Association of Neighborhood Socioeconomic Status With Withdrawal of Life-Sustaining Therapies After Intracerebral Hemorrhage

Abstract

Background and Objectives

Mortality after intracerebral hemorrhage (ICH) is common. Neighborhood socioeconomic status (nSES) is an important social determinant of health (SDoH) that can affect clinical outcome. We hypothesize that SDoH, including nSES, contribute to differences in withdrawal of life-sustaining therapies (WLSTs) and mortality in patients with ICH.

Methods

We performed a retrospective study of patients with ICH at 3 tertiary care hospitals between January 2017 and December 2022 identified through the Get with the Guidelines Database. We collected data on age, clinical severity, race/ethnicity, median household income, insurance, marital status, religion, mortality before discharge, and WLST from the electronic medical record. We assessed for associations between SDoH and WLST, mortality, and poor discharge mRS using Mann-Whitney U tests and χ² tests. We performed multivariable analysis using backward stepwise logistic regression.

Results

We identified 868 patients (median age 67 [interquartile range (IQR) 55–78] years; 43% female) with ICH. Of them, 16% were Black non-Hispanic, 17% were Asian, and 15% were of Hispanic ethnicity; 50% were on Medicare and 22% on Medicaid, and the median (IQR) household income was $81,857 ($58,669–$122,078). Mortality occurred in 17% of patients, and of them, 84% of patients had WLST. Patients from zip codes with higher median household incomes had higher incidence of WLST and mortality (p < 0.01). Black non-Hispanic race was associated with lower WLST and discharge mortality (p ≤ 0.01 for both). In multivariable analysis adjusting for age and clinical severity scores, patients who lived in zip codes with high-income levels were more likely to have WLST (adjusted odds ratio [aOR] 1.88; 95% CI 1.29–2.74) and mortality before discharge (aOR 1.5; 95% CI 1.06–2.13).

Discussion

SDoH, including nSES, are associated with WLST after ICH. This has important implications for the care and management of patients with ICH.

Introduction

Mortality after intracerebral hemorrhage (ICH) is common, occurring in 30%–40% of patients.^1,2 Withdrawal of life-sustaining therapy (WLST) is independently associated with mortality after ICH.² Social determinants of health (SDoH), such as area of residence,^3,4 affect health outcomes in general and, particularly, outcomes in cardiovascular and neurovascular disease.^5-10 For example, neighborhood socioeconomic status (nSES) is associated with mortality in patients hospitalized for neurologic disease.^11,12 Unfortunately, outcome disparities based on SDoH seem to be worsening.¹³ We sought to determine how WLST after ICH is affected by SDoH and how this contributes to mortality before discharge.

Methods

We performed a retrospective cross-sectional study of patients with ICH at 3 urban tertiary care hospitals in Brooklyn, Manhattan, and Long Island between January 2017 and December 2022 identified through the Get with the Guidelines (GWTG) database. We identified independent variables from data prospectively collected for the GWTG database and through automated electronic medical record (EMR) data retrieval. This study was approved by the NYU Institutional Review Board, protocol s19-01144. Written consent was not obtained from patients in this study because the protocol was exempt because of its retrospective nature.

Social Determinants of Health

SDoH were chosen based on currently available variables in our EMR. Information regarding race/ethnicity, zip code, occupation, marital status, religion, and insurance was obtained from the EMR. Race and ethnicity were either self-reported on admission or determined by hospital staff when self-reporting was not possible. Marital status, religion, and occupation, when available, were self-reported. A dummy variable of companion was created for marital status, defined as patients who reported they are married, have a significant other, or a partner.

Zip codes were linked to median household income from the 2021 US Census American Community Survey 2021 5-year estimate, which was the most current, comparable census data available at the time of analysis and are reported in 2021 inflation-adjusted dollars.^14,15 Incomes were divided in 4 quartiles, ≤$58,669, $58,670–$81,857, $81,858–$121,688, and >$121,688. We also assessed the percentage of high-income households within a zip code, defined as households with incomes in the top 10% of all US households (≥$200,000).¹⁵ We further stratified nSES as a binary variable of the median household income by zip code less/greater than the national median household income of $70,784 in 2021.¹⁴ The median household income in New York State was $75,157 in 2021.¹⁴

Insurance was recorded as Medicare, Medicaid, dual, private, or no insurance/self-pay. Dual-eligible beneficiaries, a group with unique health disparities,¹⁶ was coded as its own group.

