It is an empirical question whether malpractice law leads doctors to neglect to take appropriate precautions, or to elect to take inappropriate precautions, or both. Empirical research on the effects of the malpractice system and tort reforms is of three types. The first arm of the literature surveys physicians about their opinion of the role of the malpractice system in determining medical treatments (for example, Studdert et al., 2005
). Although opinion surveys indicate that physicians believe that the existing malpractice system leads to both positive and negative defensive medicine, this approach only provides information about physicians’ self-reported perceptions: it does not measure actual medical decisions.
A second arm of the literature estimates the effects of tort reforms on claim frequency, payments to claimants, malpractice premiums, and other proxies for “malpractice pressure”—the incentives for healthcare providers to shield themselves from legal liability. This arm of the literature reports two main findings: One lesson is that economic loss, rather than fault, is consistently the most important characteristic of claims in determining the probability and size of award (Danzon and Lillard, 1983
; Farber and White, 1991
; Brennan, Sox, and Burstin, 1996
). The other lesson is that “direct” reforms—those that directly reduce expected malpractice awards—reduce measures of malpractice pressure. The most important reforms of this type are caps on damages and “collateral source offsets.” Caps on damages limit a defendant’s financial liability (or some element of liability, like pain-and-suffering or punitive damages) in a successful lawsuit. Collateral source offsets revoke the common-law default rule that the defendant must bear the full cost of the injury suffered by the plaintiff even if the plaintiff were compensated for all or part of the cost by an independent or “collateral” source (such as an insurance company).1
In contrast, “indirect” reforms have had a less-consistent impact on liability and hence on malpractice pressure. One of the most important reforms of this type is limits on joint and several liability. By default, the doctrine of joint and several liability allows a plaintiff to collect all of his damages from any defendant, regardless of the extent of the defendant’s fault. Limits on joint and several liability either abolish the doctrine of joint and several liability, exclude damages for pain and suffering from it, or bar its application if defendants did not act in concert. Other important indirect reforms impose mandatory periodic payments, which require damages in certain cases to be disbursed in the form of an annuity that pays out over time and limits on the contingent fees that plaintiffs’ attorneys can charge. Examples of studies that have estimated the effects of reforms on malpractice pressure include Danzon (1982
); Sloan, Mergenhagen, and Bovbjerg (1989)
; Thorpe (2004)
; Avraham (2007)
and Born, Viscusi, and Baker (2009)
Taken alone, however, estimates from these studies are only the first step toward answering the policy question of interest. They show how tort reforms affect doctors’ incentives; they do not show how tort reforms affect doctors’ behavior. To assess the efficiency of precautionary behavior induced by the liability system, we must compare how the costs of precaution, and losses from adverse events, respond to changes in the legal environment.
A third arm of the literature seeks to make this comparison by quantifying how treatment decisions and patient health outcomes respond to malpractice pressure. The largest segment of this arm measures malpractice pressure with malpractice premiums, claims frequency, or claims severity. Studies using this approach generally find evidence of positive defensive medicine, or unproductive care (for example, Rock, 1988
; Harvard Medical Practice Study, 1990
; and Localio et al., 1993
; but see Baldwin, Hart, Lloyd, Fordyce, and Rosenblatt, 1995
; Kim, 2007
), but no evidence of negative defensive medicine, that is no failure to supply care that could be beneficial (for example, Baicker and Chandra, 2005
; Dranove and Gron, 2005
However, concerns about unobserved differences between providers and between small geographic areas qualify these results. The aforementioned studies use information on claims or premiums at the level of individual doctors, hospitals, or areas within a single state over a limited time period to measure malpractice pressure, but malpractice laws within a state at a given time are constant. Thus, variation in their measures of malpractice pressure may be due to unobserved factors that are correlated with the cost and outcomes of care. For example, the claims frequency or insurance premiums of a particular provider or area may be relatively high because the provider is relatively low quality, because the patients are particularly sick (and hence prone to adverse outcomes), because the patients had more “taste” for medical interventions (and hence were more likely to disagree with their provider about management decisions), or because of many other factors. Since these factors are extremely difficult to capture fully in observational datasets, estimates from these studies represent a combination of the true effect of malpractice pressure on treatment decisions or outcomes and unobserved differences in providers, patients, and areas.
