Are doctors more "defensive" than studies show?

Aaron Carroll takes a whack at the alleged myth that medical tort reform has the potential to significantly reduce healthcare costs. Citing two studies published in the September 2010 issue of Health Affairs, he reports findings from the first that our medical liability system accounts for only an estimated 2.4% of total U.S. healthcare costs (a nontrivial $55.6 billion), and from the second that tort reform would reduce costs by a mere one tenth of one percent.

Carroll originally published this post as a guest blogger for Ezra Klein, who has often cited similar findings -- as did Obama, in the bipartisan healthcare summit he called in February 2010.  Personally, I am deeply suspicious of the Republican passion for tort reform, medical and otherwise, which is part and parcel of their aversion to holding any industry accountable for any damages it causes.  Nonetheless, claims that the medical liability system has only a modest effect of doctors' behavior always arouses my skepticism.  This skepticism comes in large part from listening to my wife, a nurse-midwife at an inner city hospital, who is forever frustrated by the obstetricians' hair-trigger for Caesarian sections. Her colleagues are plainly strung taut by the constant threat of lawsuits; whenever anything goes wrong everyone anticipates a suit. I strongly suspect that the threat of liability affects caregivers' behavior in ways that do not show up in the research.  And I have questions about both of the studies Carroll cites, and about Carroll's presentation of them.

First, the two studies are to a certain extent at cross-purposes.  Carroll as noted above, reports that the first study [purchase required], led by Michelle M. Mello in a team including Atul Gawande, pegs the cost of our "medical liability system" at 2.4% of total healthcare costs and finds that the vast majority of that cost comes from the practice of defensive medicine.  Carrol then notes that the second study, led by J. William Thomas, pegs the potential savings from tort reform at just .1% of total costs.  But in the Mello study, defensive medicine costs and tort reform savings are presented as one and the same thing. The authors estimate of total, system-wide costs is based entirely on a series of studies by Kessler and McClellan conducted from 1984--1994. All the results, and the Mello authors' extrapolations from them, use those two terms interchangeably (my emphasis):

Kessler and McClellan examined the effect of tort reforms that directly reduce expected malpractice awards—such as caps on noneconomic damages—on Medicare hospital spending for acute myocardial infarction and ischemic heart disease from 1984 to 1990.⁷ The reforms lowered hospital spending by 5.3 percent for myocardial infarction and 9.0 percent for heart disease....

In subsequent work examining data through 1994, Kessler and McClellan found that such direct reforms reduced hospital spending by 8.3 percent, but this estimate was based only on data about myocardial infarction.²⁹ In a further analysis incorporating information about levels of managed care through 1994, they estimated that direct reforms reduced hospital spending by 3.8 percent for myocardial infarction and 7.1 percent for heart disease... ³⁰

In our analysis, we used a value of 5.4 percent for the effects of defensive medicine on hospital spending, a conservative assumption that represents the lower of Kessler and McClellan’s original estimates and the midpoint between their latest estimates.

The Thomas study is based on much more current and vastly more expansive data: a database of more than 400 million paid medical and pharma claims from CIGNA, 2004-2008, from which they ended up tabulating costs for 15.4 million treatment episodes. This study does not measure direct effects of tort reform, however. Instead, it correlates episode costs with increases in malpractice premiums. While it finds significant relationships in many high risk treatment categories, it finds the impact of premium changes to be extremely modest:

Across all specialties, reductions in medical malpractice premiums would lead to statistically significant savings in 2.0 percent of the conditions analyzed, but these are high-volume situations, comprising 35.8 percent of all episodes. However, the magnitude of savings that could be realized is small, accounting for less than 1 percent of medical care costs in every specialty. Across all thirty-five specialties, savings associated with a 10 percent premium reduction in medical malpractice premiums would be just 0.132 percent.

Extrapolating from the older and far more limited studies, in other words, Mello et al forecast almost twenty times as much savings from tort reform as do Thomas et al. That leaves me suspecting that the conclusions to be drawn from both studies are, shall we say, rather tentative. And while the Thomas study is based on a vastly larger data set, I'm not convinced that its conclusions are more reliable. In fact I'm not convinced that the Mello study does not significantly underestimate the current system's effect on doctors.

The Thomas study writeup is purely statistical: it offers no analysis or speculation regarding the relationship between malpractice insurance premiums and treatment decisions. It directly eschews such speculation:

There is a caution that applies to our cost estimate, as well as those of all empirical analyses of defensive medicine, including the CBO study referred to above.24 Empirical analyses such as ours attempt to quantify relationships between differences in physicians’ perceived liability risk and associated behavior—in our case, behavior that influences health care costs. If some level of defensive behavior is shared by all or most  physicians and is independent of variations in perceived risk, it will not be detected in studies such as this. However, we believe that such behavior would not be influenced by changes to the medical liability environment, and that the costs associated with such behavior are unlikely to be affected by tort system  reforms, regardless of the level of reduction in medical malpractice insurance premiums.

