Hospice use and end‐of‐life care among older patients with esophageal cancer

Abstract Background Hospice and end‐of‐life health care utilization among patients with esophageal cancer are understudied. We used the Surveillance, Epidemiology, and End Results (SEER)‐Medicare linked database to analyze hospice use and end‐of‐life treatment patterns. Methods We included patients diagnosed with esophageal adenocarcinoma or squamous cell carcinoma between 2000 and 2011 and who had died by December 31, 2013. We evaluated patterns of hospice enrollment, chemotherapy receipt, radiation receipt, acute care hospitalizations, and intensive care unit (ICU) admissions at end of life. We used multivariate logistic regression to evaluate possible associations with hospice use, late ICU admission, and late chemotherapy receipt. Results Our study included 6449 patients; 3597 (55.8%) enrolled in hospice. Among hospice enrolled patients, 31.4% enrolled in the last 7 days of life. Hospice enrollment increased over time, from 43.2% in 2000 to 59.6% in 2013. Patients who were older, female, with stage IV disease, or those with higher socioeconomic status were more likely to enroll in hospice. Among all patients, 19.1% had an ICU admission within the last 30 days and 4.6% received chemotherapy within the last 14 days of life. Those who were Black or Asian (compared to White), married, or had a comorbidity score >1 were more likely to have a late ICU admission. Males and younger patients were more likely to receive chemotherapy at end of life. Conclusion Hospice enrollment rates among patients with esophageal cancer have increased over time; however, a significant percentage of patients enrolls near the end of life. Further research is needed to improve understanding of how end‐of‐life care decisions for these patients are made.


| INTRODUCTION
Esophageal cancer incidence in the United States has risen over the past 2 decades, with an estimated 16 940 new cases and 15 690 deaths expected in 2017. 1,2 Despite recent advances in treatment options, the overall survival outlook for these patients remains poor, with a 5-year survival of just 18.8%. 1 While those with localized disease experience better 5-year survival rates (42.9%), approximately 39% of patients are diagnosed with metastatic disease. 3 Thus, as incidence continues to climb, efforts to improve health care delivery and outcomes for these patients are critically needed.
Previous studies have shown that hospice enrollment among patients with cancer has been increasing, but a substantial percentage still do not receive hospice services, or receive services near the end of life. [4][5][6] Conversely, aggressive end-of-life cancer treatment has been steadily increasing, despite often providing limited benefit. 7,8 Notably, hospice care, sometimes considered under the purview of palliative care, is often defined as a service that provides comprehensive care for patients with terminal illness and their families, to provide services ranging from symptom management to bereavement. 9 Importantly, hospice services can help manage the difficult symptoms patients may experience and provide greater psychological support. [10][11][12] In addition, patients who have had end-of-life discussions regarding palliative and hospice care are more likely to use the services rather than undergo aggressive end-of-life treatment. 11,13 Therefore, despite improvements in hospice utilization and clear evidence supporting the benefits of hospice services, ongoing research is needed to identify ways of ensuring that patients receive appropriate referral for hospice services to alleviate suffering and improve end-oflife care.
Patients with esophageal cancer experience high rates of pain, dyspnea, 14 and dysphagia 15 ; however, the best management of these patients at the end of life remains unclear. Notably, little research exists to describe their patterns of hospice utilization. We sought to study patients with esophageal cancer to examine trends in hospice enrollment over time using the Surveillance, Epidemiology, and End Results (SEER)-Medicare linked database. In addition, we aimed to determine which patient and clinical characteristics were associated with hospice enrollment and aggressive end-of-life treatment, defined as acute care hospitalizations within the last 30

| Statistical analysis
We examined patient and clinical characteristics that may predict use of hospice and end-of-life treatment: age, sex, race/ethnicity (White, Black, Hispanic, Asian/Other), marital status, SEER region (Northeast, South, Midwest, West/Hawaii), urban location (big metropolitan, metropolitan/urban, less urban/rural), ecological socioeconomic (SES) status, AJCC stages (I, II, III, IV), and comorbidity score (0, 1, 2+). We imputed ecological SES status using ZIP code-level median household income from US census data provided in SEER-Medicare to derive quintiles. Charlson comorbidity scores were calculated using the Deyo adaptation of the Charlson comorbidity index for the 13-month period prior to cancer diagnosis. [18][19][20] Survival was defined as the time from diagnosis date to date of death. We used chi-square tests to compare the distribution of patient characteristics among patients with and without hospice enrollment.
We examined the hospice enrollment and end-of-life treatment prevalence by year and used Cochran-Armitage tests to analyze trends over time. Multivariable logistic regression models were used to identify associations between patient and clinical characteristics and hospice enrollment in the entire cohort and late hospice enrollment among hospice enrolled patients. We used multivariate logistic regression to analyze 2 indicators of aggressive treatment: chemotherapy receipt within the last 14 days of life and ICU admission within the last 30 days of life, for a total of 4 separate multivariate models. To correct for false positive determinations of significance that can occur when running multiple tests, we applied a Bonferroni correction for each of the 4 models. We used a P value of 0.0125 (0.05/4) to test the significance of the model; if significant, a P value <.05 was used to determine significance of the independent variables.
Statistical significance was defined as P value <.05 in a 2-sided test. Analyses were performed using SAS software, version 9.4 (SAS Institute, Inc., Cary, NC).

