marketwatch - atul gawande€¦ · of malpractice claims. the pool of medical in-juries (stage 1)...

M A R K E T W A T C H

MARKETWATCH

Disclosure Of Medical Injury To Patients: AnImprobable Risk Management StrategyMovement toward full disclosure should proceed with a realisticexpectation of the financial implications and prudent planning tomeet them.

by David M. Studdert, Michelle M. Melio, Atui A. Gawande, Troyen A.Brennan, and Y. Ciaire Wang

ABSTRACT: Pressure mounts on physicians and hospitals to disclose adverse outcomes ofcare to patients. Although such transparency diverges from traditional risk managementstrategy, recent commentary has suggested that disclosure wiil actually reduce providers' li-ability exposure. We tested this theory by modeling the iitigation consequences of disclo-sure. We found that forecasts of reduced iitigation volume or cost do not withstand closescrutiny. A policy question more pressing than whether moving toward routine disclosurewiil expand litigation is the question of how large such an expansion might be. [Health Af-fairs 26, no. 1 (2007): 215-226; 10.1377/hithaff.26.1.215]

ONE OFTHE MOST intriguing aspects aspects of their care.'of the modern patient safety move- However, forthrightness about injuries andment is the mounting pressure on errors is at odds with the traditional approach

physicians and hospitals to be more open and to risk management in health care, which em-honest with parients when things go wrong phasizes caution, minimal comment, and evenin care.' There is broad consensus that disclo- cover-up. Wili more disclosure therefore meansure of unanticipated outcomes is desirable, more litigation? An emerging view in policyRegulators have begun to require it.̂ The ra- discussions asserts just the opposite: namely,tionale is clear: The experience of other in- that the bunker mentality of traditional riskdustries, such as aviadon and nuclear power, management is flawed, it stokes rather thansuggests that openness about error is crirical staves off litigation, and an ancillary benefit ofto development of effective prevention strate- disclosure will be its salutary impact on pro-gies. There are also compelling ethical rea- viders' liability exposure. Much recent com-sons for telling parients the truth about all mentary has propounded this view."*

David Studdert ([email protected]) is a Federation Fellow and professor in the Schools of Law andPopulation Health at the University of Melbourne, Australia. At the time this paper was written, he was anassociate professor in the Department of Health Policy and Management, Harvard School of Public Health, inBoston, Massachusetts. Michelle Mello is an associate professor in that department. Atul Gawande is an assistantprofessor in the Department of Surgery, Brigham and Women's Hospital also in Boston. Troyen Brennan is chiefmedical officer at Aetna in Hartford, Connecticut. Claire Wang is a research associate in the Department of HealthPolicy and Management, Harvard School of Public Health

HEALTH AFFAIRS - Volume 26, Number 1 215

DOI 10.1377/hltha£f.26.1.215 C2007 Project HOPE-Tlie People-to-People Health Foundation, Inc.

H E A L T H T R A C K I N G

The notion of disclosure as an effective riskmanagement tool is grounded in the belief thatsome patients who would have sued will not ifearly and candid disclosure occurs, becausethey will come to understand that their injurywas not attributable to neghgence, or wiU feelless anger toward a provider who deals withthem honestly, or both. It is also sometimesposited that providers' candor wiU induce pa-tients who do sue to settle their claims for lessmoney. However, the opposite consequencealso warrants careful consideration; After be-ing confronted with information about theirinjury and its cause, some patients who wouldnot otherwise have sued might be moved to doso. No research to date has evaluated disclo-sure's impact in terms of the balance between"deterred" and "prompted" claims.

We hypothesized that the number and costof prompted claims would negate—and possi-bly even trounce—any deterrent effect of dis-closure on litigation. Our skepticism stemsfrom two empirical insights gained in previousresearch. First, the vast majority of patientswho sustain medical injury never sue, whichcreates a huge reservoir of potential claims.̂

Second, socio-legal researchers have identifiedthe failure of aggrieved people to recognizetheir condition or attribute it to an externalcause as important factors in explaining whythey do not seek legal redress.* To test the hy-pothesis, we modeled the litigation conse-quences of disclosure by combining existingdata on the epidemiology of medical injuriesand malpractice claims with expert opinionabout patients' likely reactions to disclosure.

