Racial and ethnic inequities of palliative care use among advanced Non‐Small cell lung cancer patients in the US

Abstract Background With early intervention, palliative care (PC) can improve quality of life and increase survival among advanced‐stage non‐small cell lung cancer (aNCSLC) patients. However, PC is often offered late in the cancer treatment course and is underused. We characterized racial/ethnic inequities and the role of healthcare access in PC use among patients with aNSCLC. Methods We used data from the 2004–2016 National Cancer Database, including adults aged 18–90 years with aNSCLC (stage 3 or 4 at diagnosis; n = 803,618). Based on the NCCN guidelines, PC includes non‐curative surgery, radiation, chemotherapy, pain management, or any combination of non‐curative care. We examined PC use by sociodemographic and health care‐level characteristics. To evaluate the independent associations of race/ethnicity and health care access characteristics with PC, we estimated adjusted odds ratios (aOR) with 95% confidence intervals (95% CI). Covariate adjustment sets varied by exposure determined using directed acyclic graphs. Results Our population was 55% male and 77% non‐Hispanic/Latinx (NH)‐White, with a mean age of 68 years. Overall, 19% of patients with aNSCLC used PC. Compared to NH‐White patients, NH‐Black (aOR:0.91,95% CI:0.89–0.93) and Hispanic/Latinx (aOR:0.80,95% CI:0.77–0.83) patients were less likely to use PC, whereas Indigenous (AI/AN) (aOR:1.18,95% CI:1.06–1.31) and Native Hawaiian/Pacific Islander (aOR:2.08,95% CI:1.83–2.36) patients were more likely. Overall, compared to the privately‐insured, uninsured (aOR:1.19,95% CI:1.11–1.28) and Medicaid‐insured patients (aOR:1.19,95% CI:1.14–1.25) were more likely to use PC. Conclusion PC is underutilized among NH‐Black and Hispanic/Latinx patients with aNSCLC. Insurance type may play a role in PC use among patients with aNSCLC.


| BACKGROUND
In the United States (US), lung cancer is the leading cause of cancer-related death in men and women, 1 often due to late-stage disease at detection. When diagnosed at advanced-stage, non-small cell lung cancer (aNSCLC) is a debilitating disease that results in a high burden of symptoms and poor survival; five-year survival after diagnosis of distant NSCLC tumors is only 5%. 2 patients with aNS-CLC frequently experience a high symptom burden during their treatment course, and these symptoms differ across illness trajectories and treatment regimens. 3 For example, following a thoracotomy, pain is reported among 90% of NSCLC patients. 4,5 Other symptoms occurring after surgical interventions may include aching and burning sensations or neuropathic symptoms, reported in 40-69% of patients. 6,7 NSCLC patients also frequently experience cough, dyspnea, wheeze, and hemoptysis, reported in 40 to 85% of patients. 8 High patient reported symptom burden greatly affects quality-of-life and is associated with shorter survival among patients with aNSCLC, underscoring the significance of addressing these outcomes. 9,10 Palliative care can alleviate symptom burden, improve patient quality-of-life, increase overall length of survival, and can also address care satisfaction from the perspective of both the patient and their caregiver. [11][12][13][14][15][16][17] The National Comprehensive Cancer Network (NCCN) recommends intervention with palliative care should begin at cancer diagnosis to provide enhanced patient-centered care and improve aNSCLC outcomes. 18 However, there are significant barriers to timely receipt of palliative care among patients with aNSCLC. The most significant barrier is misconception of palliative treatments among both patients and oncologists; accepting palliative care is frequently conflated with poor prognosis. 19 Due to these barriers racial inequities in palliative care receipt exist in the US; however, existing research has largely focused on palliative care in the context of hospice care or care provided during the end-of-life. 20 Although evidence regarding patterns of palliative care use early in the care continuum specifically is limited, prior studies show that patients from minoritized communities are least likely to both discuss and use hospice care compared to others. 19 Little is known regarding the role of health care-level access factors in the context of palliative intervention use among patients with aNSCLC, specifically when elucidating racial and ethnic inequities. Evaluating patient-level health care access in the context of palliative care is important, as identifying potential health care-level determinants may inform targeted intervention development to improve access to equitable palliative care. Further, it is important to contextualize the societal structures from a health care perspective that influence racial inequities in palliative care use. Differences in cancer care quality by racial/ethnic groups do not have a biological etiology, but rather are manifestations of structural and societal barriers minoritized communities face in the U.S. 21 Our objective was to characterize racial and ethnic inequities in palliative care use among patients with advanced NSCLC. We further estimated odds of receiving curative treatment among those patients who received and did not receive palliative care. We characterized overall and race/ethnicity stratified associations of health care access with palliative care utilization. We considered the following measures of health care access, patient's type of insurance, distance from patient to provider, type of cancer treatment facility, and Medicaid expansion status of the state where treatment occurred. We hypothesized indicators of low health care access, for example being uninsured or living a further distance from their provider, would be associated with lower palliative care use. Insights gained from this research will support the future development of health care system or provider-level interventions to optimize equitable palliative cancer care services.

