Hospice Referral Rate Disparities of American Indian/Alaska Native Kidney Transplant Recipients with End-Stage Kidney Disease: A Retrospective Cohort Analysis

Abstract

Introduction:

American Indian/Alaska Native (AI/AN) persons disproportionately suffer from end-stage kidney disease caused by diabetes (ESKD-D). Kidney transplant is the most desirable option to treating ESKD-D, but remains unattainable for many AI/AN persons, especially in rural South Dakota (SD). Additionally, palliative and hospice care options for AI/AN with any serious illness in SD are largely inaccessible. Moreover, receiving kidney transplant potentially affects hospice referral because of the desire to prolong transplant function. Therefore, the purpose of this study was to compare hospice use rates among AI/AN and non-Hispanic White (NHW) persons with ESKD-D prior to death and determine if differences in referral rates are present for those with and without a prior kidney transplant.

Methods:

Retrospective cohort analysis of United States Renal Data System data from 2000–2021. Data for persons with hospice care, transplant status, place of death, and race were analyzed using chi-squared tests with Yates’ continuity correction and the Cochran-Mantel-Haenszel test.

Results:

AI/AN persons with ESKD-D were less likely to receive hospice care prior to death compared to NHW persons in both transplant (P<0.001) and non-transplant (P<0.001) groups. When comparing transplant and non-transplant groups by hospice use, persons with no previous transplant were more likely to receive hospice care prior to death (P<0.001).

Conclusion:

These results confirm the assumptions of significant differences in hospice care use among AI/AN versus NHW who have ESKD-D, including differences between those with a prior transplant. There is a need to expand palliative/hospice care services for persons with a prior kidney transplant.

INTRODUCTION

American Indian/Alaska Native (AI/AN) persons disproportionately suffer from diabetes and end-stage kidney disease (ESKD) compared to non-Hispanic whites (NHW).¹ Over two-thirds of all ESKD in AI/AN persons is caused by diabetes (ESKD-D).² For persons with ESKD-D, the significant burdens of dialysis negatively affect quality of life (QOL).^3,4 Furthermore, AI/AN persons with ESKD-D are diagnosed between 1.8 and 12.6 years earlier and live longer on dialysis than their NHW counterparts⁵ which compounds the negative impact on QOL by prolonging time on dialysis. Kidney transplant is the most desirable option to treat ESKD, but kidney transplant remains unattainable for many AI/AN persons, especially in South Dakota (SD).⁶ There are several challenges inhibiting transplant access⁷ that are largely determined by social drivers of health (SDOH; e.g., rural and remote geography,⁸ extensive poverty^9,10) and structural determinants (e.g., chronically underfunded healthcare services,^11–13 racial discrimination^14–16). Most AI/ANs in SD live in the Western half of the state in rural and frontier areas where dialysis is often the only feasible treatment option and driving to the nearest transplant center takes 5–6 hours one-way.¹⁷ The chronic underfunding of Indian Health Service (IHS) facilities also affects transplant access as these facilities ultimately approve a referral for transplant consideration. Although IHS provides comprehensive health services for 2.8 million of the 3.7 million AI/AN’s living in the US, 20% rely exclusively on IHS, meaning they have no additional health coverage and are considered uninsured.^18,19 The remaining 80% have additional health coverage through private or public insurance.^18,19 Regardless of primary insurance coverage, kidney transplant surgery and life-long immunosuppressant drug therapy is expensive (average surgery cost is $442,500;²⁰ average standard drug therapy cost is $25,000 per year²¹). When IHS is involved in payment for transplant related services, it takes sparse financial resources away from other high need illnesses.

Because of the significant challenges to obtaining a kidney transplant, the issues of diminished QOL from prolonged dialysis persist. Palliative care is an ideal option to help balance the burdens of dialysis with QOL goals; however, the same social and structural determinants that limit kidney transplant access also limit access to specialty services like palliative or hospice care for AI/AN populations. Furthermore, there is evidence that prior recipients of a kidney transplant may be less likely to receive palliative/hospice care due to a desire to prolong transplant function as long as possible.^22,23 Yet, it is unclear how race/ethnicity and transplant status impacts hospice care use prior to death in ESKD-D populations.