Study Data And Methods• Conceptual nnodel. Exhibit 1 conceptu-

alizes the impact of disclosure on the volumeof malpractice claims. The pool of medical in-juries (Stage 1) sphts into two groups; thoseattributable to neghgent care and those due tonormegligent causes (Stage 2). Some injuredpatients from each group sue, and some do not(Stage 3), with the total number of suers ulti-mately defining the number of claims (Stage4). Disclosure's impact occurs at Stage 3. Tran-sitions A and C represent deterred claims;transitions B and D are prompted claims. Thenet impact of disclosure on claims volume de-pends on the relative size of these four transi-

EXHiBiTlConceptuai Model Of Impact Of Disclosure On The Litigation Behavior Of Patients WhoExperience Severe iVIedicai injuries

stage 1 Severe medical injuries535,772

Stage 2 Not due to negligence355,244

Due to negiigence180,528

Stage 3 Do not sue340,111

Sue15,133

Disclosure

Sue30,266

Do not sue150,262

Stage 4Total maipractice claimsinvolving severe injuries

45,399

SOURCE: Data derived from various sources, as described in the online Technicai Appendix: http://content.heaithaffairs.or^cgi/content/fuli/26/V215/DCl.

216 January/February 2007


tions.• Injury and claim estimates. We used

previous research and publicly available datasources to derive armual narional estimates foreach cell in Exhibit 1. Estimates of injury prev-alence and type came from the New YorkMedical Practice Study (NYMPS) and theUtah-Colorado Medical Practice Study(UTCOMPS), the leading population-basedstudies of medical injury in the United States.^Claims estimates were drawn primarily fromthe Narional Pracririoner Data Bank.̂

The estimates and our analyses focus on se-vere injuries, defined as injuries with a score of4 (major-temporary disability) or higher onthe Narional Associarion of Insurance Com-missioners (NAIC) severity scale.' Two con-siderarions led us to narrow the analysis in thisway. Eirst, there are no strong systemwideprevalence estimates for minor injury." Sec-ond, minor injuries are rarely the basis ofclaims." They are poor candidates for lawsuitsbecause attorneys, who are generally paid on acontingent fee basis, have little economic in-cenrive to bring them. Therefore, there is lim-ited scope for disclosure to affect lirigarionover minor injuries.

• Estimates of compensation costs.We used average compensation costs of$141,469 and $33,683 per claim for severe in-jury claims with and without negligence, re-specrively.'̂ These are uncondirional averages,not averages among paid claims only.

Aside from its impact on claims volume,commentators have suggested that disclosuremight reduce average payments." To examinethis effect, we projected the impacts of disclo-sure separately under two cost assumprions.In the first analysis, we used the fuU average-cost figures for severe injury claims with andwithout negligence, as noted above; in the sec-ond analysis, we reduced these figures by 40percent.

The reducrion we appHed was based onavailable data on the composirion of medicalmalpractice payments.''' Although they arelimited, these data suggest that noneconomic("pain and suffering") damages consritute ap-proximately 40 percent of total payments. Pa-

rients negoriating with hospitals for compen-sarion following a disclosure are unlikely to bewilling to forgo reimbursement for economiclosses but might be willing to accept greatlyreduced pain-and-suffering compensarion aspart of an expeditious settlement.'^ Hence, 40percent approximates the upper limit of pri-vately and socially acceptable reducrions incompensarion costs.

• Transition parameters and survey ofexperts. Transirions A, B, C, and D in Exhibit 1jointly determine the net impact of disclosureon Hrigadon. Their respecrive magnitudes areunknovwi. To obtain expected ranges for them,we surveyed medico-legal experts in Septem-ber 2005.

The experts were a convenience sample ofseventy-eight people whom we recognizedfrom their publicarions or professional experi-ence as having relevant experrise. They camefrom fourteen states and the following profes-sional categories: senior patient safety (31)and/or legal researchers (17); hospital-basedrisk managers or quality assurance directors(25); senior staff from malpracrice Uability in-surers (12); plaintiffs' attorneys (8); and hospi-tal execurives or general counsels (7). (Thesecategories are not mutually exclusive.) Thesample contained seventeen pracricing physi-cians and eleven pracricing attorneys.

A written survey presented four hypotheri-cal scenarios designed to elicit percentage esri-mates for each of the transirions (Exhibit 2).After each hypotherical, the survey asked, "If100 patients experienced this sequence ofevents, how many wovJd react to the disclo-sure in the way that [this padent] did?" Re-spondents were directed to mark their bestguess, lowest reasonable number, and highestreasonable number on a rating scale, with re-sponse oprions running from 0 to 100 in five-unit increments.

