Impact of Hospice on Spending and Utilization Among Patients With Lung Cancer in Medicare

Abstract

Objectives: To compare patterns and understand drivers of spending and utilization by month in the last 6 months of life between patients with lung cancer who used hospice versus those who did not.

Study design: Retrospective cohort analysis using 2009 to 2013 Medicare claims.

Methods: We used a 10% random sample of Medicare fee-for-service beneficiaries with lung cancer who died between 2010 and 2013 (43 789 beneficiaries). Patients were assigned to 2 groups depending on whether they used hospice care in the last 6 months of life. The following outcomes were constructed by month: (1) all-cause Medicare spending, (2) indicator of hospitalization, (3) indicator of emergency department (ED) visit, (4) number of part B chemotherapy claims, and (5) number of radiation therapy sessions. We used a combination of propensity score matching and regression analysis to compare outcomes between the 2 groups.

Results: Patients who used hospice had significantly lower spending in the last month of life compared to nonusers (US$16 907 vs US$26 906, P < .00). The spending differences in the last month of life between the 2 groups were largely driven by lower likelihood of hospitalization (54% vs 80%, P < .01) and ED visits (15% vs 22%, P < .01) and fewer chemotherapy (0.12 vs 0.32, P < .01) and radiation therapy sessions (0.80 vs 1.26, P < .01). During the other months in the 6-month period, spending was similar between the 2 groups.

Conclusions: Timely hospice enrollment of patients with lung cancer-the group suffering from high morbidity and mortality among patients with cancer-may lead to significant cost savings.

Keywords: Medicare; cost impact of hospice; end-of-life utilization trends; length of hospice use; lung cancer; monthly end-of-life costs.

Figure1.

Comparison of mean spending by month in the last 6 months of life between lung cancer patients that used hospice versus those that did not use hospice Notes: ^a Unadjusted spending by month for the unmatched sample. ^b Using the matched sample, spending was adjusted for patient and market characteristics for each monthly measure using GLM model with log link and gamma distribution. Standard errors accounted for clustering within a ZIP. Value are predicted values among hospice users and non-users after setting all the other covariates to their mean values; Hospice users; Hospice non-users;

Figure 2

Comparison of utilization of select services by month in the last 6 months of life between lung cancer patients that used hospice versus those that did not use hospice Notes: ^a Unadjusted utilization by month for the unmatched sample.; ^b Emergency department; ^c Using the matched sample, utilization was adjusted for patient and market characteristics for each monthly measure. Standard errors accounted for clustering within a ZIP. Value are predicted values among hospice users and non-users after setting all the other covariates to their mean values; Hospice users; Hospice non-users;

Figure 2

Comparison of utilization of select services by month in the last 6 months of life between lung cancer patients that used hospice versus those that did not use hospice Notes: ^a Unadjusted utilization by month for the unmatched sample.; ^b Emergency department; ^c Using the matched sample, utilization was adjusted for patient and market characteristics for each monthly measure. Standard errors accounted for clustering within a ZIP. Value are predicted values among hospice users and non-users after setting all the other covariates to their mean values; Hospice users; Hospice non-users;