The United States Renal Data System (USRDS) is the data repository for all persons in the US who initiate dialysis. In addition to data collection at the time of dialysis initiation, additional data is gathered at the time of death which specifies whether individuals received hospice services prior to death or not. The USRDS does not track palliative care use and is not able to provide a date when hospice care started prior to death. The death report also contains the location of death (hospital, home, nursing home, or dialysis unit), one of the metrics that has been used for determining hospice care quality.^24–27 Despite the limitations in the dataset, there is great value in exploring hospice use disparities at a general level using the USRDS dataset. The purpose of this study was to compare hospice use rates among AI/AN and NHW persons with ESKD-D prior to death and determine if differences in referral rates are present for those with and without a prior kidney transplant (herein referred to as non-transplant).

METHODS

This manuscript reports the findings of a retrospective cohort analysis of data obtained from the USRDS. The development of this manuscript was guided by the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) criteria for cohort studies.²⁸ The study was approved by the South Dakota State University Institutional Review Board.

Cohort Derivation

Using the 2022 USRDS Standard Analysis Files,²⁹ the PATIENTS and DEATH files were merged using the USRDS ID. All adults aged ≥ 18 years from the combined file who initiated chronic hemodialysis or peritoneal dialysis between 2000 and 2021 were retrospectively identified for inclusion. Records without a diabetes-related primary cause of ESKD were excluded (ICD-9 and ICD-10 codes for ESKD-D found on the CMS-2728 forms³⁰). Persons who did not identify as AI/AN or NHW were excluded from the analysis. Furthermore, if the DEATH record hospice indicator was missing or listed as unknown, the record was excluded. From this cohort, we created 4 groups for analysis: 1) prior kidney transplant + hospice [4,083 records], 2) prior kidney transplant + no hospice [12,344 records], 3) non-transplant + hospice [91,529 records], and 4) non-transplant + no hospice [226,502 records]. The total number of records included was 334,458.

Variables of Interest

Hospice Use

The hospice use indicator is reported within the DEATH file and is used to indicate if a person was referred to and used hospice care prior to death (options include yes, no, or unknown). This data is collected from the CMS-2746 form (ESKD death notification). There is no palliative care indicator on this form, so it is unknown if persons who received hospice prior to death had received palliative care prior to hospice initiation. Additionally, it is not known how soon hospice was initiated before death.

Place of Death

The DEATH file also contains the location of patient death (hospital, dialysis unit, home, nursing home, or other). The location of death was compared to the hospice indicator and transplant history to see if there are meaningful differences in the location of death based on the hospice indicator and transplant status.

Statistical Analysis

All the data analysis in this study were conducted using R 4.2.2³¹ at the 95% confidence interval (α = 0.05). Pearson chi-squares test is a statistical method used for analyzing contingency tables (often 2×2 or Kx2) to test if the observed value and expected values for the cells in the table are close to each other or there exists a statistical difference between them.³² The statistics can be calculated using

$${\sum _{{\{}_i}}^{\frac{{\left(O_i-E_i\right)}^2}{E_i}}~{\chi }_{df}^2$$

where i is the number of categories, O_i and E_i are the observed value and the expected value for the i^th category, respectively, and df denotes the degrees of freedom of the chi-square statistic. The obtained statistics is compared to the chi-square distribution table and if p-value is smaller than the type one error, then the null hypothesis is rejected in favor of the alternative.³³ Despite its benefit, Pearson chi-squares tries to approximate a discrete distribution by using a continuous distribution which can lead to incorrect results, especially for small sample sizes. To account, the Yate’s correction subtracts 0.5 in the numerator of the test statistic which has been shown that this correction leads to a more unbiased results.³⁴ Cochran-Mantel-Haenszel test is similar to Pearson chi-square only for a 3 level table such as $K_1\times K_{2 }\times K_3$.³⁵

RESULTS

Table 1 provides a side-by-side comparison of the transplant versus non-transplant groups. Notable differences between these two groups are the age of diagnosis (18–44; NHW) in the transplant compared to non-transplant groups being 40.8% and 5.6%, respectively. This suggests that NHW persons are more likely to receive a kidney transplant at a younger age than AI/AN persons.