One way to address these concerns is to identify the effects of malpractice pressure with variation in tort law reforms across states and over time. This technique yields unbiased assessments of the impact of malpractice pressure under the assumption that the adoption of malpractice reforms is uncorrelated with unobserved differences across states in the determinants of treatment decisions and health outcomes. This assumption can be criticized; for example, see Danzon (2000)
and Congressional Budget Office (2006)
In the first paper to use this approach (Kessler and McClellan, 1996
), my coauthor and I match longitudinal data on essentially all elderly Medicare beneficiaries hospitalized with serious cardiac illness from 1984, 1987, and 1990 with information on the existence of direct and indirect law reforms from the state in which the patient was treated. We found that reforms that directly limit liability—such as caps on damages—reduced hospital expenditures by 5 to 9 percent in the late 1980s, with expenditure effects that are greater for patients with ischemic heart disease than for patients suffering heart attack (acute myocardial infarction). (Because ischemic heart disease is a less-severe form of illness, those patients may have more “marginal” indications for intensive treatment, leading to a greater scope for defensive practices.) In contrast, reforms that limit liability only indirectly were not associated with any substantial expenditure effects. Neither type of reforms led to any consequential differences in mortality or the occurrence of serious complications. The estimated cost-effectiveness ratio associated with liability-pressure-induced intensive treatment was over $500,000 per additional one-year survivor, with comparable ratios for recurrent heart attacks and heart failure. This level is well above the generally accepted value of a year of life saved of around $100,000 (Cutler and Richardson, 1998
). Thus, this study finds that treatment of elderly patients with heart disease does involve defensive medical practices, and that limited reductions in liability can reduce this costly behavior without ill effects on health.
Two studies identify the mechanism through which tort reforms affect physician behavior, which helps predict the effects of existing reforms under different market conditions and of new, untried reforms. In Kessler and McClellan (2002b)
, we match longitudinal Medicare data with law reforms and data on health insurance markets to explore the ways in which managed care and liability reform interact to affect treatment intensity and health outcomes. We report that direct reforms like caps on damages reduce defensive practices in areas both with low and with high levels of managed care enrollment. At least for patients with less-severe cardiac illness, managed care and direct reforms are substitutes in the effort to reduce defensive practices, so the reduction in defensive practices that can be achieved with direct reforms is smaller in areas with high managed care enrollment.
In Kessler and McClellan (2002a)
, we integrate four unique data sources to illuminate how reforms affect malpractice pressure and how reform-induced changes in the incentives provided by the liability system affect treatment decisions, medical costs, and health outcomes. That paper matches by state and year the longitudinal Medicare data discussed above (updated to include all years from 1984 to 1994) with data on law reforms, physician-level data on the frequency of malpractice claims from the American Medical Association, and malpractice-claim-level data from the Physician Insurers Association of America. We show that direct reforms improve medical productivity primarily by reducing malpractice claims rates and compensation conditional on a claim, which suggests that other policies that reduce the time spent and the amount of conflict involved in defending against a claim can also reduce defensive practices. A representative finding is that, at least for cases involving elderly heart disease patients, an untried reform that would reduce the legal-defense burden on physicians and hospitals by one-quarter—which is within the range of policy possibilities—could be expected to reduce medical treatment intensity by approximately 6 percent but not to increase the incidence of adverse health outcomes. In the same population, a policy that expedited claim resolution by six months across-the-board could be expected to reduce hospital treatment costs by 2.8 percent without greater adverse outcomes. This finding is consistent with Kessler and McClellan (1997)
, where we report broad differences in physicians’ perceptions of the impact of malpractice pressure in states with and without liability reforms.
