INTRODUCTION

The Asian American population numbers approximately 20 million and represents the fastest-growing ethnic minority in the United States (US).1 Accompanying this rapid growth is the increasing number of Asian individuals facing serious illness and receiving end-of-life (EOL) care.2 Although EOL disparities are evident for all racial and ethnic minorities,3, 4 most efforts to understand and alleviate these disparities have focused on African American and Hispanic communities.5,6,7,8,9,10,11,12 As a result, there remains a paucity of literature on EOL experiences of Asian Americans.

The limited evidence describing Asian Americans’ EOL care reveals differences when compared with White Americans. In the last months of life, research demonstrates that Asian Americans are less likely to use hospice, more likely to be admitted to the intensive care unit (ICU), and more likely to die in hospital.13,14,15,16,17,18,19 However, the generalizability and applicability of these study findings may be limited by geographic restraints,15, 19 temporal scope,13, 16, 17and disease type.14, 18 Furthermore, it is unclear how national policies (e.g., Patient Self-Determination Act) and the growing palliative care presence may have impacted these EOL differences.21 As a result, there is a pressing need to appraise the progression and current state of medical care for Asian Americans at the EOL.

Among medical interventions used at the EOL, invasive mechanical ventilation (IMV) holds a controversial role. Despite its life-saving potential in the setting of acute respiratory failure,25 its use in the setting of serious illness is associated with increased healthcare costs,26 delirium,27 in-hospital death,26 and difficult decisions to withdraw life-sustaining intervention.28 Furthermore, emerging evidence suggests that high intensity, life-sustaining treatment at the EOL may be discordant with the wishes of Asian Americans.22,23,24 Given the potentially burdensome nature of IMV at the EOL, we sought to investigate the differences in IMV use between Asian and White Medicare beneficiaries. Specifically, we aim to (a) compare IMV use between Asian and White Medicare beneficiaries hospitalized in the last 30 days of life, (b) compare IMV use between Asian and White Medicare beneficiaries across select admitting diagnoses, and (c) explore differences in EOL care (i.e., in-hospital death, hospice enrollment, and health care transitions) between Asian Americans and White Medicare decedents.

METHODOLOGY

Data and Study Population

We conducted a retrospective cohort analysis of Medicare fee-for-service beneficiaries aged ≥ 66 years who had a hospital admission in the last 30 days of life. The Medicare fee-for-service cohort represented a 20% random sample of Medicare Parts A and B beneficiaries who died between 2000 and 2017. Due to the increasing enrollment of beneficiaries in Medicare Advantage plans, we also identified a 20% random sample of Medicare Advantage decedents from 2011 to 2017 for sensitivity analysis.29 Within each of these two cohorts, we identified four sub-cohorts based on primary admitting diagnosis using International Classification of Diseases, Ninth (ICD-9-CM) and Tenth Revisions (ICD-10-CM) codes.30 Whereas we identified congestive heart failure (CHF), chronic obstructive pulmonary disease (COPD), and cancer based on primary admitting diagnosis, we identified dementia based on the primary or first nine secondary admission diagnosis codes. Consistent with prior literature, we excluded patients with an admitting diagnosis of cardiac arrest, evidence of pre-existing tracheostomy, a death date between October 1, 2015, and December 31, 2015, and those who were hospitalized in institutions with less than five beneficiary hospitalizations per year.30 The Institutional Review Boards of Oregon Health and Science University and Brown University approved this study and waived patient consent requirements.

Measures

We used validated ICD-9/10CM procedure codes (96.7x/5A1935Z,5A1945Z,5A1955Z) to identify IMV use among hospitalized patients in the last 30 days of life.31 We extracted socio-demographic characteristics from the Medicare Beneficiary Enrollment file which included age, race/ethnicity, sex, residence, and Medicaid-eligibility status. Race/Ethnicity data were based on voluntary, self-reported data collected by the Social Security Administration (e.g., White, Asian, etc.);32 we excluded beneficiaries with “unknown” or “other” ethnicity. We abstracted comorbidities from ICD-9/10-CM codes listed on the bill for hospitalization in the last 30 days of life. We also obtained additional EOL care measures from Medicare claims data include in-hospital death, admission to intensive care unit (ICU), hospice enrollment at death, hospice enrollment in the last 3 days of life, and healthcare transitions in the last 3 days of life. Finally, we obtained hospital-level characteristics (e.g., hospital care intensity index,33 hospital bed size, facility type, hospital deaths, ICU use) by linking the Medicare Provider Number of the hospital bills with the publicly available 2017 Dartmouth Atlas Hospital Care Intensity data.34

Analysis

We used descriptive statistics to characterize patient demographics by ethnicity and IMV use with 95% CIs calculated using Clopper-Pearson methods. We interpreted comparisons as different if their 95% CIs did not overlap. We performed multivariable random-effects logistic regressions to compare IMV use between Asian and White fee-for-service decedents clustered by hospital and adjusting for age, sex, admitting diagnosis, and Charlson comorbidity index.35 We adjusted for secular trends by creating dichotomous year of death variables with the year 2000 as the reference point.30 We also conducted subgroup analyses to investigate the association of Asian ethnicity with IMV use across admitting diagnoses (CHF, COPD, cancer, and dementia), and we tabulated EOL care measures by ethnicity for descriptive comparisons.