The survey instrucdons defined serious in-jury as "injury that leaves the parient with ei-ther permanent disability, or with temporarydisability that is very severe while it lasts." Thesurvey did not specify the circumstances of thedisclosure, such as whether an apology orcompensation was offered, patients were

HEALTH AFFAIRS - Volume 26, Number J 217


EXHiBiT 2Four Hypotheticai Scenarios Regarding Patient injury, Disciosure, And SuingBehavior From A Survey Of Experts

Transition represented* Scenario

A Patient A sustains a serious injury. The injury occurs as a result of negligentmedical care. Patient A plans to sue. A timeiy disciosure then occurs whereinthe patient is told that an injury has occurred. A full explanation of how andwhy the injury occurred is given. After the disclosure, Patient A changes hismind about litigation and decides not to sue.

B Patient B patient sustains a serious injury. The injury occurs as a result ofnegligent medicai care. Patient B has no plans to sue. A timeiy disciosurethen occurs wherein the patient is told that an injury has occurred. A fuiiexplanation of how and why the injury occurred is given. After the disclosure.Patient B changes her mind about iitigation and decides to sue.

C Patient C sustains a serious injury. The injury is caused by nonnegiigentmedical care. Patient C plans to sue. A timely disclosure then occurs whereinthe patient is toid that an injury has occurred. A fuii explanation of how andwhy the injury occurred is given. After the disclosure, patient C changes hermind about litigation and decides not to sue.

D Patient D sustains a serious injury. The injury is caused by nonnegiigentmedical care. Patient D has no pians to sue. A timely disciosure then occurswherein the patient is told that an injury has occurred. A full explanation ofhow and why the injury occurred is given. After the disciosure, patient Dchanges her mind about litigation and decides to sue.

SOURCE: Data derived from survey responses and the authors' own analyses.NOTES: Experts were asked to place three check marks on the scaie provided beneath each scenario: their best guess at thenumber of patients (out of 100) who would change their plans to sue or not sue: their estimate of the lowest reasonablenumber; and their estimate of the highest reasonable number. For further detaiis of the scenarios and instructions in thesuney, see the oniine Technical Appendix: http://contenthealthaffairs.Org/cgi/content/fuil/26/l/215/DCl.

•The letters correspond to the transitions shown in Exhibit 1.

aware of their injury prior to the disclosure, or tional Academies, and its use in pohcy researchanimosities surrounded the event. Rather, the is increasing, with applications ranging fromsurvey directed respondents to consider their evaluations of care dehvery options to the in-practical experience with disclosure and hd- fluence of financial confhcts of interest.'^gation, acknowledged that some of these fac- In this study we used a Monte Carlo ap-tors might be present some of the dme, and proach to incorporate uncertainty associatedasked respondents to contemplate a typical with the four transidon parameters, as de-series of disclosures at their insdtudon (or in- tected in the experts' responses. Thus, rathersdtudons with which they had experience).'* than estimating single values for the impact of

• Monte Carlo simuiations. When the routine disclosure on the volume and cost ofinputs of a predicdon model are uncertain, the claims, our model generated probabihty distri-model's outputs should take account of and re- budons of these outcomes,fleet that uncertainty. Monte Carlo Simula- Specifically, we fit a beta distribudon baseddons extend scientific judgment beyond de- on each expert'sjudgments (lowest reasonableterminisdc point estimates by calculating the estimate, best guess, highest reasonable esd-likelihood of various outcome scenarios, based mate) for each transidon (A, B, C, and D)."on the degree of uncertainty around the in- This probabihty distribudon represented theputs.'^ The approach has been embraced by the expert's uncertainty regarding the parameterEnvironmental Protecdon Agency and the Na- in quesdon and was bounded by the upper and



lower estimates.^" The model then sampledprobabilistically from the four transition dis-tributions and appHed the transition percent-ages selected to the relevant injury counts tocalculate a projected number of claims. We re-peated this sampling procedure 1,000 times foreach expert and then combined all experts'predictions into a single probability distribu-tion of the expected number of claims and as-sociated costs. This approach, a modified formof the method proposed by Peter Doublet andcolleagues, gives each expert equal weight andaccounts for potential within-subject correla-tion.^

We calculated means and medians of the fi-nal probability distributions. We also exam-ined the probabilities of changes in volumeand cost in relation to the status quo. AH anal-yses were conducted using SAS version 9.2.

Study Results• Expert survey. Sixty-five experts com-

pleted the survey (response rate: 83 percent).The experts predicted that among patientswho experienced severe injury as a result ofnegligence, disclosure would on average deter32 percent from suing and prompt claims by 31percent of patients who would not otherwisehave sued (Exhibit 3). Among patients whoseinjuries were not due to negligence, the deter-

rent impact was perceived to be greater: Dis-closure would deter an average of 57 percent ofsuers and prompt 17 percent of nonsuers.

