Ethnicity and insurance status predict metastatic disease presentation in prostate, breast, and non‐small cell lung cancer

Abstract Background Ethnicity and insurance status have been shown to impact odds of presenting with metastatic cancer, however, the interaction of these two predictors is not well understood. We evaluate the difference in odds of presenting with metastatic disease in minorities compared to white patients despite access to the same insurance across three common cancer types. Methods Using the National Cancer Database, a multilevel logistic regression model that estimated the odds of metastatic disease was fit, adjusting for covariates including year of diagnosis, ethnicity, insurance, income, and region. We included adults diagnosed with metastatic prostate, non–small cell lung cancer (NSCLC), and breast cancer from 2004 to 2015. Results The study cohort consisted of 1 191 241 prostate cancer (PCa), 1 310 986 breast cancer (BCa), and 1 183 029 NSCLC patients. Private insurance was the most protective factor against metastatic presentation. Odds of presenting with metastatic disease were 0.190 [95% CI, 0.182‐0.198], 0.616 [95% CI, 0.602‐0.630], and 0.270 [95% CI, 0.260‐0.279] for PCa, NSCLC, and BCa compared to uninsured patients, respectively. Private insurance provided the most significant benefit to non‐Hispanic White PCa patients with 81% reduction in odds of metastatic presentation and conferred the least benefit to African‐American NSCLC patients at 30.4% reduction in odds of metastatic presentation. Conclusions Insurance status provided the single most protective effect against metastatic presentation. This benefit varied for minorities despite similar insurance. Reducing metastatic disease presentation rates requires addressing social barriers to care independent of insurance.


| INTRODUCTION
While overall cancer mortality continues to decline, patients who present with metastatic disease pose a complex clinical challenge. 1 Across all cancers, metastatic presentation portends worse prognosis and commonly renders a patient incurable. Access to early prevention and diagnostic services is critical in preventing late stage presentation and improving survival outcomes. 2 Lack of insurance is a significant barrier to both early diagnosis and potential cure. 3 However, access to care is mediated by factors beyond insurance status, including social determinants of health, namely, ethnicity, income, and geography. 4,5 Providing health insurance to the previously uninsured aims to bridge the gap in cancer care disparities along socioeconomic strata. 6 Recent studies reveal increased health services utilization in states with Medicaid expansion. 7,8 Improved management of chronic illnesses such as diabetes illustrates the benefits of expanded insurance access. 9 Cancer stage at diagnosis could serve as an indicator for utilization of screening and protend success in curing early-stage cancers. 10,11 Recent studies suggest that Affordable Care Act's Medicaid expansion show a greater rates of breast cancer screening and early-stage diagnosis. 12 Despite increased access to insurance, delivery of high-quality care continues to be hampered by entrenched obstacles in minority communities largely due to social determinants of health. 13 While ethnic minority and lack of insurance have historically predicted for inferior outcomes, there is a paucity of studies examining the interplay between insurance and ethnicity as related to metastatic disease presentation. Furthermore, access to insurance does not always result in utilization of health services, therefore, it is critical to assess the needs of a population not solely based on their insurance status but also social determinants of care. To achieve cancer care equity, it is crucial to fill this knowledge gap by evaluating potential determinants of disparities among diverse patients with a range of insurance access and cancer types. Our hypothesis is that insurance or lack thereof by itself does not lead to favorable cancer presentation. Although we expect this effect to be most pronounced in minority patients who lack insurance, we anticipate that the effect will be present in other forms of insurance holders as well. The goal of our analysis is to identify trends in metastatic disease presentation and examine the role of ethnicity in conjunction with insurance.

| Data source
The NCDB is a prospectively collected, hospitalbased registry populated by data generated from the Commission On Cancer (CoC) accredited centers in the US, sponsored by the American College of Surgeons and the American Cancer Society. The database captures approximately 70% of cancers diagnosed in these hospitals. Significant data points related to patient's disease, treatment, and demographic features along with estimates of income and education levels by corresponding zip codes are included.

| Ethics statement
Ethical approval was sought and granted from the Institutional Review Board (IRB). After acquiring an IRB exemption, the NCDB for Breast, Prostate and Lung cancers was accessed. The study conforms to Consolidated Standards of Reporting Trials (CONSORT) guidelines. The diagrammatic flowchart is appended in the supplementary material for each disease site ( Figure S1).

