On 11 September, 2014 Louise Sheiner presented a paper at Brookings that challenges some of the the interpretations of Dartmouth research on geographic variation in health care. Her work suggests that patient, not provider, factors explain most of geographic (in her case, state) variation in spending. Coincidentally, Austin Frakt had already prepared a post reviewing work that comes to the opposite conclusion. That post appears below, and is not intended as a rebuttal to Sheiner’s work.
One of the most important and vexing strands of research on the American health care system began with a 1973 study of tonsillectomy in Vermont. That year, John Wennberg and Alan Gittlesohn published “Small Area Variations in Health Care Delivery,” in which they showed that in one Vermont town, 66% of children had their tonsils removed, but in neighboring towns 16-22% did.
As Aaron wrote at The New York Times,
There were no differences in the children [...]. Tonsillitis was not more common in one town, and the children were no more or less healthy before or after the procedure. The doctors just seemed to have very different attitudes toward how to take care of tonsils.
Since the Wennberg/Gittlesohn study, many investigators have examined geographic variation in health care utilization and spending and their relationships to health outcomes. A great deal of this literature is summarized in a Handbook of Health Economics chapter by Amitabh Chandra, David Cutler, and Zirui Song, addressing they key question, “What drives who gets what care?” These scholars categorize the possible factors as either demand side (e.g., income and preferences for care), supply side (e.g., quality), or situational (e.g., idiosyncratic contextual or behavioral influences).
There is no debate that large regional differences in health care provision, costs, and outcomes exist. For example, as Atul Gawande famously documented, in 2006 Medicare spent $15,000 per beneficiary in McAllen County, Texas and half that in the very similar county of El Paso, Texas. But there is considerable debate as to what drives such differences, in general, as Chandra et al. explain.
Some researchers argue that variation is accounted for by population disease burden (Zuckerman et al., 2010), but other authors argue that prevalence of diagnoses itself is endogenous across areas (Song et al., 2010; Welch et al., 2011). Most of the literature agrees that patient characteristics and preferences do not explain much of the differences across areas, and that substantial variations in treatment practices remain after controlling for patient characteristics (Anthony et al., 2009; O’Hare et al., 2010; Baicker et al., 2004).
The authors are also brutally honest, writing, “None of the theories for which there is a lot of evidence can be shown to explain a major part of cross-individual or cross-area variation in treatments.” However, their tentative conclusion is that supply side factors are more important than demand side factors in driving treatment choice.
An August 2013 paper by David Cutler, Jonathan Skinner, Ariel Dora Stern, David Wennberg is consistent with this conclusion. The authors examined the extent to which patient and physician preferences (collected by surveys) explained regional variation in health care spending.
Ultimately, the largest degree of regional variation appears to be due to differences in physician beliefs about the efficacy of particular therapies. Physicians in our data have starkly different views about how to treat the same patients, and these views are not highly correlated with demographics, background, and practice characteristics, and are often not consistent with professional guidelines for appropriate care. As much as 36 percent of end-of-life Medicare expenditures, and 17 percent of overall Medicare expenditures, are explained by physician beliefs that cannot be justified either by patient preferences or by clinical effectiveness.
In other words, many physicians believe in the value of certain types of care even when the evidence does not support that belief, and this explains a substantial fraction of spending.
Theirs is not the first work to highlight the importance of supply side (physician) factors. In Gawande’s examination of McAllen and El Paso, he found that McAllen physicians seemed to be more entrepreneurial than their counterparts in El Paso, for instance owning “imaging centers, surgery centers, or another part of the hospital they directed patients to.” Francis Lucas and colleagues found significant peer effects: “27% of respondents reported ordering a cardiac catheterization if a colleague would in the same situation frequently or sometimes.”
Clinicians’ beliefs may play an important role in geographic variation in health care and the inefficiencies such variation suggests. (This is not to the exclusion of patient factors, of course.) As such, an interesting line of research would focus on the extent to which those beliefs are malleable. Overpowering them with payment incentives could work, if those incentives are large enough, but I wonder if beliefs can be addressed head on.
Austin B. Frakt, PhD, is a health economist with the Department of Veterans Affairs and an associate professor at Boston University’s School of Medicine and School of Public Health. He blogs about health economics and policy at The Incidental Economist and tweets at @afrakt. The views expressed in this post are that of the author and do not necessarily reflect the position of the Department of Veterans Affairs or Boston University.