BEA Health Spending Analyses

I have written before about the interesting work being done at the Bureau of Economic Analysis to understand the components of health spending and the factors driving trends in that spending.  Much of the Bureau’s focus has been on analyzing spending by disease or condition and decomposing that spending into price, utilization and intensity drivers, creating what they refer to as the Health Care Satellite Account.  The Bureau recently released data on a more detailed and expanded set of medical conditions and the researchers authored an article in Health Affairs describing the most recent results.   (HA Article)   The prior released data focused on 18 conditions, the new set uses 261 conditions.  The spending analysis is based on survey and medical claims data from 2000-2014.  The article focuses on 30 conditions, which accounted for 42% of real per capita growth in health spending during the study period, going from 13% of total spending in 2000 to 23% in 2014.  This rise was a result of both more disease prevalence for the conditions and increased treatment costs.  But there is a lot of variation by condition.  Hepatitis for example had spending growth because of rising treatment costs (due to the new Hep C drugs) but almost no increase in prevalence.  On the other extreme, anxiety disorders had rapidly growing prevalence but no increase in treatment costs.  Most conditions show a mix of the two factors.

A clear driver of treatment and spending over the study period is adoption of expensive drugs as a primary treatment for many conditions.  this is very noticeable for cancer, where costly medications not only add to treatment expenses, but keep people alive and being treated for longer.  Interestingly, greater use of preventive services also was a spending driver.  Partly this may be due to the spread of new vaccines which can be quite expensive; the ones for shingles for example.  Population aging and obesity may be driving some of the prevalence increase for common conditions such as arthritis, diabetes and hypertension.  A very telling observation by the researchers is that many of the new treatments may be effective, that is they improve patients’ health, but they may not be cost-saving.  As the authors note, there is room for further improvement in their analyses, both in the data sources and in the method of analysis.  For example, Medicare Advantage data is not included.  That program accounts for a third of Medicare beneficiaries.  Similarly, actual Medicaid claims data is not used.  The researchers should be able to get much of the missing data from the large managed Medicaid and Medicare plans.  And they note the difficulty in assigning spending to conditions when high-cost patients typically have multiple diseases.  How much of that doctor visit was for diabetes or for hypertension?  But the approach is very useful and helps us understand where the drivers of health spending growth are, so that policymakers and care managers can address those.