Table 1:

Cohort Characteristics

	Previous Kidney Transplant			No Prior Kidney Transplant
	Total (N)	AI/AN (n [%])	NHW (n [%])	Total (N)	AI/AN (n [%])	NHW (n [%])
Nationwide Sample	~16,419†	~591 (3.6)†	15,828 (96.4)	318,031	10,488 (3.3)	307,543 (96.7)
Gender (%)
Female	6,091 (37.1)	260 (43.4)	5,831 (36.8)	136,510 (42.9)	5,470 (52.2)	131,040 (42.6)
Age (at ESKD Diagnosis-in years)
18–44	6,592	140 (23.4)	6,452 (40.8)	18,656	1,325 (12.6)	17,331 (5.6)
45–54	4,191	216 (36.0)	3,975 (25.1)	37,997	2,358 (22.5)	35,639 (11.6)
55–64	3,975	193 (32.2)	3,782 (23.9)	80,208	3,248 (31.0)	76,960 (25.0)
65–74	1,560	50 (8.4)	1,510 (9.5)	99,216	2,507 (23.9)	96,709 (31.5)
75+	109	0	109 (0.1)	81,954	1,050 (10.0)	80,904 (26.3)
Comorbidities (N/n with comorbidity present at dialysis start date) *
AHD	1,864	70	1,794	70,365	1,429	68,936
Cancer	472	13	459	19,221	296	18,925
CHF	3,309	151	3,158	135,844	3,420	132,424
CVA/TIA**	950	29	921	35,269	882	34,387
Tobacco User	665	16	649	20,565	662	19,903
Total Number of Additional Comorbidities (n (% of nationwide n)) +
0		309 (51.6)	8,584 (54.2)		5,264 (50.2)	115,466 (37.5)
1		170 (28.4)	3,923 (24.8)		3,239 (30.9)	113,632 (36.9)
2		41 (6.8)	1,204 (7.6)		1,292 (12.3)	56,561 (18.4)
3		<11†	194 (1.2)		246 (2.3)	13,869 (4.5)
4		<11†	17 (0.1)		32 (0.3)	1,491 (0.5)
5		70 (11.7)	1,906 (12.0)		413 (3.9)	6,327 (2.1)

^*Data available from CMS-2728 Form Box 16 (co-morbid conditions).³⁰ AI/AN: American Indian/Alaska Native; AHD: Atherosclerotic Heart Disease; CHF: Congestive Heart Failure;

^**CVA/TIA includes history of cerebrovascular accident (CVA) and transient ischemic attack (TIA); ESKD: End-stage kidney disease; NHW: Non-Hispanic White.

⁺Total number of additional comorbidities include any combination of the 5 comorbid variables included in the analysis;

^†Approximate values given per USRDS guidelines for <11 identified eligible patients in a cell (which is further broken down in Tables 2 & 3).

Table 2 explores the differences in hospice use rates and place of death between AI/AN and NHW with previous kidney transplant. There is a significant difference between NHW persons receiving hospice care (25.1%) compared to AI/AN persons (18.2%; P < 0.001). Place of death yielded significant results between the hospice=yes and hospice=no groups. Notably, there is a much lower rate of death at the hospital and much higher rate of death at home for the hospice=yes group, but rates of death in the hospital and receiving hospice are higher when compared to the general population.³⁶

Table 2:

Hospice Use Differences – Previous Kidney Transplant

	Total (N [%])	AI/AN (n [%])	NHW (n [%])	P value
Hospice Use Indicator				<0.001 *
Yes	~4,075 (25.0)†	~101 (18.2)†	3,974 (25.1)
No	12,344 (75.0)	490 (81.8)	11,854 (74.9)
Place of Death^ (Hospice = Yes)				<0.001 **
Hospital	1,147 (28.18)	34 (31.2)	1,113 (28.0)
Home	1,545 (37.97)	46 (42.2)	1,499 (37.7)
Nursing Home	~521 (13.1)†	<11†	521 (13.1)
Dialysis Unit / Other	~846 (20.78)†	>19 (>17.0)†	827 (20.7)
^Place of death (Hospice = Yes) – 14 missing data (all NHW)
Place of Death^^ (Hospice = No)				<0.001 **
Hospital	8,002 (64.8)	325 (66.3)	7,677 (64.8)
Home	2,995 (24.3)	111 (22.7)	2,884 (24.3)
Nursing Home	640 (5.2)	25 (5.1)	615 (5.2)
Dialysis Unit / Other	696 (5.4)	28 (5.3)	668 (5.4)
^^Place of death (Hospice = No) – 11 missing

Bold values indicate statistical significance.

^†Approximate values given per USRDS guidelines for <11 identified eligible patients in a cell;

^*Chi-squared test with Yates’ continuity correction (race and hospice indicator);

^**Cochran-Mantel-Haenszel test (race, hospice indicator, & place of death)

In Table 3, we show the differences in hospice care rates and place of death between non-transplanted AI/AN and NHW persons. In this group, there is a significant difference between NHW persons receiving hospice care (29.1%) compared to AI/AN persons (19.7%; P < 0.001). Similarly to the transplant group, those in the non-transplant group and receiving hospice had higher rates of dying at their home compared to those without hospice care.