| Data source
To conduct our analysis, we leveraged the 2004-2016 National Cancer Data Base (NCDB) Participant Use Files (PUF). The NCDB is a US clinical oncology database collecting electronic health record patient-level data of cancer patients treated at one of over 1500 Commission on Cancer (CoC) accredited institutions. 22,23 Achieving CoC accreditation indicates that the hospital is able to provide a multidisciplinary team approach care, long-term patient monitoring and follow-up, access to the latest advancements to cancer treatment, availability of psychosocial support and survivorship for patients, and prioritizes continuous quality improvements in care. The NCDB databases captures over 29 million independent cancer cases, which is estimated to include about 72% of all patients with newly diagnosed cancer. 22 The data included in the NCDB registry are standardized and collected from patient electronic health records. Certified tumor registrars (CTR), trained in cancer registry operations, conduct data abstraction from patient's electronic health records using standardized methods to be included in the NCDB. 24 The study was approved by the Institutional Review Board at Duke University under a general protocol (IRB#: Pro00102834).

| Study population
We included patients with advanced (primary stage III and IV at diagnosis) non-small cell lung cancer (aNSCLC) at cancer diagnosis. Patients diagnosed between January  1, 2004, to December 31, 2016 with NSCLC based on the  International Classification of Diseases for Oncology,  Third Edition topography codes C340, C341, C342, C343,  C348, and C349 were included. We excluded patients with missing or unknown cancer stage (n = 99,804, 6.5%). Additionally, patients with missing data on palliative care utilization (n = 3380, 0.4%) were also excluded.

| Study variables
Our primary outcome was palliative treatment, which is characterized by the NCDB as treatments with noncurative intent based on whether the patient's medical records explicitly mentioned that the goal of the provided treatment is palliation and not cure. Specifically, treatment is categorized as palliative if the treatment was provided to "prolong a patient's life by controlling symptoms, to alleviate pain, or to make the patient more comfortable. 25 " The NCDB ensures that types of palliative care include noncurative pain management, surgery, radiation, or systemic treatment administered to alleviate symptoms. Patients who receive palliative care may also contemporaneously undergo curative cancer treatment. The NCDB includes information on any palliative care from patients' clinical medical records during their treatment at the reporting facility. Data included in the NCDB do not capture hospice services or referral and were not included in the definition of palliative care.
Receipt of first-line curative cancer treatment was defined as surgery, radiotherapy, systemic therapy, or any combination of these therapies. For the present analysis, several health care access variables were evaluated as described here. Type of insurance was defined as the patient's primary payor at the time of first treatment and/or cancer diagnosis. Distance from patient to provider was defined as greatest circle distance or the distance (miles) between the patient's residence (based on the latitude or longitude of the centroid of either the patient's zip code centroid or city if the zip code was not available) and the provider's hospital location (based on hospital street address). Type of cancer treatment facility included community cancer programs, comprehensive community cancer programs, academic or research programs, or integrated network cancer programs. The NCDB includes a measure capturing Medicaid expansion status of state the patient and categorizes states as non-expansion states, early expansion states (2010-2013), January 2014 expanded states, and late expansion states (after January 2014). As the Affordable Care Act was enacted in 2011, we included data starting from January 1st, 2011 to December 31st 2016 when modeling Medicaid expansion as our main exposure. 26 We have included the study characteristics of this sub-sample in Supplementary Table 1.
We defined race and ethnicity using the following categories based on data availability of the NCDB: Non-Hispanic/Latinx White (NH-White), NH-Black, Hispanic/ Latinx, Asian, American Indian/Alaskan Native, Native Hawaiian/Pacific Islander, and other Race. When evaluating individual health care access measures, we compared estimates across racial and ethnic groups, specifically NH-White, NH-Black, Hispanic/Latinx, and Asian patients with aNSCLC with cancer, to ensure adequate sample size for statistical modeling.