Outcomes

Our primary outcomes were WLST and discharge mortality. Charts were manually reviewed for WLST, which was defined by either documentation of discussion and agreement by patient surrogate to stop all life-sustaining therapy, palliative extubation, and/or admission to hospice. Neurology-trained neurointensivists conducted goals-of-care conversations for all intubated patients, or as clinically relevant, at each hospital site. Mortality included in-hospital deaths and patients discharged to hospice.

Our secondary outcomes were functional outcome on discharge and discharge disposition. We defined functional outcome on discharge by discharge modified Rankin Scale (mRS) score as documented in the discharge summary by a board-certified neurologist, dichotomizing poor discharge mRS as mRS 4–6. Discharge disposition included “home, with or without services,” “acute inpatient rehabilitation,” “subacute rehabilitation (SAR) or skilled nursing facility (SNF),” “long-term acute care hospital (LTACH),” or “transfer to another acute care hospital.” Patients who “left against medical advice” were counted as a home disposition. We also looked at discharge as a dichotomized variable defined as good (discharge to home, acute rehabilitation or transfer to another acute care hospital) or poor (discharge to SAR, SNF, LTACH, hospice, or with mortality).

Other Covariates

Comorbidities including medical history of hypertension, diabetes, and coronary artery disease were extracted from the medical record. The NIH Stroke Scale (NIHSS)¹⁷ and ICH Score¹⁸ were used as markers of clinical severity on admission. Hospital length of stay (LOS) and days on a ventilator were used as markers of illness severity.

Statistical Analysis

We summarized the characteristics of the study population using frequencies and percentages for categorical variables and medians with interquartile ranges (IQRs) for continuous variables. Missing variables were dealt with using population median imputation.¹⁹ Mann-Whitney U tests and χ² tests were used to analyze continuous and noncontinuous variables, using dummy variables for each category of SDoH. We assessed for interaction effects using analysis of variance. We then built multivariable models to identify covariates associated with each outcome, initially using all covariates with p > 0.1 in univariable analysis and including SDoH categories as a single group and then covariates were backwards eliminated. Once the strongest model was identified, we ran multivariable analysis for each SDoH category and individual variable (eTables 1 and 2, links.lww.com/WNL/D344) for both mortality and WLST prediction. We performed receiver operating characteristic curves for median household income within a zip code to predict WLST.

Finally, we ran an additional univariable and multivariable analyses using the most conservative model that included all factors associated with WLST (age, ICH score, ventilator days, hospital LOS) in patients who lived in zip codes in which median household income was below the national median. The significance threshold was set at a 2-sided p-value of 0.05, and 95% CI were reported for all odds ratios (ORs). All analysis was performed on SPSS v28.

Results

From 11,732 patients enrolled in GWTG across 3 campuses over 6 years of interest, we identified 868 patients (median age 67 [IQR 55–78] years; 43% female) with ICH. Most patients went home after admission for ICH (32%), 28% went to acute rehabilitation, and 20% went to a SNF (Figure 1). The median mRS at discharge was 4 (2–5) (Figure 2). Mortality occurred in 150 (17%) patients, and of them, 84% were due to WLST. Medical history of hypertension, diabetes, and coronary artery disease did not result in a difference of WLST nor mortality, although each were individually associated with poor mRS (p < 0.01 for all).

Figure 1. Discharge Disposition and Mortality.

LTACH = long-term acute care hospital; NH = non-Hispanic; SNF = skilled nursing facility.

Figure 2. Modified Rankin Scale at Discharge.

NH = non-Hispanic.

Factors Associated With WLST, Mortality, and Clinical Outcome

Age

Older age was associated with WLST (median age of 77 years [64–86] vs 66 years [53–77], p < 0.001) (Table 1). Discharge mortality was also associated with older age (p < 0.001; eTable 1, links.lww.com/WNL/D344). In addition, patients with poor mRS were significantly older (69 [59–80] years vs 61 [48.5–73.5] years, p < 0.001, Table 2).