Table 3:

Hospice Use Differences – No Prior Kidney Transplant

	Total (N [%])	AI/AN (n [%])	NHW (n [%])	P value
Hospice Use Indicator				<0.001 *
Yes	91,529 (28.8)	2,071 (19.7)	89,458 (29.1)
No	226,502 (71.2)	8,417 (80.3)	218,085 (70.9)
Place of Death^ (Hospice = Yes)				<0.001 **
Hospital	25,447 (26.8)	644 (31.1)	24,803 (27.7)
Home	31,766 (33.5)	789 (38.1)	30,977 (34.6)
Nursing Home	15,887 (16.7)	236 (11.4)	15,651 (17.5)
Dialysis Unit / Other	18,325 (20.0)	398 (19.2)	17,927 (20.0)
^Place of death (Hospice = Yes) – 104 missing data
Place of Death^^ (Hospice = No)				<0.001 **
Hospital	149,878 (66.2)	5,616 (66.7)	144,262 (66.1)
Home	46,581 (20.6)	1,903 (22.6)	44,678 (20.5)
Nursing Home	5,711 (2.5)	245 (2.9)	5,466 (2.5)
Dialysis Unit / Other	24,220 (10.7)	649 (7.7)	22,083 (10.8)
^^Place of death (Hospice = No) – 112 missing

Bold values indicate statistical significance.

^*Chi-squared test with Yates’ continuity correction (race and hospice indicator);

^**Cochran-Mantel-Haenszel test (race, hospice indicator, & place of death)

In Table 4, we combined transplant and non-transplant groups into the same chi-square by race and hospice care indicator. Within the five comparisons made, two were statistically significant. First, there is a much lower rate of hospice care among those who have received a kidney transplant. Second, there is a significant increase in hospice care rates for NHW in the two groups (previous kidney transplant [25.1%] versus non-transplant [29.1%; P < 0.001]). The same comparison in AI/AN group was not significant (previous kidney transplant [18.2%] versus non-transplant [19.7%; P = 0.381]).

Table 4:

Comparison of Hospice Use Rate Differences – Kidney Transplant vs Non-Transplant

	Previous Kidney Transplant	No Previous Kidney Transplant	P value*
Hospice = Yes (all data)	4,083	91,529	<0.001
Hospice = No (all data)	12,344	226,502
AI/AN & Hospice Use=Yes	109	2,071	0.381
AI/AN & Hospice Use=No	490	8,417
NHW & Hospice Use=Yes	3,974	89,458	<0.001
NHW & Hospice Use=No	11,854	218,085
AI/AN & Hospice Use=Yes	109	2,071	0.099
NHW & Hospice Use=Yes	3,974	89,458
AI/AN & Hospice Use=No	490	8,417	0.155
NHW & Hospice Use=No	11,854	218,085

Bold values indicate statistical significance.

^*Chi-squared test with Yates’ continuity correction (race and hospice indicator)

DISCUSSION

This manuscript presents the findings of a retrospective cohort study using USRDS data among AI/AN individuals diagnosed with ESKD-D nationally. We found that hospice care rates for NHW are higher for non-transplant recipients compared to transplant recipients, but for AI/AN, there was no significant difference in hospice care between the transplant and non-transplant groups. The burdens of SDOH on AI/AN persons not only limits their ability to obtain a transplant,⁷ but also limits palliative and hospice care access once diagnosed with a serious illness, like ESKD.^8–16

Palliative care would ideally begin after a person receives a diagnosis of chronic kidney disease (pre-dialysis). However, there is a persistent perception that palliative care is synonymous with hospice and there is hesitation to start this care because of longevity on dialysis.^37–42 The misunderstanding of what palliative care offers leads to delays in uptake and fails to support adults with ESKD their family caregivers. Despite the benefits that palliative care can offer for QOL for persons with ESKD, typical treatment models for both chronic kidney disease and ESKD in the US do not include palliative care until a person is near end-of-life, when hospice care is most appropriate.^43,44 Furthermore, use of palliative and hospice care for persons with ESKD is much lower compared to the general population and the use statistics decreases further for ESKD patients who have received a kidney transplant.^22,23,45 There has been a persistent lack of access to palliative and hospice care for AI/AN persons living with ESKD despite increasing needs for these supportive services.⁴⁶ This is reflected in our results that show AI/AN persons have lower rates of hospice care use prior to death than NHW persons in both transplant and non-transplant groups. When comparing differences in hospice use between NHW and AI/AN persons across Medicare beneficiaries, 50.8% of NHW receive hospice care while only 33.5% of AI/ANs receive hospice care.⁴⁷ Therefore, it is critical to improve access to palliative and hospice services within AI/AN social and community contexts to improve equity.