| Study cohort
The NCDB for breast, prostate, and lung cancer was evaluated to identify patient-related factors that affect metastatic cancer presentation despite access to insurance. Breast and Prostate cancer are unique in that both can be diagnosed early through readily available screening which conceivably would reduce the rate of metastatic presentation. 14,15 Alternatively, lung cancer typically presents as metastatic. We selected NSCLC (non-small cell lung cancer) as it is the most common subtype. We sought to contrast highly prevalent cancer types with various risk factors, gender specificity, and potential for early detection and cure.
The NCDB for prostate, breast, and lung cancers were queried for all patients ≥ 18 years old from 2004 to 2015. Clinical stage was determined using the 7th Edition of American Join Committee on Cancer (AJCC) staging manual. Patients with noninvasive histology and missing demographic data were excluded. Additional covariates analyzed include age, gender (dichotomized for lung cancer), insurance status, and Charlson Deyo Comorbidity scores. The facility types in the NCDB are as follows: community cancer program, comprehensive community cancer program, academic/research program (includes NCI designated comprehensive cancer centers), integrated network cancer programs, and "other/unknown." Outcome variable was the rate of metastatic presentation. This was to capture the most severe instance of cancer care disparity at the onset of an irreversibly poor prognosis. The rate of stage IV presentation is assumed in this analysis to provide a composite endpoint for available determinant of disparity in the database.

| Statistical analyses
Comparison of demographic features was made for each site between patients with localized disease (stages I-III) versus those with metastatic disease (IV). Bivariate analysis was conducted using the Pearson Chi-Square test to identify differences in sociodemographic features. Multivariable logistic regression was utilized to assess the direct association of insurance with metastatic presentation. Adjusted odds ratio for insurance status, with and without ethnic stratification, was calculated while controlling for facility type and region, median income, comorbidity score, education, demographic area, and year of diagnosis. P< 0.001 were considered significant. Finally, subgroup analysis was conducted for each cancer site based on ethnic strata and insurance status with a separate logistic regression model adjusted with the above demographic features within each ethnic cohort. All statistical analyses were performed using SAS 9.4 (Cary, NC). For specific P-values please refer to tables. Patients without available data points for the sociodemographic features used in the analysis were censored per diagram ( Figure S1).  23.57% vs NHW: 9.66%). In breast and prostate cancer, patients who presented with metastatic disease were more likely to reside in zip codes with lower educational attainment and median income. There appeared to be relative parity in these features regardless of stage at presentation for NSCLC patients.

| Analyses of effect within each ethnicity identify Insurance as a significant predictor of Metastatic presentation
To determine the role of insurance in odds of metastatic presentation in each cancer site, we analyzed each ethnic group separately stratified by insurance status using a multilevel logistic regression model.

| DISCUSSION
In this analysis of the National Cancer Database, we have consistently identified the effect of insurance status in predicting metastatic presentation in three common cancer types (breast, prostate and non-small cell lung cancer). Additionally, insurance status and ethnicity affect the odds of presenting with metastatic cancer in all three sites. Finally, the degree of advantage conferred by insurance status appears to differ across ethnicities. A decade ago, a landmark study addressed this very question in multiple cancers using the NCDB. 16 Our results, using a subset comprised of fewer cancer types, mirror those findings and, significantly, reveal a persistent trend reflecting disparity despite comparable insurance status that exists across years and disease sites.
Our findings suggest that insurance coverage is a powerful protective factor against metastatic cancer presentation. Access to healthcare can be a critical barrier for cancer patients and insurance status has previously been linked to patients presenting with more advanced disease across different cancer types. 16  In our study, Medicaid recipients had decreased odds of metastatic presentation in all three cancers compared to uninsured patients but consistently lagged patients with private insurance. This is in line with the fact that Medicaid recipients typically have access to inferior care compared to those with private insurance. 17,18 This may partially leads to poorer quality of care for minority patients, given that a higher proportion of these patients receive Medicaid. 19 Moreover, minority patients, who historically have had worse cancer outcomes due to a multitude of factors including limited or inferior access to health services, appear to not benefit as much from insurance access as NHW patients even adjusting for available demographic features. It is therefore imperative that any effort to address cancer care disparity attempts to understand the factors which impede access to early detection in minority populations aside from insurance.