Table 1.

Factors Associated With WLST

	Total (N = 868)	No WLST (n = 741)	WLST (n = 127)	p Value
Age, y, median (IQR)	67 (55–79)	66 (53–77)	77 (64–86)	<0.001
Female	378 (44)	320 (43)	58 (46)	0.6
English primarya	625 (72)	540 (73)	85 (67)	0.2
Race/ethnicitya				0.03
Asian	148 (17)	131 (18)	17 (14)	0.2
Black, non-Hispanic	141 (16)	132 (18)	9 (7)	<0.01
White, non-Hispanic	362 (42)	304 (42)	57 (46)	0.4
Hispanic or Latino	132 (15)	113 (16)	19 (18)	0.6
Marital statusb				0.04
Married	420 (49)	362 (50)	58 (47)	0.5
Partner	29 (3)	27 (4)	2 (2)	0.2
Any companion	449 (53)	389 (53)	60 (48)	0.3
Single	227 (27)	194 (27)	33 (27)	1.0
Widowed	110 (13)	86 (12)	24 (19)	0.02
Divorced or separated	64 (8)	58 (8)	6 (5)	0.2
Lives in a zip code in which
Real median household income <$70,784	324 (37)	294 (40)	30 (24)	<0.01
% of households with income >$200,000, median (IQR)	13.9 (7.2–24.9)	13.7 (7.2–24.6)	15.2 (9.6–31.2)	<0.001
Median household income of zip code, $, median (IQR)	81,857 (58,669–122,078)	77,350 (58,487–117,581)	88,687 (73,489–136,431)	<0.001
Religionc				0.7
None	167 (20)	151 (21)	16 (13)	0.04
Any Christian denomination	481 (57)	401 (56)	80 (66)	0.046
Catholicism	304 (36)	246 (34)	58 (48)	<0.01
Other Christian denomination	177 (21)	155 (22)	22 (18)	0.4
Judaism	98 (12)	87 (12)	11 (9)	0.3
Buddhism, Taoism, or Hinduism	34 (4)	28 (4)	6 (5)	0.6
Islam	21 (3)	18 (3)	3 (2)	1.0
Other religion	26 (3)	21 (3)	5 (4)	0.5
Employmentb				<0.01
Unemployed	248 (29)	218 (30)	30 (24)	0.2
Employed	223 (26)	201 (28)	22 (18)	0.02
Retired	380 (44)	307 (42)	73 (58)	<0.01
Insuranced				<0.001
Medicare	433 (51)	350 (48)	83 (69)	<0.001
Private	289 (34)	265 (37)	24 (20)	<0.001
Medicaid	193 (23)	171 (24)	22 (18)	0.2
Dual	39 (5)	29 (4)	10 (8)	0.04
None or self-insured	16 (2)	13 (2)	3 (3)	0.6
Comorbiditiesa
CAD or MI	431 (50)	361 (50)	70 (59)	0.1
Hypertension	789 (91)	672 (91)	117 (94)	0.4
Diabetes	284 (33)	243 (34)	41 (36)	0.7
Admission NIHSS, median (IQR)	7 (3–15)	7 (2–13)	12 (7–27)	<0.001
Admission ICH score, median (IQR)	1 (0–2)	1 (0–2)	3 (1–4)	<0.001
Ventilator days, median (IQR)	6 (2–15)	7 (2–18)	4 (2–11)	0.03
Hospital LOS, d, median (IQR)	6 (3–13)	6 (3–14)	4 (2–10)	<0.001

Abbreviations: CAD = coronary artery disease; ICH = intracerebral hemorrhage; IQR = interquartile range; LOS = length of stay; MI = myocardial infarction; NIHSS = NIH Stroke Scale; WLST = withdrawal of life-sustaining therapy.

Data are presented as n (%) unless otherwise indicated.

^an = 859.

^bn = 852.

^cn = 839.

^dn = 846.

Table 2.