| Statistical analysis
We descriptively summarized patient sociodemographics and clinical cancer characteristics as percentages stratified by receipt of palliative care. We also conducted descriptive analyses to characterize racial and ethnic differences in palliative care use using Pearson Chi-squared tests. We evaluated racial and ethnic disparities in palliative care overall and by sex using multivariable logistic regression modeling after adjustment for age (continuous), months from cancer diagnosis at last contact or death, insurance type, median income, Charlson-Deyo comorbidity score, cancer care facility type, census region, year of cancer diagnosis and cancer grade. We also evaluated racial ethnic inequities in first-line curative treatment receipt among patients with aNSCLC overall, and stratified by palliative care receipt status using the same adjustment set. Next, we estimated associations of health care access measures with palliative care use using multivariable logistic regression modeling. To identify the minimal sufficient adjustment sets for each health care access measure exposure model, we used directed acyclic graphs or DAGs 27 constructed based on prior literature describing determinants of palliative care. The final DAGs are included in the Supplementary Material. For each model, we calculated cluster-robust standard errors to account for nonindependence within clusters at the facility level to adjust for correlated patient characteristics within hospitals. All analyses were performed with Stata statistical software, version 15.0 (StataCorp).

| RESULTS
Overall, we included 899,521 patients with aNSCLC with an average age of 67.9 years (SD = 11.1) ( Table 1). About 45% of patients with aNSCLC were female, and 76% were NH-White. Eleven percent of patients with aNSCLC were NH-Black, 3% were Hispanic/Latinx, and T A B L E 1 Characteristics among patients with advanced stage (Stage 3/4 or metastatic) non-small cell Lung Cancer with known palliative care receipt status (n = 899,521) We estimated associations of palliative care utilization by race/ethnicity overall and stratified by sex ( Figure 1). Overall, NH-Black patients had 9% relative lower odds (aOR:0.91, 95% CI:0.89-0.93) and Hispanic/Latinx patients had 20% relative lower odds (aOR:0.80, 95% CI:0.77-0.83) of using palliative care compared to NH-White patients.
Conversely, American Indian/Alaskan Native (aOR:1.12, 95% CI:1.01-1.25) and Native Hawaiian/Pacific Islander (aOR:2.11, 95% CI:1.84-2.41) had higher odds of palliative care use compared to their NH-White counterparts. When stratified by sex, similar trends were observed in across most racial/ethnic categories patients. When we evaluated odds of receiving first-line curative treatment (surgery/radiotherapy/systemic therapy) among all patients with aNSCLC, we observed that compared to NH-White patients with aN-SCLC, NH-Black (aOR:0.91, 95% CI: 0.89-0.92), Hispanic/ Latinx (aOR: 0.72, 95% CI: 0.70-0.74), Asian (aOR: 0.85, 95% CI: 0.82-0.89) adults were less likely to receive first-line curative treatment (Figure 2). These associations were consistent among those who did and did not receive palliative care. these associations were consistent across sex. Patients, including both men and women, residing farthest (over 45 miles) from their provider were 11% less likely to use palliative care compared to those residing <2 miles from their provider (aOR:0.89, 95% CI:0.80-0.98). There were no associations between cancer treatment facility type and palliative care utilization. Across racial/ethnic categories, patients with aNS-CLC were mostly Medicare insured or insured with private insurance/managed care. NH-Black and American Indian/Alaskan Native patients with aNSCLC mostly (>60%) lived in areas with low median household income, and over half lived in states without Medicaid expansion (Table 3). Table 4  were also more likely to utilize palliative care compared to privately insured patients. Compared to patients treated at a comprehensive community cancer program, our data suggest Hispanic/Latinx (aOR:1.40, 95% CI:1.05-1.84) patients with aNSCLC treated at Integrated Network Cancer Programs were more likely to use palliative care.