• Impact on claim volume. Based on theexperts' predictions about injured patients' re-sponses to disclosure, the model computed a 5percent chance that total claim volume woulddecrease or remain unchanged and a 95 per-cent chance that it would increase (Exhibit 4).These probabilities correspond to the cumula-tive size of the bars shown in the distribution.For example, the bars to the left of the statusquo, in aggregate, account for approximately 5percent of the total area of the distribution,and the bars to the right account for approxi-mately 95 percent. The distribution also indi-cates a 60 percent chance that comprehensivedisclosure of severe injuries would at leastdouble the annual number of claims nation-wide and a 33 percent chance that volumewould increase by threefold or more. The me-dian of the distribution is an increase of 70,974claims (to 127,723 claims in total), which repre-sents a 125 percent increase over the currentlevel.

• Impact on compensation costs. Un-der the assumption that average payments donot change, the model predicted a 6 percentchance that total direct costs of compensationwould decrease or remain unchanged under

EXHIBIT 3Expert Survey Responses: Deterred And Prompted Claims Following Disclosure OfMedical Injury

Injuries due to negligence

Scenario A: percent of claims deterred Scenario B: percent of claims prompted

MeanStandard deviationMedian

MeanStandard deviationMedian

Lower

181410

Best guess Upper

32 4823 2525 40

Injuries not due to negligence

Scenario (

412140

!: percent of claims deterred

57 7323 2160 • 75

Lower

191510

Scenario D:

865

SOURCE: Data derived from survey responses and the authors' own analyses.

Best guess Upper

312025

492245

percent of claims prompted

171315

291325

HEALTH AFFAIRS - Volume 26, Number 1 219


EXHIBIT 4

Monte Carlo Simuiation Of impact Of Disciosure Of IVIedicai Injury On Volume Of

iViedlcai IVIaipractice Ciaims

-100 0 100

Percent increase

200 300 400 500 600

Frequency distribution (percent)

10

8

6

4

2

0

Status quo56,749" r

i HI i'*'I

^ S 1 S i ')

Median127,723

_ .ii _ £ . .V

n •

10,000 60,000 110,000 160,000 210,000 260,000 310,000 360,000 410,000

Annual incidence of ciaims

SOURCE: Data derived from the authors' analyses.» The simulations used the estimated total volume of aii claims nationwide (56,749), not just the subset invoiving serious injury(45,399), as the status quo. Details of the basis of these estimates are provided in Part A of the oniine Technical Appendix:http://content.heaithaffairs.org/cgi/content/fuii/26/V215/DCl.

routine disclosure and a 94 percent' chance least double and a 24 percent chance that they

that they would increase (Exhibit 5). There is a would increase by threefold or more. The me-

45 percent chance that total costs would at dian of the distribudon is an increase of $5.5

EXHIBiT 5

Monte Carlo Simuiation Of impact Of Disciosure Of Medicai Injury On Compensation

Costs (Assuming Tiiat The Average Payment Size is Unciianged)

-100

Percent increase

100 200 300 400

Frequency distribution (percent)

5

Median$11.3 biiiion

Status quo3 $5.8 biiiion

iiu.0.5 3.0 5.5 8.0 10.5 13.0 15.5 18.0 20.5 23.0 25.5 28.0 30.5

Compensation costs (biliions of doliars)

SOURCE: Data derived from the authors' analyses.



billion (to $11.3 billion), a 95 percent increaseover current compensation costs.

Under the assumption that disclosure re-duced average payments by 40 percent, a netincrease in costs remains more likely than a de-crease or no change (72 percent versus 28 per-cent), and there is a 34 percent chance that to-tal costs would at least double (Exhibit 6). Themedian of the distribution indicates an in-crease of $1.4 billion (to $7.0 billion), a 24 per-cent increase in total compensation costs.

• Sensitivity anaiyses. Reductions in inci-dence ofiryury. Although there is little evidencethat rates of adverse events and negligencehave decreased appreciably over the pasttwenty years, with ongoing attention to errorprevention and quality improvement, theymight do so in the future. Indeed, widespreaddisclosure practices could help drive this re-sult. How would reductions in the incidenceof serious medical injury affect our projec-tions?

To explore this, we cut by one-third thenumber of serious injuries attributable to neg-ligence and reran the simulations. The changemodestly affected volume: There remained ahigh probability (90 percent) that claim vol-ume would increase, and large increases werelikely (median of the distribution was an in-

crease of 50,376 claims). The likelihood of costincreases also remained high (84 percent);however, the expected size of those increaseswas about half (median value of the distribu-tion was an increase of $2.7 biUion) that pre-dicted by the original model.

