Racial Differences in Palliative Care and Hospice Among Adults with Chronic Obstructive Pulmonary Disease

Introduction

Chronic obstructive pulmonary disease (COPD) is a serious respiratory illness associated with debilitating symptoms including breathlessness, fatigue, and anxiety.^1,2 Individuals with COPD often experience frequent hospitalizations and high-intensity care at the end-of-life (EOL).^3,4 International professional societies recommend palliative approaches in COPD.^5–7 Although people with COPD experience a higher symptom burden and worse functional status than patients with other serious illness such as cancer, people with COPD are less likely to receive specialty palliative care (PC) or hospice and more likely to die in the hospital or the intensive care unit (ICU).^8–13 Furthermore, evidence of inequities within COPD exist. For example, Black individuals and those with lower socioeconomic status are likelier to experience COPD exacerbations and have worse quality of life than White individuals and those with higher socioeconomic status.^14,15 However, whether racial and socioeconomic differences exist in COPD at the EOL specifically remains unknown.

Methods

This retrospective cohort study used medical records from individuals aged ≥40 years in the Georgia Death Index (GDI) who died between 2014 and 2018 with International Classification of Diseases (ICD), Tenth Revision codes J42.xx, J43.xx, or J44.xx listed as underlying cause of death, received care from Emory Healthcare, and had a documented spirometry test. We did not review each spirometry value to confirm a diagnosis of COPD, as race-based corrections during the study period may have introduced additional selection bias and worsened disparities in diagnosis.¹⁶ However, we restricted our sample to decedents who had undergone spirometry to increase specificity, as spirometric testing is an essential step in the diagnosis of COPD.^17,18 Individuals in the GDI were matched to medical records in the Emory University Clinical Data Warehouse using Social Security Number, date of birth, and last name. We determined COPD diagnosis from the GDI. As not all individuals died in an Emory Healthcare hospital, we obtained date of death from the GDI. This study was deemed by the Emory University IRB to not constitute human subjects research as it used decedents’ medical records.

Outcomes were PC consultation (PCC) and hospice order. PCC was determined by the presence of ICD-9 code V66.7 or ICD-10 code Z51.5. To determine PCC timing, we used the date of the first clinical encounter associated with the PCC diagnostic code. Hospice orders were determined by the presence of orders in the medical record. Patient-level independent variables included race, sex, age at death, insurance, death year, spiritual care consult (SCC), and ICU use in the last 30 days. Race and age at death were obtained from the GDI; insurance was obtained from the medical record. Ethnicity was not included due to inconsistent recording. ZIP codes were obtained from the GDI and linked to Rural-Urban Commuting Area Codes to create metropolitan, micropolitan, small town, and rural categories.¹⁹ We linked ZIP codes to the social deprivation index (SDI), a composite measure of area-level deprivation based on demographic characteristics from the American Community Survey.^20,21 SDI ranges from 1 (lowest) to 100 (highest) and was classified into quartiles: Q1 (1–25), Q2 (26–50), Q3 (51–75), and Q4 (76–100).²²

We used Pearson’s 𝑋² test to evaluate differences in the distribution of decedent characteristics and race. We estimated prevalence ratios (PR) and 95% confidence intervals of each EOL outcome using Poisson regression with robust standard errors adjusting models for sex, age at death, marital status, insurance status, rurality, social deprivation, SCC and ICU use. Analyses were conducted using Stata17 (StataCorp).

Results

Of 1,188 decedents, 80.2% were White (n=953) and 19.8% were Black (n=253). (Table 1). Mean death age was 75.3 years and was lower among Black than White decedents (72.7 vs 76.0 years).

Table 1.