We performed sensitivity analyses to account for potential variations across Medicare plans. Using a cohort of Medicare Advantage decedents who died between 2011 and 2017, we conducted multivariable random-effects logistic regressions to compare IMV use between Asian and White decedents overall and across admitting diagnoses. We also captured EOL care measures for descriptive comparisons between ethnicity and with the Medicare fee-for-service data.

We conducted additional exploratory analyses to account for hospital-level variations in care. Pooling Medicare Advantage and fee-for-service decedents from 2011 to 2017, we explored both a random-effects and a fixed-effects model. The random-effects model examined the overall or between-hospital variation in IMV use, and the fixed-effects model examined the association of Asian ethnicity and IMV within a hospital.36, 37 These regression analyses adjusted for Medicaid-eligibility status in addition to sociodemographics, admitting diagnosis, comorbidities, and secular trends. Additionally, we tabulated hospital-level characteristics of the pooled decedents using chi-squared, or independent sample t tests, as appropriate. Data analysis was done using Stata.v15.0 (StataCorp).

RESULTS

Our Medicare sample included 2,142,329 White fee-for-service (55.1% female; mean [SD] age, 82.4 [8.1] years), 28,430 Asian fee-for-service (50.8% female; mean [SD] age 82.6 [8.1] years), 275,645 White Advantage (52.2% female; mean [SD] age, 81.8 [8.2] years), and 6,392 Asian Advantage (48.8% female; mean [SD] age, 82.9 [8.5] years) decedents who were hospitalized in the last 30 days of life (Table 1). Pneumonia/sepsis represented the leading cause of admission for all cohorts, followed by cancer, congestive heart failure (CHF), and chronic obstructive pulmonary disease (COPD). Charlson comorbidity scores did not differ significantly between cohorts. However, Asian decedents enrolled in both Medicare fee-for-service and Advantage plans had a larger proportion of Medicaid-eligible beneficiaries compared to White decedents (Table 1).

Table 1 Characteristics of Medicare Decedents by Plan, Race, and Reception of Invasive Mechanical Ventilation
Full size table

Comparison of IMV Use between Asian and White Medicare Fee-for-Service Decedents

Overall, 16.3% of White and 25.9% of Asian Medicare fee-for-service decedents received IMV, a 9.6% absolute difference (Table 1). From 2000 to 2017, the proportion of IMV use increased for both White fee-for-service (% receipt=14.7% vs 19.2%) and Asian fee-for-service (% receipt=25.0% vs 26.5%) decedents (Fig. 1). Compared to Whites, Asian fee-for-service decedents had an increased adjusted odds ratio (AOR) of 1.42 (95%CI: 1.38–1.47) for IMV. In sub-analyses, Asians’ AOR for IMV trend higher for non-cancer diagnoses (congestive heart failure AOR=1.75, 95%CI:1.47-2.08; dementia AOR=1.93, 95%CI: 1.70–2.20; and chronic obstructive pulmonary disease AOR=2.25, 95%CI: 1.76–2.89) than cancer (AOR=1.32, 95%CI: 1.15–1.51) (Fig. 2).

Figure 1
figure 1

IMV use at the end of life for hospitalized Asian and White Medicare fee-for-service from 2000 to 2017 and MA decedents from 2011 to 2017. Overall, 16.3% of White fee-for-service (solid black line), 25.9% of Asian fee-for-service (dashed black line), 19.9% White Medicare Advantage (solid grey line), and 28.1% Asian Medicare Advantage (dashed grey line) decedents received IMV. There was an increase in the proportion of White fee-for-service (15.2 to 19.2%), Asian fee-for-service (25.0 to 26.5%), White Medicare Advantage (19.3 to 20.7%), and Asian Medicare Advantage (27.9 to 28.5%) decedents receiving IMV from 2000 to 2017. Abbreviations: IMV, invasive mechanical ventilation.

Full size image
Figure 2
figure 2

Forest plot showing the adjusted odds ratio of hospitalized Asian Medicare Advantage and Fee-for-Service beneficiaries for IMV at the EOL, by admitting diagnosis and Medicare plan. Comparison to White beneficiaries is denoted by the black vertical line corresponding to an adjusted odds ratio of 1.00. Each odds ratio is derived from a random-effects logistic regression model adjusting for age, sex, admitting diagnosis, comorbidities, and secular trends. Adjusted odds ratio for Medicare Advantage decedents also accounted for Medicaid-eligibility. Abbreviations: IMV, invasive mechanical ventilation; EOL, end-of-life; and CI, confidence interval.

