Skip to main content

Variation in Medicare Spending–The Search for Causes Continues

By June 1, 2011Commentary

The seemingly interminable controversy over the reasons for geographic variation in Medicare spending generates new research at a rapid pace.  The latest is an article published in the journal Health Services Research, exploring the relationship between medical spending and health.   (HSR Article) The Dartmouth Atlas work is the progenitor of this line of research and those authors suggest that there is substantial spending variation with no better quality in the areas with higher Medicare spending.  The current study looked at 17,438 beneficiaries in the Current Beneficiary Survey, which provides rich demographic and other data.  The primary outcome was defined as final health status, which was compared with medical spending in the prior three years.

Final health status was measured by being alive or not at the end of the observation period and by an index of functional status.  Medical spending is Medicare spending, from claims data, either alone or coupled with out-of-pocket spending, as reported in the survey.  Average total spending was about $27,000, with Medicare accounting for 60% of that.  The heart of the study is use of a statistical and experimental design method aimed at reducing the inherent bias in tracking health outcomes and spending that is caused by people in poor health having higher spending.  Sort of a chicken and egg problem.  A traditional analysis of the data in this study shows that in fact people with higher spending have poorer health outcomes.

Using the instrument variable design, however, leads to a finding that in fact greater medical spending leads to better health outcomes.  Total medical spending tends increase with education, age and family income, in the traditional analysis, but not in the instrument variable one.  The traditional analysis suggests that more spending is associated with a lower likelihood of survival, but the instrumental analysis shows the opposite effect.  If this research is accurate, it would lead to being more careful about wholesale reductions in Medicare payments to providers in high-spending areas.  Instead, policymakers might focus on spending variation for specific diseases or types of services.

Leave a comment