Decedent Characteristics by Race

Characteristics	Overall N=1,188		White N=953 (80.2%)		Black N=235 (19.8%)		White v Black
	N	%	N	%	N	%
Sex							.544
Female	621	52.3	494	51.8	127	54.0
Male	567	47.7	459	48.1	108	46.0
Age at Death							<.001
Mean (SD)	75.3 (10.4)		76.0 (10.3)		72.7 (10.4)
<54	24	2.02	13	1.3	11	4.9
55–64	177	14.9	137	14.4	40	17.0
65–79	548	46.1	424	44.5	124	52.8
80 +	439	37.0	379	39.8	60	25.5
Marital Status							<.001
Married	597	50.3	513	53.8	84	35.7
Single	139	11.7	89	9.3	50	21.3
Divorced/Separated	146	12.3	104	10.9	42	17.9
Widowed	253	21.3	201	21.1	52	22.1
Unknown	53	4.5	46	4.8	7	3.0
Insurance							.206
Private	396	33.4	315	33.1	81	34.5
Medicare	530	44.7	422	44.3	108	46.0
Medicaid	53	4.5	38	4.0	15	6.4
Miscellaneous	8	0.7	7	0.7	1	0.4
Not Recorded	201	16.9	171	17.9	30	12.8
Rural Urban							<.001
Metropolitan	1,001	84.3	776	81.4	225	95.7	<.001
Micropolitan	150	12.6	140	14.7	10	4.3
Small town	21	1.8	21	2.2	0	0.0
Rural	16	1.4	16	1.7	0	0.0
Social Deprivation
Quartile 1 (lowest)	183	15.4	166	17.4	17	7.2
Quartile 2	275	23.2	245	25.7	30	12.8
Quartile 3	409	34.4	358	37.6	51	21.7
Quartile 4 (highest)	321	27.0	184	19.3	137	58.3
End-of-Life Care Use
Palliative Care Consultation (PCC)	284	23.9	201	21.1	83	35.3	<.001
Days from first PCC to death	Median (IQR) 58.8 (12.0, 289.4)		Median (IQR) 46.5 (11.7, 219.1)		Median (IQR) 101.3 (13.3, 519.0)		.049
Hospice Order	203	17.1	149	15.6	54	23.0	.007
Spiritual Care Consultation (SCC)	374	31.5	264	27.7	110	46.8	<.001
ICU in last 30 days	123	10.4	75	7.9	48	20.4	<.001
Hospitalization in last 90 days	296	24.9	206	21.6	90	38.3	<.001
Hospital Death	58	4.9	35	3.7	23	9.8	<.001

Note: P-value determined using X2, t-test, and Wilcoxon rank sum tests. Column percentages presented.

Nearly 24% (n=284) of individuals received a PCC and 17% (n=203) received an order for hospice. Median time from PCC to death was shorter among White than Black decedents (46.5 vs 101.3 days. Nearly 10% (n=123) of individuals experienced an ICU stay and 5% (n=58) died in-hospital.

In an unadjusted model, Black individuals were more likely to receive PCCs (adjusted Prevalence Ratio(aPR) 1.68; 95%CI 1.35–2.07) and a hospice order (aPR 1.47, 95% CI 1.1–1.9) relative to White individuals. However, these results were no longer statistically significant after adjusting for ICU use. (Table 2 and Supplemental Table 1)

Table 2:

