New Results: Hospitals that Spend More on Patients Have Lower Mortality Rates

In the last two years, the results in several published papers suggest that higher, and more costly, hospital treatment intensities produce lower mortality rates. These results are in stark opposition to claims that because there is no observable relationship between variations in end-of-life hospital spending and mortality, a large fraction of high intensity care is wasteful. They also cast doubt on related claims that 20 to 30 percent of U.S. medical spending does no good.

In a refreshing departure, Kaestner and Silber (Milbank Quarterly, December 2010) begin by noting that it is curious that such waste would persist despite the existence of numerous market participants, including large for-profit corporations that self-insure medical expenditures, large health insurers in competitive markets, and large managed care organizations, with both the “expertise and financial incentive” to eliminate wasteful spending. They are therefore surprised that the available evidence (much of it from the Dartmouth group) suggests that huge waste still exists. They are also surprised that so “many people believe that the cause of the problem is easy to identify” and that it can be cured simply by “using capitated and bundled payments that encourage integrated systems of care and restrain the use of unproven treatments.”

Using several techniques to address the reverse causality problem that likely biases estimates of the relationship between treatment intensity and mortality, they find that a 10 percent increase in inpatient spending is associated with a 3.1 to 11.3 percent increase in 30 day survival. They conclude that more credible assessments of the causal relationship between spending and health are needed, and that believing that inefficiency is rife is inconsistent both “with economic incentives and a growing body of empirical evidence.”

Doyle et al. (NBER Working Paper No. 17936, March 2012) used differences in ambulance assignments to compare outcomes for emergency patients at high and low cost hospitals in New York State. They found that conventional estimates likely underestimate the returns to treatment intensity because they cannot account for patient selection. When patient selection is accounted for, higher-cost hospitals are associated with improved patient mortality outcomes and the returns are substantial. High cost hospitals reduce mortality by 2 to 2.3 percent relative to the average. A two-standard deviation increase in hospital costs is associated with a 15 to 30 percent decline relative to the average.

Teaching and high tech hospitals both tend to have higher average costs. Being taken to a teaching hospital reduces average mortality by 3.9 percent, over 10 percent of average mortality. Being taken to a high-tech hospital reduces average mortality by 4.7 percent. Hospital ratings using the best practices measures for heart failure, heart attack, and pneumonia endorsed by CMS had no statistically significant effect on mortality.

Finally, Bernato et al. (Medical Care, February 2010) analyzed the Pennsylvania Health Care Cost Containment Council discharge data set from April 2001 to March 2005 to estimate a Cox survival model. They also found that admission to hospitals that used more intensive care and life-sustaining resources was associated with a moderately higher likelihood of survival.

These papers join a small literature of earlier papers with similar findings. These include Silber et al. who found that more aggressive treatment in Medicare surgical patients was associated with lower mortality. Cooper (2009, Health Affairs) used broad measures of system quality to show that states with higher per capita total health spending had better quality care. Higher per capita Medicare beneficiary spending apparently is a sign of a disproportionately high social burden and lower health spending overall. Cooper (2009, JAMA) reviews the context of the debate and suggests the distribution of poverty may be responsible for many of the unexplained variations in both health inputs and health outcomes. More specific variations with respect to poverty are detailed in Cooper et al. in the Journal of Urban Health (2012). Doyle (2007, NBER Working Paper No. 13301) found that people visiting Florida who endured health emergencies had lower mortality if they went to hospitals in high-spending areas than if they went to hospitals in low-spending ones.

Comments (16)

Trackback URL | Comments RSS Feed

  1. Studebaker says:

    New Results: Hospitals that Spend More on Patients Have Lower Mortality Rates

    This isn’t rocket science. Hospitals that spend more are (by definition) “wasting resources” on most people. But, at the margin, there are a few outliers that are saved who otherwise would not survive. If someone else is paying your bill (or you’re one of the outliers) you like the Big Spending hospitals. If you’re paying your own bills, you may want a provider who’s more conservative. There’s nothing inherently wrong with big spending or low spending hospitals. What is wrong with our system is that consumers are so far removed from the decisions that spending is out of sync with actual market demand (i.e. quantity of care demanded at a given price).

  2. Cindy says:

    “Teaching and high tech hospitals both tend to have higher average costs. Being taken to a teaching hospital reduces average mortality by 3.9 percent, over 10 percent of average mortality.”

    I wonder about this. It would seem that teaching hospitals will have higher costs largely because they have to fund instructional resources. I wonder how much of the “better patient outcomes” are correlated with just the higher volume of available students/residents/etc. versus non-teaching hospitals where less personnel is available to check up on patients.

  3. Peterson says:

    When my father had a stroke, he was taken to a local hospital to be cared for. The doctor there was very nice and seemed to very smart and helpful. He came in regularly (3-4 times a day) to check on my father, and was very imformative/attentive. However, after he was released my Dad continued to have severe headaches, and other pains that were stroke related. My mother then checked my father into the teaching hospital that she worked at in order to see if they could solve the problem. I remember much more people coming in to check on my father. The student that was assigned to him was probably around once every hour or two, and the physician 2-3 times a day. This, along with routine nurse checkups, I believe led to a much better diagnosis of the problem. After being discharged, my father stopped having anymore pains, and has not had any stroke related problems since.

