The Agenda

Guest Post: Arpit Gupta on Joseph Doyle on High-Cost Hospitals and Health Care Reform

Editor’s note: Arpit Gupta has kindly agreed to share this thoughts on new research by Joseph Doyle that has sobering implications for our efforts to restraining medical cost growth without undermining the quality of care.

If you’ve paid any attention to the health reform debate, you are likely familiar with statistics suggesting that America spends more than any other country on healthcare without a corresponding boost to longevity. This pattern holds up within the United States as well. As the Dartmouth Atlas team has argued, large regional differences in health spending don’t seem to bear any relation to differences in health outcomes.

Statistics such as these form an important basis for the health-reform push. If much of our current health spending is wasteful and doesn’t contribute to better medical outcomes, it should be easy to cut this waste and spend the proceeds on broader access to healthcare. The impetus for best practice programs and the IPAB cost-cutting board in the PPACA are based on assumptions that top-down bureaucratic efforts can painlessly achieve the sizable cost savings — 20-30% by some accounts — suggested by America’s seemingly deplorable health statistics.

The chief statistical problem with these sorts of large-scale comparisons is that higher medical costs can come from a variety of legitimate sources. In particular, higher medical costs may be the result of treating sicker patients. In terms of background patient characteristics, regions and countries differ dramatically for reasons relating to demography, patient sorting, social behaviors (smoking, diet, etc.) and a million other difficult to understand reasons. Is it a coincidence that many of the hospitals Dartmouth considers pricey — such as those in McAllen, Texas, the centerpiece of work by Atul Gawande in The New Yorker — are located in high-cost urban areas and serve patients with high rates of poverty and chronic illness? Though research on the regional variation of medical spending is interesting, it doesn’t add up to a convincing case that medical waste is prevalent and easy to purge.

Joseph Doyle, a health economist at MIT, has recently worked on a pair of fascinating papers that probe this issue much further, complicating the typical narrative.

His first paper examines the treatments of tourists in Florida. Visitors in Florida who fell ill and visited high-cost hospitals had lower mortality rates than those patients who went to low-cost hospitals. Doyle suggests that the lower mortality rate of high-cost hospitals might be explained through their more intensive care and greater overall use of medical procedures.

What’s fascinating is that the outcomes for these Florida visitors were vastly different from the outcomes of Florida natives, who didn’t seem to do better in high-cost hospitals relative to low-cost hospitals. 

One straightforward explanation for the difference between Florida natives and tourists is that patient bases around different hospitals are not identical. High-cost hospitals may serve a sicker population base than low-cost hospitals. The sicker patient base will make it appear as if the health spending is delivering little value. However, a much better test comes from comparing how hospitals handle patients who are coming from a more equivalent background — out-of-state tourists. For this sample, it seems that high-cost hospitals are actually spending more in ways that deliver value.

Doyle’s latest paper — with co-authors John Graves, Samuel Kleiner, and Obama Administration favorite Jonathan Gruber — comes to similar conclusions. This paper examine patients who go to one hospital instead of another due to ambulance delivery decisions. In one specification, the authors look at people living on either side of ambulance dispatch boundaries. Patients on one side of a dispatch boundary are very similar to patients on the other side; however they go to different hospitals when they dial 911. Comparing the outcomes of patients to went to high-cost hospitals, relative to nearly identical patients who went to low-cost hospitals, therefore provides a statistically rigorous measure of hospital impact. Any health outcome difference between those two groups of patients can be reliably assigned to hospital characteristics.

The conclusion once again is that higher-cost hospitals deliver better medical outcomes. A 10% increase in hospital costs is associated with a 4% decline in the mortality rate one year after hospital admission. Doyle’s statistical strategy also uncovers some interesting facts about what happens inside hospitals. Popular indicators of hospital quality — whether a hospital is using “appropriate care” for heart attacks for instance — don’t relate to actual health outcomes. Rather, teaching hospitals and early adopters of new technology seem to deliver better outcomes. The relationship between hospital cost and treatment is primarily driven by quantity rather than price — more treatments result in better care, but higher medical prices don’t. Finally, medical costs seem to exhibit diminishing returns.

From a statistical point of view, Doye’s work points to the importance of what economists term endogeneity. Different groups and populations can vary in a complicated fashion, making similar statistical inference — for instance with cross-country graphs or statistics — difficult. As an example, high-crime areas tend to have more cops than low-crime areas. A simple analysis might therefore conclude that having more police doesn’t reduce crime. The problem here is simple to grasp — cops tend to focus on high crime areas, and reduce crime relative to having no cops. Yet similar inference problems tend to crop up in many areas of social science, often in non-obvious ways. Though the public tends to look to economists to predict financial crises and opine on debt crises; economists are typically not very good at these tasks. Instead, the value of economics, such as it is, really rests on being able to cut through difficult questions of statistical inference — as Joe Doyle has done rather well here.

Doyle’s results also undercut the claims of Dartmouth Atlas researchers and point to different conclusions for healthcare reform. To be sure, Jonathan Gruber is a co-author on this research, and likely believes that the major thrust of the Obama Administration’s approach to health reform is consistent with this research. Also, Doyle’s work so far focuses on emergency care and might not be relevant to other aspects of health spending. However, a plausible reading of his work is that common assumptions regarding the prevalence of large amounts of wasteful spending won’t hold up. Spending more money on newer technology or additional treatments may well save lives; and cutting medical spending may come at a steep cost. This doesn’t mean that cost-containment on healthcare will be impossible, but it highlights the painful dilemmas a top-down bureaucratic board will face in trying to identify legitimate treatments and deny care. Additionally, this work highlights the importance of technological improvements in improving medical care. It’s essential that health reform preserve incentives for the development and dissemination of new medical technologies; a point of view that’s sometimes lost when thinking about medicine as purely an issue of distributing the current supply of medical resources.  

An alternate strategy for health reform might start with the approach John Cochrane recently laid out in the Wall Street Journal. Deregulation of hospitals and insurance markets might create legitimate markets, while both lowering costs and speeding the adoption of new live-saving technology through competitive pressures. A market-led approach to health reform is hardly without risks of its own. But whatever strategy we take, it’s certainly worth better understanding the real determinants of cost and mortality in the American medical system.


The Latest