| DISCUSSION
Despite the documented benefits of palliative care toward improved survival and higher patient-reported quality of life, our study demonstrates that only about one in five patients with aNSCLC utilized palliative care in the US. NH-Black and Hispanic/Latinx patients with aNSCLC were less likely to utilize palliative care compared to NH-White patients, which suggests inequitable access to high-quality cancer care. Inequities observed across racial/ethnic groups in cancer care may stem from structural and societal barriers to equitable access to health care in the US. As such, it is important to characterize the role different aspects of social determinants may play in access to care. Our analyses suggest several health care access measures may influence palliative care use among patients with aN-SCLC, including insurance status and the type of cancer treatment facility. Identification of multi-level determinants of palliative care use introduces the opportunity for intervention development to improve palliative care uptake and deliver high-quality equitable care.
The benefits of early palliative care intervention among patients with cancer were first demonstrated among patients with metastatic non-small cell lung cancer in a randomized controlled trial. 13 The trial demonstrated that patients with aNSCLC who received early palliative care had improved survival (11.6 months vs. 8.9 month), a better quality of life and reported fewer symptoms of depression than patients who received standard of care (i.e. no palliative care). Despite these documented benefits, we found that few patients with aNSCLC used palliative care. Prior research evaluating the prevalence of palliative care use among patients with aNSCLC varies, potentially due to underlying differences in the demographic distribution of the patient population. For example, in a recent study of patients with advanced-stage lung cancer receiving care at the Veterans Health Administration (i.e. in patients insured by the Veterans Affairs), more than half of patients diagnosed between 2007 to 2013 received palliative care. 16 Another study leveraging data from Surveillance, Epidemiology, and End Results-Medicare found that 15% of Medicare-insured metastatic NSCLC patients received palliative care within the first year after diagnosis, which is similar to our findings. 28 Given the significant symptom burden experienced by most patients with aNSCLC, 29 the low use of palliative care in the present study and prior research is of concern. However, we were unable to investigate timeliness of palliative care delivery due to lack of data characterizing date of palliative care provided or sequence of palliative care provision with concurrent curative care. Future work to delineate potential racial/ethnic inequities in time to palliative care receipt among metastatic cancer patients is important and should be prioritized to inform intervention development. Racial/ethnic inequities exist in palliative care use among patients with aNSCLC, suggesting inequitable delivery of high-quality cancer care. Our study demonstrates that NH-Black and Hispanic/Latinx patients with aNS-CLC are less likely to use palliative care compared to their NH-White counterparts. Prior work has focused on racial inequities of palliative care use in the context of end-of-life care, [30][31][32][33] and similar to our work, demonstrated that NH-Black patients are less likely to receive referral to hospice or to receive palliative care near the end-of-life. Several reasons have been attributed to the inequities observed in palliative care use at the end-of-life among patients with cancer from minoritized communities, including poor patient-provider communication regarding the role of palliative care. 34 Provider training and ability to initiate timely discussions regarding end-of-life care and decisions toward fostering positive quality-of-life through hospice or complementary supportive care is critical. Although providers support early intervention with palliative care, less than one in five report they would refer aNSCLC patient at diagnosis. 35 In the context of caring for patients with cancer from minoritized communities, providers are shown to apply implicit bias during clinical-decision making and may assume patients have limited ability to understand medical decision-making. 36 In addition to provider-level barriers to palliative care use, patient-level attitudes and perceptions regarding the role of palliative care and treatment near the end-of-life can also impact racial/ethnic minority patient's decisions. Recent work has focused on the significance of palliative care use in the When contextualizing the results of these analyses, it is important to consider the types of patients captured in the NCDB. Although the NCDB captures 70% of patients from diverse regions of the country and a large majority of aNSCLC cases in the US, the NCDB is not populationbased. The NCDB only captures patients who received care at a Commission on Cancer (CoC)-accredited US hospital, which includes over 1500 programs in the US and Puerto Rico. The CoC is a consortium of 56 professional organizations, such as the American Cancer Society, and the Centers for Disease Control and Prevention, which reviews the services and resources available at a specific hospital to ensure they address the full spectrum of the cancer care continuum from prevention to end-of-life. As such, CoC-accredited hospitals are not representative of all clinics in the US, particularly clinics where palliative care may not be made available to patients. However, despite the high-resource setting of CoC hospitals, it is still important to acknowledge that the data on palliative care services are of uncertain accuracy. Non-curative surgery, radiation, and systemic therapy as documented by the NCDB may be considered as palliative care resources and do not meet the criteria for early intervention with palliative care. Further, information regarding hospice referral is unavailable and should be evaluated in future studies focused on palliative care use across the care continuum, including end-of-life.
In the U.S, lung cancer is one of the most common cancers among adults and the leading cause of cancer-related death. Racial inequities exist in mortality and outcomes among patients with lung cancer, particularly affecting NH-Black and Hispanic/Latinx adults. Opportunities exist for improving outcomes after lung cancer diagnosis through palliative care, as demonstrated by the survival benefits when palliative care is introduced early in the cancer treatment trajectory of a patient. However, we continue to observe racial inequities in receipt of palliative care, with only 20% of all adults diagnosed with aNSCLC using palliative care. Future research integrating implementation science principles into intervention development to address palliative care inequities should be prioritized. In particular, interventions informed by research characterizing palliative care use among patients with aNSCLC receiving curative care or those who have declined curative treatment would be of high priority given the potential missed opportunity to treat curable cancers if diagnosed at an early stage. High-quality and equitable cancer care delivery through equitable use of palliative care among patients with cancer should be prioritized in the US to mitigate the long-standing effects of racism and historic structural barriers to access to care.

FUNDING INFORMATION
This project was conducted without funding support.

SUPPORTING INFORMATION
Additional supporting information can be found online in the Supporting Information section at the end of this article.