Skip to main content

Advanced Search

Advanced Search

Current Filters

Filter your query


Publication Types


Area of Focus
Date Range

Area of Focus

Controlling Health Care Costs

Changing Diagnosis Patterns Are Increasing Medicare Spending for Inpatient Hospital Services

An elderly patient admitted to the emergency room waits to be seen by a doctor.

Changes in diagnosis patterns are leading to an increase in Medicare spending for inpatient care. As the charts that follow illustrate, in examining some of the most frequent and fastest-growing episodes, we find high episode growth rates for DRGs (diagnosis-related groups) with higher reimbursement. This implies that overall Medicare spending increases are being driven more by rising volumes of these highly reimbursed services rather than by growth in payments per se. Incentives related to Medicare reimbursement may therefore be spurring changes in medical coding.

Key Findings

  • The frequency of sepsis and severe heart failure has increased, while coding for pneumonia and less severe heart failure has declined.
  • These shifts in diagnoses are not the result of demographic changes.
  • The increase in the frequency of episodes is higher for those with higher reimbursement, suggesting that payment incentives and coding may be contributing to spending increases.

Methodological Notes

  • Motivation: We sought to examine the rise in Medicare spending for inpatient services given the decline in overall inpatient utilization. Sepsis and heart failure were chosen because they were included in the diagnosis-related groups (DRGs) with the largest absolute increase in episode spending from 2008 to 2017.
  • Definitions: We defined an inpatient episode to include the following services and spending: total traditional Medicare spending for care in the 30 days leading up to an acute hospital admission, the duration of the hospitalization, and the 90 days post hospital discharge. To measure episode-related spending for hospitalizations that occurred during the calendar years 2008–2017, we analyzed 2007–2018 Medicare Part A, B, and D claims data and Minimum Data Set records for beneficiaries living in the 50 states and Washington, D.C. Beneficiaries needed to have traditional Medicare Part A and B for the entire length of the episode to be included in the study. Spending for beneficiaries who died during the episode window had spending measured until death. If the start date of a claim fell within the episode window, then that claim was attributed to an episode. All acute hospitalizations initiated a new episode, unless the hospitalization was a readmission within 90 days for the same DRG as the beneficiary's previous hospitalization. Nonacute hospitalizations in inpatient rehabilitation facilities, psychiatric facilities, and other long-term settings did not trigger new episodes. Since almost every hospitalization was considered the start of a new episode, beneficiaries could have overlapping episodes. To avoid double-counting spending for concurrent episodes, we weighted spending proportionately during the overlapping periods. For example, if a new episode began a month before a prior episode ended, then the spending for the overlapping month was weighted so that half of the spending was attributed to the earlier episode and half was attributed to the later episode. Up to three overlapping episode windows were accounted for in our approach. Fewer than 6 percent of index hospitalizations had four or more overlapping episodes. In this situation, all spending for overlapping time was attributed to the third episode.
  • In slides where it is applicable, the race of beneficiaries was determined using the RTI race variable, which uses an algorithm developed by the Research Triangle Institute to more accurately capture minority groups. The “Other” variable, in these slides, is a combination of RTI race variables for “Hispanic” and “Other.”
  • Limitations: Results may not be generalizable to other populations, including Medicare Advantage beneficiaries. While our results suggest “upcoding” may be occurring, further studies using clinical data and/or using a comparison group would be necessary to attribute the change definitively to coding behavior.

Publication Details



Melinda B. Buntin, Mike Curb Professor of Health Policy and Chair of the Department of Health Policy, Vanderbilt University School of Medicine


Melinda B. Buntin et al., Changing Diagnosis Patterns Are Increasing Medicare Spending for Inpatient Hospital Services (Commonwealth Fund, Apr. 2021).