Baseline characteristics
Minority patients' patterns of cancer screening utilization and preventative health behaviors differ from that of NHW patients both in frequency of access and quality of care. 20 at the same rate as their NHW counterparts 22 or undergone "high-quality" mammograms. 23 Likewise, in prostate cancer, AA men undergo less-frequent PSA screening -a critical difference as they are 70% more likely to be diagnosed with prostate cancer and have a PCa mortality rate 2.4 times that of NHW. 24 Compounding missed opportunities for early detection and timely intervention are systemic disparities, including that minority-serving hospitals often lack dedicated cancer screening programs with expertise and technologies comparable to the academic centers serving more affluent populations. 25 Recent evidence has shown minority patients' care differs even within the same hospital compared to NHW patients. [26][27][28] Pain and neurological symptoms are more commonly overlooked in AA and H patients, a lapse that may lead to dismissal of nonspecific symptoms that could warrant further workup. 29,30 Our results complement similar studies showing, despite the same insurance, disparities along ethnic lines exist among patients whether in rates of metastatic presentation or treatment utilization. Attitude of health-care providers toward minority patients, 31 deficiencies in cancer prevention in minority serving hospitals, and possible discrepancy in health literacy in minority populations 32 can be additional confounders in metastatic cancer presentations.
We included NSCLC, a site with significant risk of metastasis at presentation, which differs from the other two cancer types in that screening tests are in their infancy and not widely utilized. While low-dose CT screening is promising, 33 a majority of patients present with metastatic disease in the NCDB, likely due to pathophysiology of disease overwhelming other demographic features. However, non-White patients consistently continue to have a higher risk of presenting with metastatic cancers. Interestingly, private insurance offers the smallest risk reduction against metastatic presentation of NSCLC for all ethnic groups.
Significant efforts have been undertaken to define biological determinant of health-care disparity. However, these biologic factors cannot fully explain the rates of metastatic presentation in various ethnicities. 34,35 For example, AA patients experience different odds of metastatic prostate cancer presentation in different regions (Table 1 and Figure S2). It is unlikely that a unique underlying biological driver can account for geographic disparity.
The disproportionate rates of metastatic presentation in minority patients suggest that barriers to early detection are rooted in a complex set of factors related to region and quality of insurance. Protective effects of insurance are tempered by structural obstacles facing minority patients, and providing insurance, albeit a critical step, may not sufficiently alter odds of metastatic presentation without addressing barriers to early detection, quality primary and oncologic care, and improved health literacy.

| STRENGTHS AND LIMITATION
The strength of this analysis lies in the use of a large, curated database serving a diverse cohort of patients representing the majority of cancer diagnosis in the US population. The NCDB enables direct examination of the role of various insurance types on specific cancer presentation accounting for patient-level and sociodemographic features. Two significant developments highlight NCDB's usefulness since the prior publication addressing issues of insurance and ethnicity came out a decade ago. Since 2009, CoC accredited centers were required to report metastatic presentation in patients at the time of diagnosis. Additionally, with increasing participation of minorities in diverse geographical regions, a more wholly representative demographic is captured. 36,37 Nevertheless, this analysis has limitations inherent to any retrospective database research. Several unmeasured confounders are present. There are significant differences in the experience of Hispanic and African American patients in the health-care system. Language barriers, institutional biases, and immigration status can define the experiences of some while not pose a barrier to others. Additionally, self-reported ethnicity could be difficult to capture in such large datasets. African-American ethnicity may be reported adequately while Hispanic patients may be misrepresented. Additionally, while zip code-level data rarely provide a complete view of a single patient's social setting, in this and many other reports using NCDB, the census-based demographics have been used to draw conclusions regarding individual-level care. This cannot fully capture the complex interaction of patients with social determinant of their care within their community. Finally, metastatic cancer manifests in a spectrum of low to high burden disease. This inevitably affects the patient's utilization of health services. The current report cannot sufficiently characterize and adjust for different modes of presentations. Furthermore, the NCDB itself has certain drawbacks. Although diverse facility types and regions are represented in the NCDB, not all hospitals are CoC accredited. Therefore, depending on the demographics of nonaccredited centers, the conclusions may slightly differ.