Background: Marked regional variation exists in end-of-life (EOL) care intensity and expenditures. Hospice use is expected to reduce EOL care expenditures, but literature provided mixed results on its financial impact. Determining the interplay of length of hospice enrollment on EOL care expenditures among regions with different levels of EOL care intensity could shed light on the relationship between hospice enrollment duration and cost saving. Methods: Using the Surveillance, Epidemiology, and End Results–Medicare database, we identified 90,465 decedents who had cancer diagnosed in 2004-2011 and died within 3 years of diagnosis as a result of cancer. We calculated decedents’ last 6 months expenditures in the hospital referral region (HRR) areas, and assigned HRRs into five quintiles. We also calculated each decedent’s length of hospice enrollment. Using hierarchical generalized linear models, we estimated risk-standardized EOL expenditures for each HRR and determined the associations between length of hospice enrollment and EOL expenditures across quintiles. Results: Compared with decedents in the lowest quintile expenditure areas, decedents in the highest quintile expenditure areas had shorter average hospice enrollment (8.3 days vs. 13.7 days; p < .001). Among decedents residing in the highest areas, increased hospice enrollment were associated with a decrease in the mean adjusted expenditure per decedent, from $59,167 for not hospice users, $52,456 for hospice users less than 3 days, $48,784 for hospice users 4-7 days, $45,393 for hospice users 8-56 days, to $41,713 for hospice users more than 56 days. However, there was no substantial difference in expenditures by length of hospice enrollment in HRRs with low levels of end-of-life expenditures ($37,731, $34,561, $32,759, $32,787, and $33,265, respectively). Length of hospice enrollment could explain 7.8% of variation in EOL expenditures between the highest and lowest quintile areas. Conclusions: Substantial geographic variation in length of hospice enrollment among cancer decedents exists. Early hospice enrollment has the potential to decrease expenditures in areas with high EOL expenditures, but not in areas with low EOL expenditures.