Factors Associated With Poor Discharge Outcome

	Good mRS (n = 298)	Poor mRS (n = 570)	p Value
Age, y, median (IQR)	61 (49–73)	70 (59–80)	<0.001
Female	126 (42)	252 (44)	0.6
English primarya	226 (76)	399 (70)	0.1
Race/ethnicitya			0.8
Asian	50 (17)	98 (18)	0.8
Black, non-Hispanic	56 (19)	85 (15)	0.2
White, non-Hispanic	125 (42)	236 (42)	1.0
Hispanic or Latino	47 (16)	85 (16)	1.0
Marital statusb			0.1
Married	134 (45)	286 (51)	0.1
Partner	9 (3)	20 (4)	0.7
Any companion	143 (48)	306 (55)	0.1
Single	94 (32)	133 (24)	0.01
Widowed	34 (11)	76 (14)	0.4
Divorced or separated	24 (8)	40 (7)	0.6
Lives in a zip code in which
Real median household income <$70,784	113 (38)	211 (37)	0.8
% of households with income >$200,000, median (IQR)	13.9 (7.2–25.1)	13.9 (7.5–24.9)	0.6
Median household income of zip code, $, median (IQR)	81,142 (58,669–120,699)	81,857 (58,669–123,401)	0.6
Religionc			0.04
None	79 (27)	88 (16)	<0.001
Any Christian denomination	151 (52)	330 (60)	0.04
Catholicism	95 (33)	209 (38)	0.2
Other Christian denomination	56 (19)	121 (22)	0.4
Judaism	29 (10)	69 (13)	0.3
Buddhism, Taoism, or Hinduism	6 (2)	28 (5)	0.04
Islam	6 (2)	15 (3)	0.6
Other religion	13 (5)	13 (2)	0.1
Employmentb
Unemployed	76 (26)	172 (31)	0.1
Employed	116 (40)	107 (19)	<0.001
Retired	101 (34)	279 (50)	<0.001
Insuranced
Medicare	120 (41)	313 (57)	<0.001
Private	129 (44)	160 (29)	<0.001
Medicaid	67 (23)	126 (23)	1.0
Dual	11 (4)	28 (5)	0.4
None or self-insured	6 (2)	10 (2)	0.8
Comorbiditiesa
CAD or MI	119 (41)	312 (57)	<0.001
Hypertension	258 (87)	531 (94)	<0.01
Diabetes	81 (28)	203 (38)	<0.01
Admission NIHSS, median (IQR)	3 (1–7)	10 (7–20)	<0.001
Admission ICH score, median (IQR)	1 (0–1)	1 (1–3)	<0.001
Ventilator days, median (IQR)	1.5 (1–3)	7 (2–16)	<0.001
Hospital LOS, d, median (IQR)	4 (2–6)	9 (4–19)	<0.001

Abbreviations: CAD = coronary artery disease; ICH = intracerebral hemorrhage; IQR = interquartile range; LOS = length of stay; MI = myocardial infarction; mRS = modified Rankin Scale; NIHSS = NIH Stroke Scale.

Data are presented as n (%) unless otherwise indicated. Poor mRS defined by a modified Rankin score of 4–6 points.

^an = 859.

^bn = 852.

^cn = 839.

^dn = 846.

Sex and Race/Ethnicity

Of the patients in this cohort, 16% were Black non-Hispanic, 17% were Asian, and 15% were of Hispanic ethnicity. WLST was significantly less likely among Black non-Hispanic patients WLST (7% vs 18%, p < 0.01). Mortality was lower in Black non-Hispanic patients (9%) compared with survival (18%) (p = 0.04). Female sex was not associated with any of the outcomes on univariable analysis, but female patients were less likely to undergo WLST after adjusting for age, LOS, ventilator days, and ICH score (adjusted OR [aOR] 0.44; 95% CI 0.2–0.97). Hispanic patients were less likely to be discharged to acute rehabilitation (11% vs 18%, p = 0.006) but more likely to go home (20% vs 14%, p = 0.04; Figure 1).

