Predictors of Hospice Utilization Among Patients with Lymphoma

Background: Although there is an increasing focus on the provision of quality end of life (EOL) care for cancer patients, little is known about the factors that influence hospice use for patients with lymphoma. Utilization of palliative services is also felt to be lower in hematologic oncology. We aimed to characterize the prevalence and predictors of hospice enrollment in a large population of patients with B-cell non-Hodgkin lymphoma (NHL).

Methods: We conducted a retrospective analysis using the Surveillance Epidemiology and End Results (SEER)-Medicare database. Patients ≥ 65 years of age who were diagnosed with any B-cell NHL, lived for at least 30 days, and died between January 1, 2000 and December 31, 2010 were eligible. We divided the cohort into three groups: those with indolent NHL (chronic lymphocytic leukemia and follicular, marginal zone, and lymphoplasmacytic lymphoma), aggressiveNHL (diffuse large B-cell, Burkitt, primary mediastinal B-cell, and primary effusion lymphoma), and mantle cell lymphoma (MCL). MCL was considered a unique group because, while it is largely incurable like indolent NHL, it often has a more aggressive clinical course. We first characterized the prevalence and covariates of hospice use in the cohort. Next, we fit a multivariable logistic regression model to characterize factors associated with hospice use, including NHL type (indolent, aggressive, MCL) and patient sociodemographic characteristics with p < 0.05 in univariable analyses.

Results: A total of 18, 815 patients were eligible, of which 9,666 had indolent NHL, 8,241 had aggressive NHL, and 908 had MCL. Of the total cohort, 7,820 (42%) enrolled in hospice, with 6,450 (34% of the total cohort) enrolled ≥ 3 days before death. There were significant differences in hospice use by NHL type, with 39% of indolent NHL patients, 44% of aggressive NHL patients, and 48% of MCL patients enrolling in hospice respectively (p < 0.001). Multivariable associations of patient characteristics with hospice use are shown below.

Figure 1

*Multivariable model adjusted for all covariates in table and time from diagnosis to death.

View Large

The median time from diagnosis to death was significantly longer for patients who enrolled in hospice (37.5 months) compared with those who did not enroll (31.8 months, p < 0.001); effects were similar in analyses stratified by lymphoma type.

Conclusions: In this large cohort of patients with B-cell NHL, the rate of hospice use ≥ 3 days before death (34%) was substantially lower than the 2007 national average of 43% for all cancer decedents (National Hospice and Palliative Care Organization Facts and Figures, 2012). This finding is provocative: it suggests either the need for improvements in enrollment and/or that our current hospice model is not meeting the specific needs of lymphoma patients in the United States. We also found disparities in hospice use similar to those previously observed in solid malignancies, with non-white patients and those of lower socioeconomic status less likely to enroll. Moreover, patients who enrolled were more likely to have lived longer after their diagnosis, perhaps due to increased experience with—and thus desire to avoid—the burden of additional treatments. Finally, the fact that patients with MCL had the highest rate of enrollment may reflect its increased symptom burden, and often aggressive course in the context of being incurable. These may be clearer indicators of the EOL phase and the need for hospice.