April 15, 2013 Morayea Pindziak

Return of the American Pastime – Analyzing the latest health expenditure data

Last month, AHRQ released the latest data from the Medical Expenditure Panel Survey (MEPS), which is one of the few nationally-representative, longitudinal databases with patient covariates and medical costs available to the public.  Our excitement of the recent full-year 2011 data was dimmed somewhat when realized that the current release was only preliminary without the expenditure data. Though we can still analyze utilization rates and diseases up to three digits of ICD9, we are anxiously awaiting the complete data.  In the meantime, we would like to present analysis from the latest full dataset, MEPS Panel 14 (2009-2010).

One “fact” that often gets thrown around in everything from the media to bar conversations is that a small minority of Americans account for the majority of medical expenses; coupled with the fact that the United States spends more per capita on health care, this minority is often vilified.  For this analysis, we used the MEPS dataset to analysis 1) If indeed a small population does most of the health care spending and 2) Whether we can identify that population.

Big-market vs. Small-market: How skewed is medical expenditure in the United States?

For the nationally-representative sample, we divided each observation into quintiles based on self-reported expenditure in 2009 and 2010. Then, we calculated the spending of each quintile as a percentage as total spending.   We present the 2010 results in Table 1 only because 2009 followed a very similar pattern.






Total medical spending was over 1.2 trillion USD in 2010 and the top 20% accounted for over 82% (1 trillion USD) of that total.  As quintiles, each of these groups represent roughly 60 million Americans.  For the bottom 20%, total expenditure totaled less than 80 million USD, which is roughly $1.30 per person for 2010.  The second column in Table 1 shows that people in the bottom quintile spend between $0 and $34 for the entire year versus those in the highest quintile who spent over $4000 each year.  The upshot of this analysis confirms the factoids that we often hear: 20% of Americans account for over 80% of all health spending!

Hot streaks: Is high-spending persistent?

For every individual, there are years where spending is high and others where spending is low. For the top 20% in Table 1, was it just an aberration from their normal spending habits? One of the main advantages of the MEPS dataset is that it allows us to have data for individuals in subsequent years. In this section, we analyze what happened in 2010 for those who were in the top quintile in 2009.

The results for the analysis in Table 2 show over 55% (33 million) of the 61 million in the highest quintile remain in the highest quintile the following year. In fact, nearly 80% of the top spenders end up being in the top two quintiles the subsequent year.   Less than 5% move to the bottom quintile, which includes those who have zero medical expenditure in the year.  From this analysis, we conclude that high spenders consistently outspend other Americans each year.

Spending like the Yankees: who are the people in the top quintile?

Another “fact” we often hear is that the majority of the spending is done by the sick and elderly. In this section, we analyze the distribution of seniors (those aged 65 and above) and those that died during the MEPS survey.

In 2009 senior citizens represented roughly 13% of the total population; of those 40 million seniors, nearly half of them fell into the top quintile category for medical spending by Americans.  In fact nearly 80% of seniors spent more than $1200 annually.  Note that this analysis only provides associations and cannot make any statements on causality: unfortunately, we have yet to develop technologies to make people younger, so we cannot test what would happen to spending if these people became non-seniors.

Dead-ball Era: Is death costly?

In addition to calling out seniors as high spenders, the other group that receives attention for spending are those people on the last months of their lives. For this, we use the MEPS dataset to identify those that died during the survey and see how their expenditures were distributed across quintiles.

This analysis suggests that the majority of those who died during the year spent more than $4000, which put them in the top quintile. One thing to note is that the total number of deaths (nearly 4 million) is higher than commonly cited numbers.[1]  We attribute this bias to the small number of observed deaths in the survey (182), and use this to caution researchers when applying statistical weights to small numbers of observations.

From the analysis in this section, we find evidence to support the assertions than seniors and the sick spend more on medical care than others.

Postseason: Concluding Thoughts

In this post, we attempted to showcase the unique data available in MEPS.  For those in the health outcomes fields, this dataset is powerful in that it also has diagnostic codes, so it provides an easy source to see the costs associated with various diseases and conditions.  The subjects from MEPS are a subset of the National Health Interview Survey (NHIS), so researchers can actually utilize up to four years of time-series data using the publically-available linkages between the two datasets.

[1] The CDC estimates roughly 2.5 million deaths (http://www.cdc.gov/nchs/deaths.htm).