Primary Language

The primary language was English in 72% of patients (Table 1). Primary language was not associated with WLST or mortality, but patients with primary English language were more likely to have a good discharge mRS (60% vs 27% with a poor discharge mRS p < 0.001) and were more likely to be discharged to acute rehabilitation (80%) than not (68%) (p < 0.001) (Figure 1). After multivariable analysis and adjustment for ICH score and ventilator days, English language remained associated with a better discharge mRS (OR 0.29; 95% CI 0.09–0.93). Patients with primary English language were more likely to have a good dichotomized discharge disposition (381 [74%] good vs 191 [67%] bad, p = 0.03).

Religion

Most patients reported any Christian denomination (57%), and of these patients, Catholic religion was the most common (63%), although patients also self-reported as Christian (18%), Baptist (3%), Protestant (2%), Greek Orthodox (2%), Lutheran (1%), Pentecostal (1%), Episcopal (1%), Methodist (1%), Unitarian Universalist (1%), Russian Orthodox (0.8%), Jehovah's Witness (0.6%), Presbyterian (0.4%), Seventh Day Adventist (0.4%), Anglican (0.2%), Mormon (0.2%), and Nazarene (0.2%). Patients who self-reported as Christian were more likely to undergo WLST (66% vs 56%, p < 0.05). Catholic patients had more WLST and higher mortality (45% vs 34%, p = 0.02). Patients who reported no religion or atheism were less likely to undergo WLST and had lower mortality (p < 0.05 for both). Only 1% of patients did not report a religion. After adjusting for age, ICH score, hospital LOS, and ventilator days, Jewish patients were less likely to have WLST (aOR 0.15; 95% CI 0.03–0.78).

Marital Status

Most patients reported having a companion (53%), with married patients comprising 84%. Patients without companions were widowed (13%), divorced or separated (8%), or single (27%). Widowed patients were more likely to undergo WLST (19% vs 12%, p = 0.02). In just patients living in zip codes comprising patients with incomes lower than the national median (eTable 2, links.lww.com/WNL/D344), WLST was lower in patients who were married (50% vs 30%, p = 0.03) or had any companion (54% vs 30%, p = 0.01).

Occupation

In this cohort, 44% of patients were retired, and only 26% were employed (Table 1). Employed patients were less likely to undergo WLST (18% vs 28%, p = 0.02). Patients with good discharge mRS were more likely to be employed than not (40% vs 19%, p < 0.001; Figure 2). Poor discharge disposition as a dichotomized outcome was associated with employment status; 32% of employed patients had a good discharge disposition compared with 18% with a poor discharge disposition (p < 0.001). There was no difference in occupation status in patients who went to acute rehabilitation after discharge. Although almost a third of patients was unemployed (248, 29%), there was no difference in any outcome measure for these patients.

Insurance Status

Half of all patients (50%) were on Medicare and 22% on Medicaid (Table 1). Patients with private insurance were less likely to have WLST (20% vs 37%, p < 0.001) or a poor discharge mRS (29% vs 44%, p < 0.001). Similarly, patients with private insurance were more likely to be discharged alive than to die in the hospital (242 [36%] vs 33 [23%], p < 0.01) and less likely to have WLST (24 [20%] vs 267 [37%], p < 0.001). Patients with private insurance were more likely to be kept full code (21 [21%] vs 251 [37%], p < 0.01) and more likely to have a good discharge disposition (41% vs 23%, p < 0.001). Patients with dual coverage were more likely to undergo WLST (8% vs 4%, p = 0.04) and have a bad discharge disposition (6% vs 3%, p = 0.03).

Neighborhood Socioeconomic Status

The median (IQR) household income was $81,857 ($58,669–$122,078) (Table 1). WLST was more common in patients who lived in areas with higher median household income ($88,687 [$73,489–$136,431] compared with patients who did not $77,350 [$58,487–$117,581], p < 0.001). Mortality was also related to areas with higher median household income ($84,145 [$6,466,375–$136,112] vs $77,378 [$58,669–$11,921,525] in patients discharged alive, p = 0.01, eTable 1, links.lww.com/WNL/D344). There was a higher percentage of households with incomes of $200,000 or more in patients who had WLST (15.2 [9.6–31.2] vs 13.7 [7.2–24.6]) or died before discharge (15.2 [9.3–30.7] vs 13.7 [7.2–24.6], p < 0.01 for both, Table 1 and eTable 1). Patients who lived in neighborhoods in which the household income was less than the median US household income were more likely to be alive at discharge (40% vs 27%, p < 0.01) and to not have WLST (40% vs 24%, p < 0.001). There was no significant difference in income levels by zip code in discharge mRS nor in discharge home. However, patients who went to SNF/LTACH were more likely to be from areas with lower income ($77,031 [$58,487–$107,088] vs $84,002 [$58,850–$122,078], p = 0.02; Figure 1).