| Ethical considerations
This study has been approved as exempt by the Institutional Review Board at Massachusetts General Hospital. A Data Use Agreement was signed before obtaining the data from SEER-Medicare.

| Hospice enrollment
Over half of patients enrolled in hospice (3597; 55.8%). The percentage of patients who enrolled in hospice showed a steady increase over time from 43.2% in 2000 to 59.6% in 2013 (Cochran-Armitage test for trend, P < .0001) ( Figure 1). Among enrolled hospice patients, 31.4% did not enroll until 7 days before death, demonstrating a high rate of late enrollment. Notably, 6.1% of patients who enrolled in hospice lived more than 180 days on hospice. The median (IQR) age at death was similar for the 2 groups, with 77 (72-82) for those enrolled in hospice and 76 (82-72) for those who were never enrolled. Patients who were enrolled in hospice had a longer median survival than those who never enrolled (8.8 months (IQR 3.9-17.9) vs 6.9 months (2.9-16.4); Wilcoxon-Mann-Whitney P < .0001).
Our multivariable logistic regression model found that patients were more likely to have enrolled in hospice if they were 80 years or older, female, had stage IV disease at the time of diagnosis, died in later years of the study period, lived in either the South or Midwest, or had a SES quintile >2 ( Wald chi-square P = .006), and those with a Charlson score >1 were less likely to have enrolled than those with a score = 0 (OR: 0.82;    and also aligns with recommendations from guidelines and expert groups. [25][26][27] The overall rate of hospice utilization among patients with esophageal cancer was slightly lower than those found in overall Medicare cancer patients (59.5% in 2009). 6 It is also lower than rates seen in other cancer sites during a similar period, such as hepatocellular carcinoma (63.0%), 28 breast cancer (62.8%), 29 and glioma (63.0%) 30 but higher than others, such as leukemia (44.4%). 24 Collectively, our data support the need for additional research to understand barriers to hospice enrollment and to help encourage appropriate use of hospice services for patients with esophageal cancer.
Our analysis found that nearly one-half of patients with esophageal cancer were admitted to an acute care hospital within 30 days of death and 10% of our cohort had at least 2 hospital admissions.
These rates remained steady over time, demonstrating that aggressive end-of-life care continues to remain high in this population. One-fifth of patients had an ICU admission within 30 days of death, and nearly 5% received chemotherapy within 14 days of death. Female patients were more likely to enroll in hospice and less likely to receive aggressive end-of-life treatment, which may suggest differences in how patients approach treatment decisions. Black patients were more likely to receive aggressive treatment than White patients, and Asian patients were less likely to enroll in hospice. Notably, these sex 7,31 and racial/ethnic differences are consistent with other studies on disparities in end-of-life care 31,32 , and these findings further corroborate the need for efforts to address disparities in end-of-life care for patients with cancer. Interestingly, we found that marital status was not associated with hospice enrollment or late chemotherapy use, similar to an earlier study among patients with cancer. 32 Earlier studies have demonstrated that hospice enrollment differs based on geographic location. 33 Our study also suggests regional and socioeconomic differences in hospice enrollment and end-of-life treatment. Hospice enrollment was higher among patients within a higher ecological SES quintile, and among patients who lived in the South or Midwest (compared to Northeast). This contrasts with a similar study that showed no association between income and hospice enrollment in an adjusted model. 32 Patients living in the South or Midwest or living outside a large metropolitan area were less likely to have a late ICU admission. The mechanism for these findings requires additional study, but our results highlight that regional and socioeconomic differences play a role in patents' end-of-life care.
Patients diagnosed with stage IV cancer were more likely to enroll in hospice and less likely to have a late ICU admission. Clinically, this aligns with the fact that patients with esophageal cancer who are diagnosed with metastatic disease may have fewer treatment options than those diagnosed with earlier stages. It is possible that these patients receive information about hospice as a health care option earlier in their disease trajectory and at higher rates than earlier stage patients.
Alternatively, some earlier stage patients who ultimately experience disease progression may experience a more rapid decline, thus representing a barrier to early discussions about hospice enrollment.
Interestingly, we found that patients who enrolled in hospice had a longer median survival than those who never enrolled, and this is hypothesis-generating, but likely related to the fact that those who survived a longer period also had more time to be enrolled on hospice.
Our study has several limitations. It only includes patients older than 65 who resided in a SEER region and received Medicare services.
Therefore, our results may not be generalizable to other populations.
However, most patients with esophageal cancer are diagnosed above age 65 (60%), and SEER data represents one of the largest population databases available. 1 We identified patients who had died and analyzed health care utilization before death, which may be subject to biases. 34