By JEFF GOLDSMITH

In late June, 2024, two economists, Zarek Brot-Goldberg and Zack Cooper, from the University of Chicago and Yale respectively, released an economic analysis arguing that hospital mergers damage local economies and result in an increase in deaths by suicide and drug overdoses in the markets where mergers occur. Funded by Arnold Ventures  their study characterizes these mergers as “rent seeking activities” by hospitals seeking to use their economic power to extort financial gains from their communities without providing any value. 

The Brot-Goldberg-Cooper analysis was a spin-off of a larger study decrying the lack of federal anti-trust enforcement regarding hospital mergers. These two studies used the same economic model. The data were derived from the Healthcare Cost Institute, a repository of commercial insurance claims information from three of the four largest commercial health insurers, United Healthcare, Humana and Aetna (a subsidiary of struggling pharmacy giant CVS) plus Blue Cross/Blue Shield. HCCI’s contributors account for 28% of the commercial health insurance market.

The authors use a complex econometric model to manipulate a huge, multifactorial data base comprising hospital merger activity, employer health benefits data, county level employment data and morbidity and mortality statistics. This data model enabled a raft of regression analyses attempting to ferret out “associations” between the various domains of these data.

Using HCCI’s data, the authors construct what they termed a  “causal chain” leading from hospital mergers to community damage during their study period–2010 to 2015.  It looked like this: hospital mergers raise prices for private insurers-these prices are passed on to employers–who respond by laying off workers–some of whom end up killing themselves. So, according to the logic, hospital mergers kill people. Using the same methodology, the authors argued that between 2007 and 2014, hospital price increases of all sorts killed ten thousand people. 

A classic problem with correlational studies of this kind is their failure to clarify the direction of causality of data elements.  The model lacked a control group–comparable communities that did not experience hospital mergers during this period–because the authors argued that mergers were so pervasive they could not locate comparable communities that did not experience them.    

The model focused on a subset of 304 hospital mergers from 2010 to 2015, culled from a universe of 484 mergers nationally during the same period. The authors excluded mergers of hospitals that were further than fifty miles apart, as well as hospitals with low census. The effect of these assumptions was to exclude most rural hospitals and concentrate the mergers studied in metropolitan areas and cities. The densest cluster was in the I-95 corridor between Washington DC and Boston. See the map below:

According to the model, these mergers resulted in an average increase of 1.2% in hospital prices to commercial insurers, 91% of which were passed to their employer customers in those markets. This minuscule rate increase had a curiously focused and outsized effect–a $10,584 increase in the median employer’s health spending in the merged hospitals’ market.

According to the model, local employers “responded” to this cost increase by reducing their payrolls by a median amount of $17,900, all through layoffs–70% more than the alleged merger cost increase. This large overage was not explained by the authors. Moreover, the layoffs took place almost immediately, in the same year as the merger-induced increases, even though many health insurance contracts are multi-year affairs, and lock hospitals in to rates for that period.

At the end of the “causal chain,” 1 in 140 laid off people in those communities for whatever reason killed themselves through suicide or drug overdoses. By extrapolation, the authors accuse the perpetrators of overall hospital rate increases of killing ten thousand people in the affected communities during seven years overlapping the study period.