In multivariable analysis following backward elimination of covariates and adjusting for age, ICH score, hospital LOS, and ventilator days, residence in areas with higher median household income levels was still associated with WLST (aOR 1.88; 95% CI 1.29–2.74; Table 3). The relationship between higher income by zip code was also still associated with mortality (aOR 1.5; 95% CI 1.06–2.13). There was no significant association between discharge mRS and income by zip code. The area under the curve for median household income was 0.61 for WLST (Figure 3).

Table 3.

SDoH Associated With Mortality, WLST, and Poor Clinical Outcome

	aOR (95 CI)a	p Value
Median income by zip code	1.88 (1.29–2.74)	0.001
Race/ethnicity	1.12 (0.92–1.35)	0.2
Sex	0.44 (0.2–0.97)	0.04
Employment	0.66 (0.39–1.1)	0.1
Insurance	1.04 (0.76–1.41)	0.8
English primary language	0.78 (0.34–1.8)	0.6
Marriage status	0.93 (0.7–1.23)	0.6
Religionb	0.91 (0.75–1.11)	0.4

Abbreviations: aOR = adjusted odds ratio; ICH = intracerebral hemorrhage; SDoH = social determinant of health; WLST = withdrawal of life-sustaining therapy.

^aThe multivariable model included age, ICH score, hospital length of stay, and ventilator days.

^bJewish patients were less likely to have WLST (aOR 0.15, 95% CI 0.03–0.78).

Figure 3. Receiver Operating Characteristic Curve.

ICH = intracerebral hemorrhage; WLST = withdrawal of life-sustaining therapy.

Among patients living in zip codes with incomes lower than the national median (eTable 2, links.lww.com/WNL/D344), WLST was higher in patients of Hispanic/Latino ethnicity (OR 4.65; 95% CI 1.94–11.17), and lower in patients with primary English language (OR 0.24; 95% CI 0.1–0.55), and these relationships remained significant after adjusting for age, ICH score, ventilator days, and LOS (aOR 13.3; 95% CI 3.2–55.46 and aOR 0.1; 95% CI 0.02–0.49). Religion was also significantly associated with WLST in this subset of patients, and the aOR was 5.11 (95% CI 1.33–19.52) in Catholic patients.

Discussion

The correlation between income, education, and occupation within neighborhoods has been well-established.³ Previous work demonstrated a relationship between neighborhood socioeconomic measures and both health traits (e.g., hypertension, smoking status, dementia, and stroke incidence) and outcome (e.g., mortality in multiple sclerosis, outcome after carotid procedures).^{4,5,10,12,20-22} Exploration into the relationship between SDoH and WLST usage can allow us to understand why mortality discrepancies exist in certain patient populations. Although mortality has been associated with the socioeconomic features of neighborhoods in patients with acute ischemic stroke,²³ there is limited research on the relationship between SDoH and both WLST and mortality after ICH. A recent study that demonstrated low socioeconomic status was associated with increased 30-day mortality in neurologic patients assessed ischemic and hemorrhagic stroke together,¹² which is problematic because ICH and ischemic stroke mortality trends differ.²⁴ Our study is novel because we focused on the relationship between SDoH and WLST and mortality in patients with ICH only.

It is well-known that age, NIHSS admission score, ICH score, and LOS are associated with ICH mortality.²⁵ Here, we showed that patients with higher nSES have higher rates of WLST. On univariable analysis, we found that WLST and mortality were more common in older patients, patients who lived in zip codes with higher median household incomes, and patients on Medicare and less common in patients who were Black non-Hispanic or had private insurance. In a sequential empirical model including all SDoH, ICH score, admission NIHSS, ventilator days, and hospital LOS, nSES remained associated with WLST.