Full size image

Comparisons of End-of-Life Care between Asian and White Medicare Fee-for-Service Decedents

Compared to White decedents, Asian fee-for-service decedents had higher rates of in-hospital deaths (62.1%, 95%CI: 61.6–62.7 vs 51.4%, 95%CI: 51.3–51.4), intensive care unit admission (52.7%, 95%CI: 52.1–53.2 vs 40.1%, 95%CI: 40.0–40.2), lower rates of hospice enrollment (22.5%, 95%CI: 22.0–23.0 vs 31.6%, 95%CI: 31.6–31.7), and hospice enrollment in the last 3 days of life (8.9%, 95%CI: 8.6–9.3 vs 13.4%, 95%CI: 13.3–13.4). Among decedents who survived their hospitalization, Asian and White decedents had similar rates of health care transitions in the last three days of life (36.9%, 95%CI: 36.0–37.8 vs 35.8%, 95%CI: 35.8–36.0) (Fig. 3).

Figure 3
figure 3

Among White and Asian decedents hospitalized at the EOL, in-hospital death, ICU/CCU admission, hospice use, and healthcare transitions in the last 3 days by Medicare plan. Abbreviations: EOL, end-of-life; ICU, intensive care unit; and CCU, cardiac care unit.

Full size image

Sensitivity Analysis Involving Medicare Advantage and Fee-for-Service Decedents from 2011 to 2017

Overall, sensitivity analyses using Medicare Advantage data did not reveal significant changes in the direction of effects. Similar to fee-for-service decedents, Asian Medicare Advantage decedents had consistently higher rates of IMV from 2011 to 2017 (Fig. 1). Compared to White decedents, Asian Medicare Advantage decedents had higher AOR (1.66, 95%CI: 1.55–1.77) for IMV. Furthermore, Asian Medicare Advantage decedents’ AORs for IMV were higher for non-cancer diagnoses than for cancer (Fig. 2). In terms of EOL measures, Asian Medicare Advantage decedents had higher rates of death in hospital (53.6%, 95%CI: 52.3–54.8 vs 43.0%, 95%CI: 42.8–43.2), higher ICU admission (58.5%, 95%CI: 57.3–59.8 vs 56.8%, 95%CI: 56.7–57.0), lower hospice enrollment at death (32.3%, 95%CI: 31.1–33.5 vs 44.6%, 95%CI: 44.4–44.8), and lower hospice enrollment in the last 3 days of life (12.2%, 95%CI:11.4-13.0 vs 19.3%, 95%CI: 19.2–19.5) when compared with White decedents (Fig. 3). Different than fee-for-service decedents, however, Asian Medicare Advantage decedents had lower rates of healthcare transitions in the last 3 days compared to White decedents (41.0%, 95%CI: 39.8–42.2 vs 45.6%, 95%CI: 45.4–45.8) (Fig. 3).

Exploratory Analyses with Pooled Medicare Advantage and Fee-for-Service Decedents from 2011 to 2017

In pooled analyses, Asian decedents had a higher likelihood of receiving IMV than White decedents in both the random-effects (between-hospital) model (AOR=1.55, 95%CI: 1.49–1.61) and fixed-effects (within-hospital) model (AOR=1.51, 95%CI: 1.45–157). Nearly one-fifth (ρ=17.7%, 95%CI: 16.8–18.7%) of the variance for IMV use in the random-effects model were attributed to care variations across hospitals. An exploratory analysis of the hospitals attended by patients in 2017 revealed that Asian beneficiaries were more likely than White beneficiaries to receive their EOL inpatient care in the state of California, academic teaching hospitals, and be enrolled in integrated health care systems (Table 2). Furthermore, the hospitals attended by Asian beneficiaries had higher Dartmouth Atlas hospital care intensity scores, more inpatient beds, more in-hospital deaths, and more in-hospital deaths preceded by an ICU stay (Table 2).

Table 2 Characteristics of Hospitals that Asian and White Medicare Decedents Patronized in the Last Month of Life in 2017
Full size table

DISCUSSION

Achieving equitable and high-quality EOL care for racial and ethnic minorities represents a national objective of the US healthcare system.38 Despite prior evidence demonstrating that ethnic minorities experience disparate care at the EOL,3 little is known about the care Asian Americans receive. Our findings demonstrate that older Asian Americans disproportionately experience high-intensity care when hospitalized at the EOL. Specifically, when compared to White beneficiaries, Asian Medicare beneficiaries hospitalized at the EOL are more likely to receive IMV and have higher rates of ICU admissions, in-hospital deaths, and lower rates of hospice use. Confidence in these results is bolstered by the consistency of our findings across diagnoses, Medicare plans, and adjustment for within- and between-hospital effects. Although our findings likely raise more questions than answers, the striking differences in care suggest that older Asian adults may disproportionately receive burdensome care when hospitalized in the last month of life.