    I think the number of observations they have at teaching hospitals, and the larger number of interactions with the patient lead them to much better solutions than non-teaching hospitals.

  4. James Mule says:

    Doesnt educational hospital leave more room for error and mis-diagnosis?

  5. Paul Rueter says:

    Way to point this to the general public..

  6. Charlotte Spencer says:

    I would think that teaching hospitals would have lower average costs than high tech hospitals, or any other hospital for that matter. Aren’t they “employing” doctors in training that lack the experience and knowledge of most physicians working in every other “regular” hospital? Logic tells me that, just for that reason, costs paid by these teaching hospitals would be lower, since these “doctors in training” don’t get paid as much.

  7. Cindy says:


    It seems to be the case that, in a teaching hospital, the students and interns etc have a huge incentive to be very thorough because they’re trying to stand out from the crowd of other students. I don’t know how to replicate this in non-teaching hospitals — merit raises, maybe?

  8. seyyed says:

    good point, studebaker. it seems inevitable that a hospital that spends more is likey to have fewer deaths. these are doctors that push for treatments that may have a 95% chance of failure, but do it enough times and you’ll get people that survive the intense treatments.

  9. Donna says:

    Thank you, Linda, for compiling this info here!!!

  10. MarkH says:

    Doesnt educational hospital leave more room for error and mis-diagnosis?

    The data routinely shows teaching hospitals and academic centers perform better than private hospitals for a lot of the reasons above, including more staff, more advanced services, and expert care. One mistake people routinely make about teaching hospitals is the belief they are being experimented or practiced on by inexperienced personnel, but this is dismissive of the intense oversight and guidance provided by attendings. Medical students make no decisions about care, but they may, for instance, develop a plan of care, present it to an attending, and that plan will then be revised and ultimately implemented by a resident or intern. The lowest level of resident – the intern, implements many of the plans but has a lesser role in developing them than the senior residents and attending. Ultimately, the attending is responsible for all the actions of the residents and students under their supervision, so in my experience, they delegate responsibility carefully, with the most experienced residents given the most leeway, while maintaining a supervisory role on the plan for every patient, every day.

    So basically, in a teaching hospital, you have teams of people thinking about, debating, and planning your care, with supervision by a very experienced physician, immediate access to all sorts of other experts, and consultants, physicians in-house at all times, and often the more advanced technology. At a private non-teaching hospital, each attending is usually on their own, or even delegate the entirety of care to a hospitalist.

  11. Robert says:

    All this and more in this week’s exciting issue of: You Get What You Pay For!

  12. Tom C says:

    My mother was on medicare and whent into the hospital with diverciulites. She passed away 21 days later at a cost of $258k or $512 per hour 24 hours a day. She was in the ICU for 3 days then out in a room where they monitored her potential recovery. The day she died the hospital made us move her back to her home, at great pain to her? My suspician was that they did not want her to be pernounced dead in their hospital? The costs of hospital care is rediculus and something needs to be done about it!!!

  13. Ashley says:

    “They find that a 10 percent increase in inpatient spending is associated with a 3.1 to 11.3 percent increase in 30 day survival”

    That doesn’t exactly seem like a good outcome, but it must better than decreasing survival.

  14. James R Chaillet, Jr. ,MD says:

    With hospitals inpatient beds being filled with sicker patients due largely to advances in outpatient medications and diagnostic and treatment technologies and processes ( readily available MRI and CT, ambulatory surgery, almost instant cardiac catheterization and stent placement and the like)the afforementioned improvements are not surprising. With diseases affecting multiple organ systems with multiple interactions in people with generally failing and aging systems, more intense and frequent interventions and more intensive oversight is what keeps some of these people alive longer.

  15. JayB says:

    Here’s to hoping that Uwe Reinhardt et all will attempt to address the findings in this growing body of literature.

    It seems obvious that spending and outcomes can’t be meaningfully analyzed without controlling extensively for the differences in the composition of the patients that they treat. Even if such controls are in place – one would hope that it would be common practice to temper the conclusions drawn from geographic-variation data in light of these factors.

    IF such qualifications exist – one rarely sees reference to them in the presentations of Orszag et al, much less in the broader non-specialist literature.

  16. Bob Geist says:

    I hope this discussion guides everyone to Richard “Buz” Cooper’s many papers cited that debunk the Dartmouth Atlas (DA) managed care (capitation cure) bias. See especially “More Is More And Less Is Less:Richard A. Cooper Health Affairs, January/February 2009; 28, no. 1 (2009): w124”–and his other papers in that same issue. Yes more care means better quality. The Dartmoth maps of “waste” only depict high rates of spending in areas of poverty needing increased public support. The real problem is to “cure” poverty, not expect captiation pay to be some magic wand to cure a specious DA manufactured evidence-free illusion of “waste”. Bob