There are numerous potential reasons why nSES is related to outcome after stroke. Lack of access to healthy food, longer distance to acute care hospitals, limited access to safe exercise areas, and reduced social and neighborly interaction, and cohesion have all been associated with worse stroke outcomes.²⁶ nSES is associated with delayed stroke recognition,²⁷ which affects treatment and mortality.²⁸ In addition, patients with lower nSES are less likely to receive appropriate inpatient therapy after stroke.²⁹ Outcome disparities are thus a direct result of differences in access to clinical care and societal resources among communities, potentially stemming from systemic practices of racism and segregation; for example, there is a higher stroke prevalence in areas associated with a history of redlining³⁰ and a higher stroke mortality in counties that participated in slavery in the United States.³¹ Factors such as exposure to heavy metals, pesticides, and exhaust established within systems of inequity may also contribute to increased comorbidities and perpetuate health disparities.³²

Our study corroborated previous literature suggesting Black non-Hispanic patients were less likely to undergo WLST.^7,8 However, although some literature suggests similar or higher mortality rates in Black non-Hispanic patients after ischemic and hemorrhagic stroke compared with patients of other race/ethnicities,²⁶ we observed a lower incidence of mortality in Black non-Hispanic patients in our cohort, similar to studies that examined interactions of age.^25,33 Our group previously reported that in a different cohort of patients with ICH or subarachnoid hemorrhage, Black or Hispanic race/ethnicity was associated with a lower likelihood of a code status of do-not-resuscitate,³⁴ indicating that race/ethnicity may affect goals-of-care decision-making after ICH. That study also showed lower rates of palliative care consultation for Black and Hispanic patients, meriting consideration of the need for cultural sensitivity and equity when approaching goals-of-care discussions. One explanation for this disparity might stem from a mistrust of the medical profession reported by Black patients, possibly based on an experience of epistemic injustice and racism within health care.³⁵

In patients living in zip codes comprising patients with lower incomes, WLST was higher in patients of Hispanic/Latino ethnicity and in patients whose primary language was not English. This is consistent with a survey of inner-city Latino population in which focus groups described the importance of preserving patient dignity.³⁶ However in a study in the general intensive care unit population, WLST was actually lower in patients with limited English proficiency, although this study did not assess the impact of nSES.³⁷ Whether the gap between non-Hispanic Black and Hispanic/Latino patients' use of WLST in neighborhoods with low nSES is a result of differing culture and traditions, religious influence or language barriers remains to be explored.

We found a lower rate of WLST in Jewish patients after adjusting for confounders. There is a diversity of perspectives about WLST within and across different faiths. Patients who report a religious affiliation practice with varying adherence to the same laws and customs across a spectrum ranging from strictly observant to cultural identity. A religious moral argument for WLST has been based on balancing of good intention and burden of treatment. Followers of Roman Catholicism might adhere to the concept developed by Thomas Aquinas, in that the good intention should be greater than an evil arising from the action. This was further elaborated on by Francisco DiVitoria, who proposed that the burden of treatment does not prolong death (as opposed to prolonging life). The Lutheran, Anglican, and Presbyterian Church all support WLST.³⁸ In Judaism some sects are strongly against the concept of WLST while others allow for removal of “ineffective therapy.”³⁸ Withdrawal of futile treatment and unnecessarily burdensome care is supported by Islam. Eastern religions, such as Hindu traditions, focus less on bodily function and might support the avoidance of artificial life support.³⁸ These religious differences ought to be considered when discussing the patient's family and cultural values during goals-of-care discussions.

Although neurology-trained neurointensivists led the goals-of-care discussions in our study and this training background is associated with good ability to predict poor outcome,³⁹ we cannot exclude the possibility that unconscious bias contributed to our findings that WLST was more common for patients from neighborhoods with higher median household income levels and less common for Black non-Hispanic patients and patients with private insurance. WLST affects mortality, so unconscious bias when discussing end of life care may affect mortality differences leading to the potential for a self-fulfilling prophecy that certain groups of patients will have worse outcomes. These findings demand attention by clinicians conducting goals-of-care discussions after ICH. For example, using a race and socioeconomic conscious framework that can also give credence to holistic well-being of the patient, familial duties, concern for caregiver burden, and differing practices and needs for spiritually should be considered.³⁶ Exploration of core values is paramount. While financial considerations should not dictate life or death decisions, it is also important to acknowledge the cost of care after ICH, particularly for families who live in neighborhoods with inequitable resources.

