Proportion of Decedents’ Expenditures Among Recent Reductions in Medicare Expenditures

Abstract

This cohort study uses Medicare data from the Dartmouth Atlas Project to determine the source of recent changes in end-of-life Medicare expenditures.

Health care expenditures continue to increase in the United States. Historically, end-of-life expenditures have been a consistent contributor to that growth.¹ Recent studies of end-of-life practice patterns suggest that increased use of hospice care costs more than it saves,² end-of-life care intensity is increasing,³ and end-of-life intensive care unit use and care transitions are increasing,⁴ raising concerns that end-of-life care costs might be accelerating.

To investigate such concerns, we used publicly available Medicare data to calculate changes in end-of-life Medicare expenditures in the recent past.

Methods

Dartmouth College has determined that analysis of Dartmouth Atlas data is not considered human subjects research. From the Dartmouth Atlas Project website (http://www.dartmouthatlas.org), for fee-for-service Medicare enrollees aged 65 to 99 years, we obtained publicly available per-capita mean annual 2004 through 2014 Medicare part A and B expenditures for all enrollees and mean prior 2-year expenditures for enrollees who had a chronic condition and died in that year (decedents). From total fee-for-service expenditures (number of enrollees times the mean annual per-capita expenditure), we subtracted total annual expenditures for decedents (the number of decedents times the mean prior 2 years’ expenditures) to estimate survivors’ expenditures.

Using the consumer price index to convert expenditures to 2014 dollars, for decedents and survivors, we calculated the annual overall per-capita expenditures in 6 categories: inpatient (including hospital and skilled nursing facility expenditures), physician, outpatient, home health, hospice, and durable medical equipment. We calculated year-to-year changes in Medicare expenditures attributable to changes in the numbers of survivors or decedents (changes in numbers times the mean annual per-capita expenditure) and changes in practice intensity (total expenditure changes minus changes attributable to enrollee numbers).

Our analysis encumbered the year of death with the prior 2 years’ costs. Informed by patterns of Medicare utilization in the last year of life,⁵ we conducted 2 sensitivity analyses that allocated end-of-life expenditures across prior years; those did not generate appreciably different findings.

Results

The number of Medicare A and B enrollees aged 65 to 99 years decreased from approximately 27 million in 2004 to approximately 25 million in 2009 before increasing to approximately 26 million in 2014; during that time, the annual proportion of decedents decreased from approximately 4.6% to approximately 4.2%. Overall per-capita Medicare expenditures increased between 2004 and 2009 (from $9119 to $10 458 in 2014 dollars) and then decreased (to $9589 in 2014), with the proportion of per-capita costs attributable to decedents decreasing steadily. For both survivors and decedents, changes in overall expenditures due to changes in practice intensity increased until roughly 2009 and have since decreased; between 2010 and 2014, despite accounting for 4.3% of enrollees, decedents accounted for 54.6% of Medicare expenditure reductions attributable to practice changes.

For all enrollees, decreasing inpatient, physician, and durable medical equipment expenditures drove slight per-capita expenditure decreases (Figure, A). Since 2009, per-capita expenditures fluctuated slightly for survivors but decreased considerably for decedents, for whom inpatient, physician, and home health care expenditures decreased (Figure, B and C).

Figure. For 2004 Through 2014, Annual Overall per-Capita Medicare Expenditures in 6 Categories for All Fee-for-Service Medicare Enrollees Aged 65 to 99 Years, for Those Who Survived, and for Those Who Died.

Discussion

Contrary to other studies, we found that per-capita end-of-life spending is decreasing and appreciably contributing to overall per-capita Medicare spending growth moderation. End-of-life spending reductions were driven by substantially lower inpatient care expenditures (and modestly lower physician and home health expenditures) without commensurate increases in other care segments. Potential explanations include both changes in supply (physician efforts may have been redistributed to the newly insured following implementation of the Affordable Care Act) and demand (normalization of more conservative end-of-life preferences, or even financial constraints imposed by the Great Recession⁶).

While our findings are limited because they used administrative data sets, examined only the fee-for-service Medicare population, could not examine components of inpatient care, and did not study specific patients and their end-of-life spending patterns,⁵ they warrant further investigation into why end-of-life care intensity is changing and, particularly, how those changes influence end-of-life care quality.