Several more recent studies update and refine these estimates of the effects of tort reform, generally finding a small but significant impact on positive defensive medicine. Hellinger and Encinosa (2006)
find that states adopting caps on noneconomic damages have 3–4 percent lower overall health spending than states that do not. Baicker, Fisher, and Chandra (2007)
find that higher malpractice awards and premiums are associated with higher Medicare spending, especially for imaging services that are often believed to be driven by physicians’ fears of malpractice; Smith-Bindman, McCulloch, Ding, Quale, and Chu (forthcoming) confirm the latter finding with an analysis that shows greater use of emergency CT and MRI scans in states without reforms. Sloan and Shadle (2009)
find that direct reforms have a negative but statistically insignificant effect on Medicare spending, with no important effect on health outcomes. Avraham and Schanzenbach (2010)
show that the effect of reform-induced reductions in health spending translates into increases in private health insurance coverage.
Other work examines the incidence of defensive practice in obstetrics. Using national birth certificate data, Dubay, Kaestner, and Waidmann (1999)
show that malpractice-claims risk leads to increased rates of Caesarian-section births, but no better birth outcomes. Similarly, Yang, Mello, Subramanian, and Studdert (2009)
find a negative effect of caps for C-section rates; although Currie and MacLeod (2008)
(also looking at C-sections) find a positive effect of caps and a negative effect of limits on joint and several liability.
As the evidence about positive defensive medicine has accumulated, its conclusions have become more widely accepted. For example, the Congressional Budget Office (2006)
found that adoption of direct reforms leads to significantly lower Medicare hospital spending per beneficiary, but questioned whether their findings, and mine with Mark McClellan, are valid estimates of the causal effect of reforms. However, the Congressional Budget Office (Elmendorf, 2009
) revised this earlier assessment of the effects of tort reform and concluded that “the weight of the empirical evidence now demonstrates a link between tort reform and the use of healthcare services.”
Evidence of negative defensive medicine is more mixed. In one of the few papers that measures how malpractice liability leads to patient avoidance, Dubay, Kaestner, and Waidmann (2001)
examine the effect of tort reforms on prenatal care use and infant health. They find that malpractice pressure results in prenatal care beginning later in the pregnancy, although without significant harmful effects on infant health. In Kessler, Sage, and Becker (2005)
, my coauthors and I apply the approach of Kessler and McClellan (1996)
to assess the impact of reforms on physician supply. We match data from the American Medical Association’s Physician Masterfile on the number of practicing physicians in each state for each year from 1985 through 2001 with law reforms and data on healthcare markets. We find that three years after adoption, direct reforms increase physician supply by 3.3 percent, controlling for fixed differences across states and other time-varying state characteristics; using similar methods, Encinosa and Hellinger (2005)
report even larger effects. In contrast, Klick and Stratmann (2007)
and Matsa (2007)
find no effect of reforms on aggregate physician supply, but a significant positive effect on the supply of physicians in high-malpractice-risk specialties and rural areas, respectively.
Future research might examine the effects of a new type of tort reform on both positive and negative defensive medicine: restrictions on the legal discoverability of information gathered as part of private, voluntary error-reporting systems. In theory, restrictions on discoverability have an ambiguous effect on patient welfare. On one hand, limiting injured plaintiffs’ access to information could reduce incentives for appropriate care; on the other hand, it could reduce defensive practices and increase incentives for quality improvement. States have historically differed in the legal protection they gave to analyses done by hospitals, physician groups, and insurers that sought to identify the cause of medical errors. Some states restricted the extent to which such analyses could be used as evidence against defendants in malpractice cases, and some states did not (for a discussion, see Scheutzow, 1999
). In 2005, Congress passed and President Bush signed the Patient Safety and Quality Improvement Act (PL 109-41), which restricted state law discoverability of certain activities undertaken by “Patient Safety Organizations” that register with the U.S. Department of Health and Human Services. There is some empirical evidence that the state laws in effect before the adoption of this law provided less than the optimal amount of protection for voluntary error-reporting systems. In a series of case studies, the Institute of Medicine (2000)
recommended expanding peer review protections along the lines eventually adopted by the federal act. To date, however, no study has examined the effect of such “peer review protection” laws. Liang, Riley, Rutherford, and Hamman (2007)
propose several detailed hypotheses about the effects of this law on patient safety that deserve consideration. Greenberg, Haviland, Ashwood, and Main (2010)
investigate the relationship between adverse patient safety events and malpractice pressure in ways that suggest that this may be a promising area for reform.