That "belief" is not really explained.  Is a direct correlation between malpractice premiums and cost of treatment the only way to assess the potential impact of reduced medical liability?  All tort reforms are not created equal. Some might reduce liability awards more than others; others might reduce the number of suits that go forward; some states regulate insurance more tightly than others and so will insist that cost savings be passed more fully to the physician policyholders; different regions of the country are probably more or less litigious; and medical culture and incentives differs in different regions.  Perhaps the impact of the most potentially effective forms of tort reform -- if any effective means have really been tried yet (award caps are far from the whole arsenal) -- was simply lost in a sea of data? The analysis did of course include multiple controls, including for state, and there are technical appendices to which I don't have access. I may be out of my depth here. Perhaps no tort reform thus far implemented has had any significant impact on costs. But it seems to me at least possible that doctors are as influenced by the likelihood of being sued at all as by the potential size of loss (and resulting rise in insurance premium) per suit. And the former variable would not be captured by this study.

As noted above, the Mello findings regarding the impact of tort reform on practice of defensive medicine are wholly extrapolated from the Kessler/McClellan studies of  the effect of specific tort reforms that reduced expected malpractice awards on Medicare hospital spending for specific forms of heart disease from 1984-1994. That impact was found to be pretty dramatic, reducing hospital spending on one form of treatment by 8.3%. Still, there are various grounds for suspecting that even this study may be underestimating the possible impact of defensive medicine:

1) Mello et al themselves define their own nationwide extrapolation -- "a value of 5.4 percent for the effects of defensive medicine on hospital spending" -- as "conservative," a word they use repeatedly to describe their estimates.

2. The costs of defensive medicine may throw out tentacles that escape researchers' attention Carlos Badiola of Bristol Hospital, in a response to the Mello article, writes:

...I did not see any mention of the iatrogenic costs of defensive medicine.
When calculating the costs of defensive medicine, one must consider the costs of iatrogenesis associated with unnecessary testing, especially imaging. These include radiation risk, false positives, follow-up testing for "incidentalomas" (with additional radiation or invasive procedures), and the defensive interpretation of tests, including imaging studies and biopsies. Given that much of defensive medicine involves imaging, and in light of the limited specificity of imaging studies, this may result in an almost incalculable cascade of added costs.

3) Liability-driven practice of defensive medicine interacts with other factors in ways impossible to measure. Mello at all write:

Studies indicate that physicians in group practices preserve their net income in the face of malpractice premium increases by increasing both the volume of services they perform and the unit prices they charge.^37,38 About half to three-quarters of physicians’ response takes the form of higher volume; price effects are comparatively modest. 

It is impossible to determine how much of the increase in volume constitutes defensive medicine—services performed primarily to reduce liability risk—as opposed to services performed primarily to enhance revenue

 The studies cited here would seem to contradict the Thomas study.  Still, it seems at least possible that profit motive and fear of being sued interact. American doctors can justify their extremely high incomes (relative to those of doctors in other OECD countries) by pointing to their higher expenses on multiple fronts: administrative, educational, liability. Comfort with being "entrepreneurial" is probably reinforced by awareness of liability and risk. Then too, the salience of liability exposure may justify and partly mask the prospect of financial reward. Jonas B. Green of the UCLA Medical Center, in a response to the Mello article, writes:

The usual threshold for ordering of goods has been lowered in health care by invisible pricing and third-party payment. Other than the risk of testing itself—borne entirely by patients and commonly underestimated by both parties—the decision to perform a test aligns patient and physician interests. Absent other resistance in the system, a physician feeling only slightly more protected by an additional test may be all that is necessary to ensure it gets ordered. Perhaps when prices erroneously seem nominal, rational calculation of defensive risk, too, becomes distorted?

4. More broadly, fear of liability may feed into the culture of overutilization outlined by Ezekiel Emanuel, who, in a 2008 article, cites four physician-related factors that "drive overuse" of medical treatment in the U.S. These include values, imparted in medical school, that place a premium on "enumerating all possible diagnoses and tests that would confirm or exclude them"; the perverse incentives created by fee-for-service (do more, get paid more); saturation marketing from drug companies; and finally, medical malpractice laws. Significantly reducing one of these mutually reinforcing factors may not by itself change behavior immediately, or dramatically.   But in tandem with, say, more doctors going on salary and/or into Accountable Care Organizations, effective tort reform could have a multiplier effect on the culture of overutilization.