“The Price Ain’t Right” Ain’t Right Again!

When Gaynor et al. originally released their “The Price Ain’t Right” report on the relationship between hospital prices and market structure in 2015, there were a number of questions and concerns about how they reached their conclusions. Now, with a revised version of that study just released, many of those concerns remain, as well as additional ones about new claims made in the report.[i]

The result is that the report’s findings are based on old and limited data, none of which include the payor with the biggest share in most markets, and with highly uneven geographic representation. The authors make a number of leaps in their assumptions to compensate for real and significant gaps in their methodology and data. These issues raise serious questions about the results of their analysis.

The authors assert that market structure allows certain hospitals to demand contracts that load more risk on insurers such as payments based on discount off charges. This assumes that hospitals prefer discount off charge arrangements and they will leverage their market power to negotiate these types of contracts. It also assumes hospital risk aversion. However, if hospitals feel they can manage risk effectively, they may actually prefer contracts that allow them to benefit from reducing costs. The authors don’t present any evidence that hospitals are risk averse or that the supposedly preferred low-risk contracts are associated with more profitable hospitals. In fact, Figure 9 in the paper appears to show lower prices for contracts based on a share of charges.

The authors also attempted to show that mergers resulted in higher prices when merging hospitals were located near each other. However, the geographic areas used to assess hospital market structure don’t accurately measure competition. In drawing conclusions about the relationship between market structure and price, the paper primarily uses a crude 5-25 mile radius to define hospital markets. This ignores the fact that hospitals don’t draw symmetrically from all directions and that travel distances and times can vary widely for both urban and rural hospitals.

The authors don’t provide much detail on their methodology for analyzing the effects of hospital mergers. They discuss using control groups and a “difference-in-difference” approach, but don’t say much about how they implement their methodology. For instance, they note that they looked at two years before and two years after a merger. However, given that they are using data covering a 2008 to 2011 time period, it appears that they don’t have any pre data for their 2008 mergers or any post data for their 2011 hospitals. This is either a weakness in their methodology or with how they characterized it.

The authors relied once again on old claims data – covering 2008 to 2011 – from the Health Care Cost Institute (HCCI). The HCCI data is comprised of employer-sponsored health insurance (ESI) claims for three large payors, Aetna, Humana and United, and represent 27.6 percent of individuals with ESI coverage. The Kaiser Family Foundation estimates that 49 percent of Americans have ESI[ii], so the HCCI sample represents just 13.5 percent of covered lives. The HCCI data also do not include any data from Blue Cross Blue Shield (BCBS) plans. This is an astonishing gap given that BCBS plans are often the largest payor in a market area. They attempt to compensate for the lack of BCBS data by identifying markets with high and low BCBS share. However, the map of BCBS market share they present on Figure 13 in the Appendix appears to seriously undercount the BCBS share and market power. For instance, they put BCBS of Texas (HCSC) at under 30 percent statewide market share, but according to the Kaiser, HCSC holds 44 percent of the Texas individual market, 51 percent of the small group market, and 46 percent of the large group market.[iii]

Finally, buried in the introduction of the revised study, the authors include a caveat that their paper is “fundamentally descriptive.” They go on to say that unobserved factors mean any links that they found between market structure, prices and contract structure “should not be assumed to be causal.” That’s a striking statement for a report that draws broad and questionable conclusions about the links between those very factors, especially when there are clearly other factors at play here.

The AHA is commissioning a more in depth analysis of the flaws in the methods used and the conclusions reached in this report that will be available in the coming weeks – so please stay tuned.

By Melinda Hatton, AHA General Counsel, and Aaron Wesolowski, AHA Vice President, Policy Research, Analytics and Strategy