The strength of this study is the cohort size and demographic diversity. However, the patients were all hospitalized in 1 metropolitan city. One limitation is that we do not have systematically collected data on how many patients had appointed a health care proxy or completed advance directives and who participated in the goals-of-care discussions. Another important limitation is the lack of self-reported race and ethnicity, preferred sex, and precise methods of determining individual economic status (i.e., income on a personal level rather than median household income by zip code). Unfortunately, education was not captured in our data set, and it is, therefore, difficult to draw any conclusions regarding how education or cultural values relate to the decision to undergo WLST. However, previous studies comparing WLST found that education level was not associated with WLST decision-making.³⁹

SDoH, including nSES, are associated with WLST and mortality after ICH. Further research is required to investigate the relationship between mortality and WLST and the structural and social environments that can influence systemic health and hemorrhagic stroke in various neighborhoods. In addition, it is essential to examine how the utilization of WLST may affect outcomes in patients from diverse socioeconomic and racial/ethnic backgrounds. Understanding these associations is crucial for comprehending the broader consequences and factors influencing health outcomes in different populations. The strong association with insurance and mortality on univariable analysis has major implications for practice and policy. The association between mortality and areas with higher median household income suggests SDoH have important repercussions for the care and management of patients with ICH. Considering how to tailor our perspective and our language might be 1 step toward creating a more equitable approach to goals-of-care discussions after ICH.

Glossary

aOR

adjusted OR

EMR

electronic medical record

GWTG

Get with the Guidelines

ICH

intracerebral hemorrhage

IQR

interquartile range

LOS

length of stay

LTACH

long-term acute care hospital

mRS

modified Rankin Scale

NIHSS

NIH Stroke Scale

nSES

neighborhood socioeconomic status

OR

odds ratio

SAR

subacute rehabilitation

SDoH

social determinants of health

SNF

skilled nursing facility

WLST

withdrawal of life-sustaining therapy

Appendix. Authors

Name	Location	Contribution
Kara R. Melmed, MD	Department of Neurology and Neurosurgery, NYU Langone Health and NYU Grossman School of Medicine, New York	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design; analysis or interpretation of data
Ariane Lewis, MD	Department of Neurology and Neurosurgery, NYU Langone Health and NYU Grossman School of Medicine, New York	Drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data
Lindsey Kuohn, MD	Department of Neurology, NYU Langone Health and NYU Grossman School of Medicine, New York	Major role in the acquisition of data
Joanna Marmo, BSN, RN, SCRN	Department of Neurology, NYU Langone Health, New York	Major role in the acquisition of data
Nirmala Rossan-Raghunath, MSN, SCRN	Department of Neurology, NYU Langone Health, New York	Major role in the acquisition of data
Jose Torres, MD	Department of Neurology, NYU Langone Health and NYU Grossman School of Medicine, New York	Drafting/revision of the manuscript for content, including medical writing for content; study concept or design; analysis or interpretation of data
Rajanandini Muralidharan, MD	Department of Neurology, NYU Langone Health and NYU Grossman School of Medicine, New York	Drafting/revision of the manuscript for content, including medical writing for content
Aaron S. Lord, MD, MSc	Department of Neurology, NYU Langone Health and NYU Grossman School of Medicine, New York	Analysis or interpretation of data
Koto Ishida, MD	Department of Neurology, NYU Langone Health and NYU Grossman School of Medicine, New York	Drafting/revision of the manuscript for content, including medical writing for content; analysis or interpretation of data
Jennifer A. Frontera, MD	Department of Neurology, NYU Langone Health and NYU Grossman School of Medicine, New York	Drafting/revision of the manuscript for content, including medical writing for content; major role in the acquisition of data; study concept or design; analysis or interpretation of data

Study Funding

No targeted funding reported.

Disclosure

The authors report no relevant disclosures. Go to Neurology.org/N for full disclosures.