Underdisclosure. The above analyses assumethat every severe injury will be followed by adisclosure and that the disclosure would gen-erally be clear and comprehensible and wouldprovide the essential elements of what hap-pened and why. Although many would regardthese as noble goals, they are obviously unreal-istic. What impact would less than 100 per-cent disclosure have on the estimates?

If the degree of "underdisclosure" were thesame for negligent and normegligent injuries,then the likelihood of increases and decreasesfrom status quo would not change. The effecton the magnitude of the changes would corre-spond to the degree of underdisclosure. Inother words, if only half of severe injuries were(properly) disclosed across the board, themagnitude of the changes predicted wouldhalve. However, if the degree of underdisclo-sure differed by injury type—one previoussurvey of risk managers, for example, sug-gested a greater reluctance to disclose injuriescaused by neghgence—then this would alter

EXHIBIT 6Monte Carlo Simulation Of Impact Of Disciosure Of IVIedicai Injury On CompensationCosts (Assuming That The Average Payment Size Is Reduced By 40 Percent)

-100Percent increase

100 200 300 400Frequency distribution (percent)

8

6

Status quo$5.8 biillon

iVIedian$7.0 biiiion

0.5 3.0 5.5 8.0 10.5 13.0 15.5 18.0 20.5 23.0 25.5 28.0 30.5Compensation costs (billions of dollars)

SOURCE: Data derived from the authors' analyses.

HEALTH AFFAIRS - Volume 26, Number I 221


the litigation impacts we estimated, althoughthe direction and extent of those changeswould depend on the profile of the undis-closed events.̂ ^

Break-even analysis. To test the robustness ofour predictions, we conducted a "break-even"analysis in which we fixed anchor points (bestguesses and upper bounds) for the deterredclaims (transitions A and C) and then calcu-lated the levels to which prompted claims(transitions B and D) would need to descendto maintain the status quo.^' Exhibit 7 showsthe results.

Based on experts' best guesses about de-terred claims, if more than 5.4 percent of negh-gently injured patients and 3.0 percent of non-neghgently injured patients were prompted tosue, overall claims volume would increase be-yond current levels. Alternatively, anchoringdeterred claims at the experts' upper-boundestimates, if more than 12.0 percent and 7.1 per-cent of negligently and nonnegligently injuredpatients, respectively, were moved to sue, theresult would be more claims. (These break-even points fall below the experts' meanlower-bound estimates for prompted claims,as shown in Exhibit 3.) The break-even pointsfor compensation costs are fractionally higher,although very similar.

DiscussionThrough the analyses described in this pa-

per, we found the chances that disclosurewould decrease either the frequency or cost ofmalpractice litigation to be remote. On thecontrary, an increase in htigation volume andcosts was highly likely.

• The great unlltlgated reservoir. Thekey driver of the model's findings is the semi-nal and well-established insight that the num-ber of serious injuries that do not lead toclaims dwarfs the number that do.'̂ '' ("Under-claiming" also appears to be widespreadamong injuries outside the health care sec-tor.)^' Approximately eight in ten setious inju-ries due to negligence and more than nine inten other serious injuries never trigger litiga-tion. This has important implications for dis-closure. Because the stock of unUtigated inju-ries vastly outnumbers the stock of htigatedones, relatively small shifts from the former(prompted claims) will tend to overwhelmshifts from latter (deterred claims).

• Anatomy of prompted claims. The po-tential for disclosure to stimulate htigation hasreceived relatively httle attention. The theoret-ical effect can be decomposed into sequentialsteps. It begins with the notion that the reasonsome injured patients do not sue is that they

EXHIBIT 7Break-Even Analysis Of Claims Volume And Costs Following Disclosure Of MedicalInjury, Based On Experts' Best Guesses And Upper Bounds

Transitions

Anchoring at experts' best Anchoring at experts' upperguesses for deterred claims bounds for deterred cialms(32% and 57%)° (48% and 73%)°

Break-even points for claims volumePrompted claims among severe injuries

due to negligence (Transition B) 5.4%Prompted claims among severe injuries

not due to negligence (Transition D) 3.0%

12.0%

7.1%

Break-even points for compensation costsPrompted claims among severe injuries

due to negligence (Transition B) 6.0%Prompted claims among severe injuries

not due to negligence (Transition D) 3.3%

13.9%

8.3%

SOURCE: Data derived from the authors' own analyses.

