Variation in State Per Capita Health Expenditures

CMS has a little know publication, the Medicare & Medicaid Research Review, which contains excellent work.  In a recent issue, staff at the CMS Office of the Actuary published research on their model of variation in state level per capita health spending.  (MMRR Article)  The model is based on data gathered from the period 1991-2009 from a variety of sources, including actual health spending, income levels, demographic, health status and other factors.  Much of the article discusses various methodological issues in building these models, as well as alternative approaches to address potential errors.  Some of the proxies are limited, for example, health status is based on smoking and obesity rates, and provider supply is based on beds in community hospitals.  And there is possible interaction between variables, for example, states with higher incomes likely attract more health care resources.  The strongest correlates (it is important to understand that correlation may or may not be explanatory in a causation sense) are personal income per capita and increased use of technology.  This finding is consistent with other research.  When people have more income, they are inclined to spend more on health care, whether in the strictest sense they need that health care or not. And it has long been argued that medical technology–new drugs, new medical devices, new diagnostics and new procedures like robotic surgery–are a fundamental cause of health spending growth, because they either are completely new and additive treatments or because they replace less-expensive treatments.  More community hospital beds were also associated with more spending as was having a higher percent of women of child-bearing age.  In most forms of the model, having a high percent of the population over age 65 and having more people who smoke or are overweight are also factors.  Percent of people uninsured and HMO penetration did not appear to have a major effect on per capita spending.  Somewhat surprisingly, more minority residents was negatively correlated with spending.  Because it looks over longer periods of time, the report provides a useful addition to the body of knowledge about geographic variation in health spending.