How Does Health Insurance Affect Demand for Health Services?

Health economists have fun, I am sure, although I admit that reading many of their papers may lead to the opposite conclusion.  Believe it or not, they are creative types, and I enjoy learning about new methods and approaches to answering long-standing health questions.  One such question is what impact obtaining health insurance may have on the demand for health care.  We would expect that it will lead to more use of services and most past research, including a seminal Rand study, shows that to be the case.  Research from analysts at the Bureau of Economic Analysis takes a new tack in answering the question.   (BEA Paper)   A fundamental issue in such analyses is to determine the real out-of-pocket price faced by a consumer.  For someone with no insurance, that is pretty straightforward.  For a government employee or a Medicaid enrollee  it is also generally clear–you don’t pay anything.  For the rest of us, there are a series of copays and deductibles and coinsurance that vary by service and sometimes even by service location.  Consumers may have trouble figuring out what they will owe in advance, but if they can, that cost should affect demand in some way.  The data used in the study is from a commercial health plan claims database for the years 2006 and 2007.  (Note that cost-sharing has risen dramatically since this time.)  It was grouped into episodes of care.

The analysts use an instrument variable method, with the instrument being an index of prices negotiated between providers and plans in a particular geographic market.  This method is complex, but is designed to isolate cause and effect.  103 geographic regions end up being included, and price and utilization was examined across 271 episodes.  As an interesting aside, just 5 episodes–pregnancy, back joint issues, hypertension, diabetes and heart disease–account for 25% of all spending.  Another interesting statistic–16% of enrollees had no health spending; insurance is a really good deal for them!!  The researchers find an elasticity of demand similar to that in the Rand experiment–increasing the cost to a patient results in about 20% less service use.  And they confirm that the reduction is largely in episodes of treatment, not in utilization within an episode.

It is likely the perception of the impact of the cost of a health service on a specific individual that is key, not the actual amount of cost-sharing.  Don’t think Bill Gates cares much about his cost-share.  And using area-wide income measures suggests that for wealthy people there is much less of an effect.  There is enough fudging in the method here that it is not clear how precisely the actual cost share for a specific service for a particular individual is being calculated, and we can nitpick about things that might get missed, like generic drugs with a cost under the copay, for which pharmacies often don’t sent in claims.

But the analysis seems sound and consistent with common sense.  The most important implication is in regard to why national health spending growth seems so muted in the aftermath of the recession.  The answer is pretty clearly the enormous growth in consumer cost-sharing–for almost all commercial insureds, any health service now involves a deductible, coinsurance and/or a copay.  That cost-sharing is doing its job in restraining spending, but who knows if the missed care was needed.