Our findings are consistent with and extend results from prior reports by offering a population-level perspective that accounts for relevant sociodemographics, secular trends, admitting diagnoses, and hospital-level variations in care. The findings that hospitalized Asian Medicare beneficiaries have a higher likelihood of IMV use both within a hospital (i.e., fixed-effects model) and between hospitals (i.e., random-effects model) suggest that the observed EOL differences likely extend beyond individual preferences to social, institutional, and systemic factors.39

At the individual and social level, the increased IMV use may reflect Asian patients’ and caregivers’ attitudes toward serious illness care. Existing literature reveals mixed attitudes of older Asian patients toward life-sustaining treatments,24, 40,42,42 which appear to be influenced by acculturation, religious affiliation, and traditional Asian philosophies.43 These mixed attitudes contrast with our findings that the hospitals attended by older Asian patients exhibit higher care intensity metrics. A possible explanation may be that Asian Americans are geographically restricted from accessing institutions that provides lower intensity of care at the end-of-life. Another explanation may be that the observed differences reflect the preferences and influence of family caregivers. While family is consistently endorsed as a key player in EOL decision-making among Asian Americans,44,46,47,48,49,49 intergenerational differences in acculturation level complicate our understanding of familial roles at the EOL.50 Future research into the mechanism by which Asian families make EOL decisions will enable the delivery of family-centered, goal-concordant, and culturally tailored EOL care.

At the institutional level, we found that Asian Medicare beneficiaries, who are hospitalized at the EOL, are more likely to receive high-intensity care both within and across hospital settings. This pattern of healthcare utilization may relate to differential engagement with palliative care services.3 While researchers have consistently linked palliative care with higher quality and lower intensity EOL care,51,53,54,55,55 few have investigated the impact of palliative care on Asian Americans.56 What is known is that Asian Americans continue to face many barriers in accessing palliative care services, including under-reporting of symptoms, low engagement with advance care planning, low knowledge of available end-of-life services (e.g., hospice), and significant communication barriers (e.g., language discordance).57 These barriers are reinforced by clinicians’ implicit biases toward and unfamiliarity with Asian cultural nuances during EOL conversations.58, 59 Thus, an examination of the pattern of use and effectiveness of palliative care among Asian Americans may be important in exposing inequities in palliative care access and engagement.

Finally, at the policy level, several forces may contribute to the sustained differences in EOL care intensity for Asian Americans. First, high rates of misclassification and aggregation of Asian Americans into “other” and “ethnic minorities” result in under-representation of Asian Americans in large databases.60 Second, reporting of Asian American data in aggregate may mask important health disparities among the diverse and heterogeneous Asian sub-populations.60, 61 Third, historical discriminatory immigration policies and the Asian model minority myth (i.e., above-average achievement in education and socioeconomic status) perpetuate the stereotype of Asian Americans as the outsider.62, 63 Together, the under-representation, aggregation, and alienation may have increased the difficulty and diminished societal interest in understanding health disparities Asian Americans’ experience at the EOL. As periods of instability have historically resulted in scapegoating of minority groups, policymakers should be vigilant of the potential for the widening of EOL health disparities for Asian Americans during the COVID-19 pandemic.

LIMITATIONS

This is the first nationally representative cohort analysis of Asian Medicare beneficiaries’ use of intensive care and hospice services when hospitalized at the EOL, yet it has limitations. First, our data should be interpreted in the context of hospitalized, older adults at the EOL. Although more than half of Medicare beneficiaries received hospitalization in the last month of life,64 our findings may not generalize to younger patients and individuals who were not hospitalized in the last 30 days of life. Second, our study was unable to capture patient nor family preferences for EOL care. This limits our understanding of whether the observed differences in care were goal-concordant. Third, we were unable to account for the severity of disease at admission, acculturation level, religious affiliation, and education. This may have biased the findings away from null. Finally, as Asian Americans were evaluated as an aggregate group, it is unclear how our findings may apply to individual Asian sub-populations.

CONCLUSION

From 2000 to 2017, persistently higher rates of IMV use were observed among Asian American Medicare beneficiaries when hospitalized at the EOL. These differences in IMV use are accompanied by higher rates of other measures of potentially burdensome EOL care including ICU admission, in-hospital death, and lower rates of hospice use among Asian Americans. While observed differences are partly explained by sociodemographics and hospital-level variability, they may also signal the presence of underlying health disparities. Further research is needed to understand these potential EOL care disparities by exploring patient/caregiver shared decision-making preferences, evaluating the availability and effectiveness of palliative care services, and dis-aggregating Asian Americans into sub-populations.