL preferences and their underlying socio-demographic and behavioral health associations may be key to tailoring interventions aimed at ensuring that optimal EOL care is achieved among diverse populations. While we did not show that health literacy or trust in medical professionals attenuated the increased likelihood of preference for aggressive EOL care drug interventions among racial/ethnic minorities, it is possible that there are unmeasured factors not included in our study that could also explain this finding. For instance , spirituality and religious beliefs may be a driving factor behind pursuit of aggressive EOL measures,^43,44 and this was not measured in our study. Similarly, socio-cultural and family dynamics may be barriers to optimal EOL care in minority patients. Hispanic patients in particular have been identified as an ethnic group who are resistant to appointing health care proxies and participating in advanced care planning due to the high priority of “familism” and desire to not place additional burden on their loved ones.^45,46 Finally, inadequate hospice access for minority patients presents another likely contributor to inequities in EOL care and this was also not specifically measured in our study.^44,47

Data regarding preferences about EOL care may serve to help inform efforts to improve EOL care quality among diverse groups of patients. Improving patient-clinician relationships via better availability of minority medical professionals, same-language services, and overall cultural congruency of an institution within a community can positively impact the likelihood of patients pursuing hospice enrollment.^48,49 Community outreach programs to minority neighborhoods developed with community engagement, expanding social work and patient navigator services geared towards minority patients, and development of hospice and advanced care planning information responsive to cultural differences require further effort and funding support.^50–52 Additionally healthcare professionals can implement shared decision making into their practices. Data shows that shared-decision making efforts aimed at vulnerable populations affects significant improvements in patient knowledge, informed decision making, patient physician collaboration, lower physician conflict, and increased decisional self-efficacy.⁵³ All of these interventions are facilitated by data like ours that provide insights into EOL care preferences and their associated patient characteristics.

In addition to the aforementioned potential unmeasured factors in our study, there are several relevant limitations of this study that we must acknowledge. Given that this study is a mail-in survey, it is subject to several common limitations including non-response bias, low response rate, questionable comprehension capabilities for individuals with low health literacy or those who are non-English speakers, and the inability to question respondents for further information/rationale for their responses.⁵⁴ Unfortunately due to resource limitations our study was unable to provide a Spanish version of this survey thus explaining the under-representative of non-English speaking Hispanic patients in Texas which undoubtedly contributed to our low response rate.²⁰ There was significant overrepresentation of non-Hispanic respondents relative to the state population which impacts the generalizability of this data. It is possible that by surveying individuals early after their cancer diagnosis, we do not accurately capture responses that are reflective of actual EOL care preferences and choices that individuals may make when faced with the reality of terminal illness versus a hypothetical. Additionally, the instrument used to assess trust in our study was previously validated in a predominantly non-Hispanic white cohort which may limit its ability to accurately identify medical mistrust within this population. However, this instrument was selected due to the fact it did not directly reference potential racial/ethnic concerns in attempt to not alienate potential respondents of any background. Despite these limitations, considering the large size of this study and incorporation of trust and health literacy as well as socio-demographic characteristics, we feel that it contributes to our understanding of EOL care preferences and concerns among recently diagnosed cancer patients.

In conclusion, non-white racial/ethnic minorities recently diagnosed with cancer expressed generally more favorable preferences toward potential life-prolonging drugs at the end of life and were more opposed to palliative drug interventions that might shorten life and these findings were independent of health literacy, trust in medical professionals and other socio-demographic factors. These findings may be used to inform practice or policy changes to improve quality of end-of-life care specifically for minority patients.

KEY MESSAGE.

This article describes a population-based cross-sectional survey conducted in a large cohort of Texas cancer patients. Our findings show ethnic/racial minority patients have an increased preference for aggressive pharmacologic EOL care. Further work is needed to address underlying factors influencing these preferences to improve EOL care for minority populations.