"The percentages used in these anchor points are means, as shown in Exhibit 3.



are unaware they have been injured. Patientsconfuse their adverse outcome with their un-derlying disease or the expected effects of theirtreatment. Or they realize that they have suf-fered an adverse event but do not attribute itto substandard care. Previous research into ht-igation behavior has shown that such igno-rance is quite common.^'

In theory, thorough disclosure wHl rid suchpatients of their ignorance about both the ex-istence of an injury and its connection to sub-standard care. A subset of the enlightened willreact by suing. Some will seek to litigate butnot be able to find an attorney; others will se-cure representation and sue. Severe medicalinjuries are more likely to progress througheach of these stages. It is also logical that theoverall transition probabilities predicted byexperts were higher for patients whose inju-ries are due to neghgence (mean of one in threepatients) than for their nonneghgent counter-parts (one in six patients).

• Problems underlying the risk man-agement hypothesis. In hght of our findings,it is interesting to explore the roots of popularperceptions that disclosure will deter htiga-don. We attribute the perception to tenden-cies to both misread and overreach the avail-able evidence.

The evidence base To the best of our knowl-edge, only one study has sought to directly ex-amine the disclosure-claims relationship. Thiswidely cited article by Steve Kraman andGinny Hamm reported on the experience ofthe Veterans Affairs (VA) Medical Center inLexington, Kentucky, which adopted a "radi-cal pohcy of full disclosure" in the late 1980s.̂ ''The analysis compared the number and cost ofmalpractice payments made by the facilitywith those of thirty-five other VA medical cen-ters and found that Habihty payments were"moderate" and "comparable to those of simi-lar facilities." The authors attributed the resultto transparency about substandard care andtimely compensation, although they notedthat the analysis "suggests but does not provethe financial superiority of a fuU disclosurepohcy."

The authors concluded with additional ca-

veats about drawing causal inferences. Therehas been an unfortunate tendency to overlookthese caveats in subsequent references to thestudy. Allen Kachalia and colleagues recentlyelaborated several additional concerns.^' Gen-eralizability is particularly problematic be-cause federal hospitals and clinicians workingin them enjoy broad immunities from tort hti-gation.

Two survey studies pertaining to the dis-closure-claims relationship paint a more un-certain picture. A 2002 survey found that riskmanagers at a nationally representative sampleof hospitals were divided in their behefs aboutthe impact of disclosure on malpractice risk: 37percent believed that it would increase risk, 33percent beheved that it would decrease risk,and 25 percent thought that risk would notchange.^' Kathleen Mazor and colleagues sur-veyed a group of health plan enroUees abouttheir propensity to seek legal advice if harmswere not disclosed and found that failure todisclose had no statistically significant effectin three of four injury scenarios presented.'" Insum, the empirical evidence that disclosurewill reduce htigation is weak.

Erroneous extrapolation from related literature.Although few empirical studies have exam-ined the consequences of disclosure, there is animpressive hterature on why patients sue. Thisresearch has frequently been cited in supportof the view that disclosure wiU reduce htiga-tion. Such extrapolations are problematic.

In general, studies analyzing motivationsfor litigation have done so retrospectivelythrough surveys of plaintiffs, comparisons ofcharacteristics of sued and nonsued physi-cians, claims file review, or combinations ofthese approaches.^' Several studies have usedvignettes to probe key considerations in pa-tients' decisions about whether to sue." A con-sistent finding in this research is that prob-lems in the patient-physician relationship,particularly breakdowns in communication,influence htigation decisions. Several studieshave even identified the quahty of explana-tions given after injuries as a motivator."

This research estabhshes that a mix of fac-tors motivates htigation decisions. To infer.

HEALTH AFFAIRS - Volume 26, Number I 223


however, that changes to any one factor, suchas disclosure, would alone alter the claimdecision is questionable. But the more seriousflaw is extrapolation from this type of researchto conclusions about disclosure's systemwideeffects on htigation. Motivation-for-suit stud-ies are exphcitiy geared toward addressing oneside of the behavioral response: deterredclaims. An understanding of prompted claimsrequires a different research approach, onethat is focused on decision making immedi-ately after the injury and thataccounts carefully for thepossibility that without a dis-closure, the patient mightnever have known of the in-jury or its cause. This researchhas not been done.

• Study strengtiis andweaknesses. The strength ofour Monte Carlo approach toexamining the impact of disclosure on htiga-tion is that it can deal with multiple uncertain-ties simultaneously and incorporate consider-ation of "reasonable ranges" around thoseuncertainties. Weaknesses include the theo-retical nature of the exercise and the fact thatthe results must be expressed as probabihtydistributions, not simple, easily interpretableestimates. However, the approach is appropti-ate for modeling the impact of disclosure onhtigation today: There is keen interest in thenature of the effect, but it is too early to test itempirically.

