When more expensive medicine is better medicine
The theory behind the health reform law is that you can deliver better care at lower cost. Support for that idea comes from a robust body of research — most prominently the Dartmouth Atlas project — which suggests that a decent amount of health care spending is unnecessary. Hospitals can get the same (if not better) outcomes, the thinking goes, if they focused on providing just the most cost-effective treatments.
But sometimes, more expensive care is also better care. That’s what a team of health-care economists have found in a new NBER paper, which looked at Medicare patients in New York. It found that, all other things being equal, those treated in higher-spending hospitals had mortality rates 20 to 30 percent lower than those treated in low-cost facilities.
Those findings, in some ways, contradict the Dartmouth Atlas research, which has suggested that more can be less.
“We find that more intensive intervention is associated with better outcomes when compared with less intensive interventions,” said Joseph Doyle, a professor at the Massachusetts Institute of Technology Sloan School of Management and lead author of the study. “We would say that this is a cautionary tale, that we cannot necessarily cut spending and expect quality to increase.”
Why do the higher-spending hospitals deliver better results? One plausible explanation could be that they’re higher-tech, having invested more aggressively in newer, more accurate medical devices. Higher-spending hospitals could also be those that train doctors and, again, tend to invest more heavily in human capital.
The NBER researchers did find that teaching hospitals and “high-tech” hospitals were among the bigger spenders with better results. Being treated in a teaching hospital, for example, lowers the mortality risks by about 4 percent.
But they also tried controlling for those two factors. When they did, they found that the disparities persisted: Those treated in the higher-spending hospitals tended to have better outcomes than those where treatments cost less. And that leads them to conclude that in the higher-spending hospitals, the better outcomes most likely result from providing more treatments.
“The Dartmouth research finds no relationship between spending and outcomes, and concludes that we can cut spending without doing harm,” Doyle said. “We don’t know that’s quite appropriate to say, given that we find high-cost hospitals doing better jobs.”
As to why the NBER researchers found different results than the Dartmouth Atlas, Doyle said it may partially have to do with the research methods. His work compared different individuals living in New York City, who were “randomly assigned” to different hospital emergency rooms by wherever an ambulance took them.
The Dartmouth researchers have, in contrast, compared spending outcomes in different cities; they have famously worked with the New Yorker’s Atul Gawande to compare the high spending in McAllen, Tex., to nearby El Paso. Doyle, in his work, thinks that research method could be masking important contributors to spending, such as variations in how healthy each population is.
“A major issue that arises when comparing regions is that greater treatment levels may be chosen for populations in worse health,” Doyle writes in the paper. “For example, higher spending is strongly associated with higher mortality rates on the individual level. ... At the regional level, long-term investments in capital and labor may reflect the underlying health of the population as well.”
To be sure, a number of health-care systems across the country have successfully demonstrated that better care can be delivered at lower costs, largely by coordinating between various providers. These are places such as Kaiser Permanente and the Mayo Clinic, which made big investments in building new care-coordination models; they had well-laid game plans for how, exactly, they would improve the quality of care when providing less of it.
In those cases, hospitals have shown that they can generate savings. Without that coordination component, however, this new study could suggest that those gains may not materialize.