Prevalence Ratios of Palliative Care Consultation and Hospice Order by Race

	Palliative Care Consultation				Hospice Order
Variables	Unadjusted		Adjusted		Unadjusted		Adjusted
	OR	95% CI	aOR	95% CI	OR	95% CI	aOR	95% CI
Sex
Female	Ref		Ref		Ref		Ref
Male	1.02	.833–1.25	1.05	.879–1.26	.821	.637–1.06	.922	.723–1.18
Race
White	Ref		Ref		Ref		Ref
Black	1.68***	1.35–2.07	1.05	.851–1.29	1.47**	1.11–1.93	.830	.614–1.12
Age at death
≤54	Ref		Ref		Ref		Ref
55–64	1.19	.500–2.60	.661	.272–1.61	1.70	.428–6.71	1.87	.407–8.55
65–79	1.10	.494–2.43	.998	.412–2.42	1.77	.463–6.79	1.77	.395–7.96
80 +	1.23	.552–2.72	1.02	.418–2.48	2.60	.650–9.91	2.37	.525–10.66
Marital Status
Married	Ref		Ref		Ref		Ref
Single	1.26	.928–1.71	1.02	.747–1.38	1.57*	1.08–2.29	1.43	.982–2.08
Divorced/Separated	.953	.674–1.35	.804	.598–1.08	1.30	.867–1.94	1.15	.787–1.69
Widowed	1.30*	1.01–1.65	1.16	.929–1.44	1.81***	1.35–2.43	1.37*	1.03–1.83
Unknown	.678	.352–1.31	1.08	.618–1.89	.275	.070–1.09	.437	.109–1.75
Insurance
Private	Ref		Ref		Ref		Ref
Medicare	1.25*	1.01–1.55	1.35	.938–1.37	1.51**	1.15–1.98	1.30*	1.01–1.68
Medicaid	.924	.546–1.57	.968	.604–1.55	.712	.324–1.56	.786	.354–1.74
Other	.255***	.147–.444	.453**	.273–.750	.212***	.099–.454	.432*	.207–.904
Rural Urban
Metropolitan	Ref		Ref		Ref		Ref
Micropolitan	.306***	.176–.531	.469**	.277–.750	.139***	.052–.369	.237**	.094–.600
Small town	1.09	.551–2.16	.837	.492–1.42	1.24	.571–2.70	1.10	.639–1.91
Rural	.955	.406–2.25	.995	.524–1.89	.652	.177–2.40	.702	.246–2.00
Social Deprivation
Quartile 1 (SDI 1–25)	Ref		Ref		Ref		Ref
Quartile 2 (SDI 26–50)	.783	.567–1.08	.846	.637–1.12	.802	.548–1.17	.866	.613–1.23
Quartile 3 (SDI 51–75)	.825	.615–1.12	.977	.760–1.26	.711	.496–1.02	.880	.635–1.22
Quartile 4 (SDI 76–100)	.883	.653–1.19	.877	.663–1.16	.804	.556–1.16	.884	.614–1.27
ICU use in last 30 days	5.39***	4.64–6.26	3.65***	3.04–4.38	4.75***	3.82–5.92	3.24***	2.51–4.19
Spiritual Care Consult	4.01***	3.24–4.94	2.60***	2.10–3.22	4.42***	3.39–5.76	2.73***	2.06–3.62
Year of Death	.946	.881–1.01	1.02	.955–1.08	.883**	.810-.963	.940	.865–1.02

Table is based on Poisson regression models assessing the association of decedent characteristics and each EOL outcome. Models were adjusted for sex, age at death, marital status, insurance status, rurality, social deprivation, spiritual care consult, and ICU utilization. White decedents are the reference. In the regression, miscellaneous and no recorded insurance were combined as only 8 individuals were reported to have miscellaneous insurance.

Discussion

In this cohort of COPD decedents, Black individuals were more likely to receive PCCs and hospice orders than White individuals; however, after adjusting for ICU use, Black race was no longer a significant predictor of PCC or hospice order. Decedents in our sample had fewer hospitalizations and ICU stays but more hospice orders than in prior cohort studies of COPD decedents.^13,23

The need for ICU care may be a driver of PCC at the EOL. In our cohort, Black individuals were more likely to receive ICU care at the end-of-life. These findings may, in part, be driven by patient preferences, as that Black individuals have been shown to prefer more aggressive care at the EOL.²⁴ Greater use of intensive care near EOL among Black individuals may prompt PCCs for goals-of-care conversations and caregiver support.^25,26 ICU clinicians may have limited confidence, time, and skills to conduct goals-of-care conversations.²⁷ Our findings may also indicate recognition of the disparate symptom burden and palliative needs among Black individuals with COPD.

Limitations include generalizability of findings as our data were from one health system, though greater receipt of PCCs has been observed in Black individuals with cancer and heart failure from various geographic regions and health systems.^22,25,28 Other limitations include a lack of measures of disease severity, reasons for PCC, and patient preferences regarding end-of-life care. While identifying COPD via ICD codes has low sensitivity compared with manual chart review, ²⁹ adding an age cutoff and selecting decedents with documented spirometry increases the diagnostic accuracy in this sample. Additionally, there were too few decedents of other racial groups to include in the analysis. Although ICD codes have limitations in identifying PCC in electronic medical records, all decedents in our sample had documented PC encounters.^30,31

We found that fewer than a quarter of COPD decedents received PCCs, with earlier receipt observed among Black decedents. PC remains underutilized in COPD and efforts to increase early PCCs in COPD among all patients are needed.

Highlights.

• Black individuals with COPD are equally likely to receive palliative care as White individuals.

• Intensity of end-of-life care may be a driver of racial disparities in palliative care use.

• Median time from palliative care consultation to death is shorter among White vs Black individuals.

• Most decedents with COPD who receive palliative care do so within the two months prior to death.