Other aspects of our methodology have lim-itations. First, the analysis was confined to se-vere injuries. The impact of disclosure on hti-gation decisions following temporary andminor injuries might differ. For example, theseverity threshold that plaintiffs' attorneyshave for taking cases might drop in a routine-disclosure environment if the disclosure de-creases the effort and expense needed to inves-tigate claims. The expected effect of such achange on our estimates would be to increaseclaim volume and shghdy decrease the averagevalue of paid claims.

Second, our analyses focused on compensa-tion costs. Administrative costs of malpractice

"The spread ofdisclosure through

health care systemsis likely to amplify

malpracticeiitigation."

htigation—which include expenses associatedwith lawyers, experts, courts, and habihty in-surers—are substantial but were not countedUnder any plausible scenario, their inclusionwould have increased the size of the cost in-creases we projected.'"*

Third, our sample of experts represented avariety of backgrounds and perspectives, but itwas not drawn randomly or designed to begeographically or professionally representa-tive. Fourth, instead of detailing the content of

the disclosure, the survey re-ferred respondents to a "typi-cal disclosure situation inyour institution or experi-ence." If the respondent's con-ception of a disclosure didnot include apologies, offersof compensation, or someother element that reducespropensity to litigate, and

disclosures elsewhere or in the future consis-tently encompass such elements, then themodel might have underestimated deterredclaims. Finally, experts might have overesti-mated new claimants' abihty to secure legalcounsel—the final step in the transition toprompted claims.

• Implications. The spread of disclosurethrough health care systems is likely to amplifymalpractice litigation. We beheve that thepressing question is not whether an expansionwill occur, but how large it wiU be. Laws thatprohibit admission of disclosures into evi-dence will do htde to alter the outcome; dis-closure's primary impact wiU stem from theflagging function it serves for plaintiffs andtheir attorneys.

Two aspects of the way in which disclosureis executed could upset our general conclu-sion. The predictions might not hold if disclo-sure is practiced selectively. If incidents thatare likely to ttigger lawsuits, cost a lot, or both,are hidden—for example, those involvingclear-cut negligence or the most seriousharms—then the assumptions of our modelbreak down. In addition, the cost picturechanges if payments made to patients follow-ing disclosures fall well below the economic


M A R K E T W A T C

losses that patients sustain as a result of theirinjuries. Although potentially disruptive to theresults we have forecast, both of these featuresof disclosure are socially undesirable.

DISCLOSURE IS THE right thing to do;so is compensating patients who sus-tain injury as a result of substandard

care. Continuing moves toward transparencyabout medical injuries wHl expose tensionsbetween these two objectives. That severe in-juries are prevalent and that most of themnever trigger litigation are epidemiologicalfacts that have long been evident. Theaffordability of the medical malpractice sys-tem rests on this fragile foundation, and rou-tine disclosure threatens to shake it. Move-ment toward full disclosure should proceedwith a realistic expectation of the financialimplications and prudent planning to meetthem.

This work was supported by a grant from the Harvard/Kennedy School Health Care Delivery Policy Program.The authors are indebted to Tom Gallagher and MiltonWeinsteinforproviding helpful comments on an earlierdraft.

NOTES1. T.H. Gallagher et aL, "Patients' and Physicians'

Attitudes regarding the Disclosure of MedicalErrors," Journal of the American Medical Association289, no. 8 (2003): 1001-1007; J. Banja, Medical Er-rors and Medical Narcissism (Sudbury; Jones andBardett, 2005); and R. Bovbjerg, "Patient Safetyand Physician Silence," Journal of Legal Medicine 25,no. 4 (2004): 505-516.

2. Joint Commission on Accreditation ofHealthcare Organizations, "Comprehensive Ac-creditation Manual for Hospitals: Standard RI1.2.2" (Chicago: JCAHO, 2001); and NationalAcademy for State Health Policy, "State AdverseEvent Reporting Rules and Statutes," 5 Decem-ber 2005, http://wvvw.nashp.org/docdisp_page.cfm?LID=2A789909-5310-llD6-BCF000A0CC558925 (accessed 20 December 2005).

3. N. Berlinger, After Harm Medical Error and the Ethiaof Forgiveness (Baltimore: Johns Hopkins Univer-sity Press, 2005).

4. See, for example, D.N. Frenkel and C.B. Uebman,"Words That Heal," Annals oflnterml Medicine 140,no. 6 (2004): 482-483; A.W. Wu, "Handling

Hospital Errors: Is Disclosure the Best Defense?"Annals of Internal Medicine 131, no. 12 (1999): 970-972; and R. Zimmerman, "Doctors' New Tool toFight Lawsmts: Saying Tm Sony," Wall Street Jour-nal, 18 May 2004.

