Large Variation in Hospital Billing: Three Preliminary Takeaways from New U.S. Data

The NYTimes has an interesting article on variation in hospital billing. In addition to highlighting substantial dispersion for the same procedure even within local areas (e.g. “a hospital in Livingston, N.J., charged $70,712 on average to implant a pacemaker, while a hospital in nearby Rahway, N.J., charged $101,945.”), it also cited the release of awesome Medicare price and billing data from CMS.

New Data: The data provide the average amount each hospital in the US charged Medicare for every DRG (Diagnosis-related group) and how much Medicare actually paid in 2011. I merged this hospital-DRG level data with zipcode level tax return data from the IRS to link these hospital prices and payments to local incomes (the IRS data are from 2008. Presumably average zipcode incomes are highly persistent).

Here are three preliminary takeaways from an initial look at this data.

1. There’s substantial variation in the amount hospitals charge for a given procedure in the US. The first figure shows that the amount hospitals charge for a DRG varies widely from 50% to 3X relative to the average amount all hospitals in the US charge for that DRG. While there are local price, quality, demographic, etc. differences, this is pretty substantial variation.
2. Medicare typically pays about 30% of the amount hospitals charge, but there’s also wide variation in this amount. Interestingly, hospitals in zipcodes with higher average incomes tend to charge more, but Medicare pays hospitals in these areas an even lower share of the charged amount.
3. The amount hospitals charge for a given procedure (relative to the mean US charge for that procedure) increases with local income. Figure 3 shows the average ratio between hospital charges for a given DRG to the nation mean for 50 AGI bins. There are likely quality differences and other important omitted variables that affect outcomes, but the simple correlation appears to be quite strong. Roughly speaking, going from a zipcode that has mean AGI of $40-50K to one with average AGI of $80-$100K is associated with an ~ 50% increase (from .8 to 1.2 or 1.3) in the amount Hospitals charge. It may be the case that part of the story is that hospitals in richer areas have to offset higher labor and rental costs so they charge higher prices (based on mechanisms related to Bamoul’s cost disease). In addition, higher relative prices of healthcare and higher incomes both affect the demand for healthcare. I am investigating these explanations more in depth/more formally and will hopefully be able post about this more soon.