In addition to poor access to palliative and hospice care for AI/AN persons, our results show that access to hospice care after previously receiving a kidney transplant is limited compared to NHW persons. Hospice care rates did not show a significant difference for AI/AN transplant and non-transplant groups, but for NHW, the difference between transplant status and hospice care rate was significant. Since the rate of hospice care for AI/AN is roughly equivalent between transplant and non-transplant groups, there appears to be a barrier present that prohibits an increase of hospice care uptake in AI/AN persons in the non-transplant group. This barrier to hospice access is likely related to the significant SDOH that affect AI/AN person’s ability to access transplant (rural and remote geography, extensive poverty, chronically underfunded healthcare services, and racial discrimination^8–16). Prior work has demonstrated that people with ESKD and have previously received a kidney transplant or are on the waitlist receive more intensive healthcare intervention near the end-of-life²³ which may worsen suffering. Furthermore, people who died while on the transplant waitlist were also less likely to receive hospice and have dialysis stopped prior to death.²³ With limited access and late/absent initiation of hospice care near end-of-life, there is a need for a clearer understanding of how advance care planning (ACP) conversations are introduced in the ESKD population.

ACP provides a comprehensive roadmap for healthcare decisions for those who are diagnosed with a serious illness.⁴⁸ It is difficult to provide accurate/confident prognoses for those with ESKD due to barriers like the unpredictable disease course, misconceptions about its role, and insufficient discussions, limiting its integration into disease management despite its relevance in advanced stages of organ failure.⁴⁹ Furthermore, once dialysis is started, it is very difficult to stop regardless of the goals outlined in the ACP.^50–52 Advances to ACP processes are being tested in dialysis settings now,⁵² but widespread ACP is lacking in ESKD populations.^53–55 Even among those who have received a solid organ transplant, there is a reported lack of engagement in advanced care planning despite literature demonstrating clear unmet palliative needs.^23,56 Despite the lack of ACP for persons with ESKD, there has been a positive trend in the place of death. Broadly, more people now are dying in their homes (30.7%) than in the hospital (29.8%) setting, and the number of deaths in hospice facilities is nearly 10% of all deaths.⁵⁷ However, our results indicate regardless of hospice care or transplant status for AI/AN and for NHW with a prior transplant, a greater proportion of people died in the hospital setting compared to the home setting (see Tables 2 & 3), consistent with more intensive healthcare intervention prior to death.²³

Strengths and Limitations

Major strengths of the study include the use of a large national cohort of AI/AN and NHW persons with ESKD-D to answer the underlying question. Use of this dataset allowed for a comparison of transplant and non-transplant recipients receiving hospice care prior to death. The study was not without limitations. The underlying goal of the study was focused on ESKD-D as identified on the CMS Medical Evidence Report (CMS Form 2728).³⁰ Therefore, the findings of this work are limited to diabetic context for AI/AN and NHW persons. Further exploration of hospice care use should be completed for all causes of ESKD and expanded to other minoritized populations to compare how hospice care access differs across different races and ethnicities. Finally, racial misclassification is a known concern using large medical datasets.^58,59 Future studies using death data USRDS should consider matching the USRDS records with the National Center for Health Statistics to account for potential racial misclassifications. Pairing data in this fashion exceeded the scope of this project. To best overcome the limitations of the dataset used for this retrospective study, future work should use Electronic Medical Record data (such as longitudinal data through Epic Cosmos⁶⁰). This will give further context of hospice use timing, palliative care use prior to hospice, to determine if and when ACP discussions occurred, and to better understand the disease trajectory once diagnosed with ESKD.

CONCLUSION

The results of this study confirm the assumptions that there are significant differences in hospice care use among AI/AN versus NHW with ESKD-D, including disparities between those who previously received a kidney transplant. Resistance to use palliative or hospice care with persons who have received a kidney transplant may reflect an underlying thought that those to receive a transplant are willing for aggressive treatment until the end. However, palliative and hospice care are not well integrated into kidney care settings (regardless of transplant status) despite the significant palliative needs experienced by persons with ESKD. These findings reinforce the need to reimagine how palliative and hospice care can be meaningfully implemented for persons with ESKD, especially those who previously received a kidney transplant to reduce the intensity of healthcare received near the end-of-life.