5. PC. Weiler et al., A Measure of Malpractice (Cam-bridge, Mass.: Harvard University Press, 1993).

6. W.L.F Felstiner, R.L. Abel, and A. Sarat, "TheEmergence and Transformation of Disputes:Naming, Blaming, Claiming," Law and Society Re-view 15, no. 3-4 (1980): 631-654; D. Hensler et al.,Compensationfor Accidental Injuries in the United States(Santa Monica, Calif.: RAND, 1991); and M.J.Saks, "Do We Really Know Anything about theBehavior of the Tort Litigation System—andWhy Not?" University of Pennsylvania Law Review140, no. 4 (1992): 1147-1292.

7 TA. Brennan et al., "Incidence of Adverse Eventsand Negligence in Hospitalized Patients: Resultsof the Harvard Medical Practice Study I," NewEngland journal of Medicine 324, no. 6 (1991): 370-376; and E.J. Thomas et al., "Incidence and Typesof Adverse Events and Negligent Care in Utahand Colorado," Medical Care 38, no. 3 (2000): 261-271.

8. Part A of our Technical Appendix provides de-tails of how the estimates shown in Exhibit 1were derived. See http://content.healthaffairs.org/cgi/content/ful]y26/l/215/DCl.

9. M. Sowka, ed, Malpmctice Ciaims: Final Compilation(Brookfield, Wis.: National Association of Insur-ance Commissioners, 1980).

10. The NYMPS and UTCOMPS involved reviews ofinpatient medical records only They nonethelessprovide reasonable systemwide estimates of theprevalence of serious injuries because, in additionto injuries that arose from hospital care, they cap-tured injuries that necessitated hospital care,wherever they occurred.

11. D.M. Studdert et al., "Claims, Errors, and Com-pensation Payments in Medical Malpractice Liti-gation," New En^and journal of Medicine 354, no. 19(2006): 2024-2033; and B. Black et al., "Stability,Not Crisis: Medical Malpractice Claim Out-comes in Texas, 1988-2002," jourml of EmpiricalLegal Studies 2, no. 2 (2005): 207-259.

12. Part B of the online Technical Appendix detailsthe cost calculation methodology; see Note 8.

13. S.S. Kraman and G. Hamm, "Risk Management:Extreme Honesty May Be the Best Policy," Annalso/Intemal Medirine 131, no. 12 (1999): 963-967

14. A compilation of the data sources used for thiscalculation appears in the online Technical Ap-pendix; see Note 8.

15. Patients' wiUingness to accept this trade-off isthe premise of so-called Early Offer programs.

HEALTH AFFAIRS - Volume 26, Number ] 225


16. Part C of the online Technical Appendix pro-vides further details of the survey scenarios andinstructions to respondents; see Note 8.

17. K.M. Thompson and J.D. Graham, "Going Be-yond the Single Number: Using ProbabilisticRisk Assessment to Improve Risk Management,"Human and Ecolcgical Risk Assessment 2, no. 4 (1996):1008-1034.

18. J.J. Kim, T.C. Wright, and S.J. Goldie, "Cost-Effectiveness of Alternative Triage Strategies forAtypical Squamous Cells of Undetermined Sig-nificance," Journal of rfie American Medical Assodation287, no. 18 (2002): 2382-2390; and P. Lurie et al.,"Financial Conflict of Interest Disclosure andVoting Patterns at Food and Drug Administra-tion Drug Advisory Committee Meetings," Jour-nal of the American Medical Assodation 295, no. 16(2006): 1921-1928.

19. C.E. Clark, "The PERT Model for the Distribu-tion of an Activity Time," Operations Research 10,no. 3 (1962): 405-406; and D.G. Malcolm et al.,"Application of a Technique for Research and De-velopment Program Evaluation," Operations Re-search 7, no. 5 (1959): 646-669.

20. Bounding the distribution in this way differsfrom the conventional approach to fitting a betadistribudon, which uses bounds of 0 and 1.

21. P. Doubilet et al, "Probabilistic Sensitivity Analy-sis using Monte Carlo Simulation: A PracticalApproach," Medical Decision Making 5, no. 2 (1985):157-177 Part D of the online Technical Appendixdescribes in greater detail how the modelworked; see Note 8.

22. R.M. Lamb et al, "Hospital Disclosure Practices:Results of a National Survey," Health Affairs 22, no.2 (2003): 73-83.

23. To arrive at the break-even point, the percentagesof prompted claims (transitions B and D) werereduced proportionally, beginning at the bestguess.

24. P.M. Danzon, Medical Malpractice: Theory, Evidence,and Public Policy (Cambridge, Mass.: Harvard Uni-versity Press, 1985); A.R. Localio et al, "Relationbetween Malpractice Claims and Adverse EventsDue to Negligence: Results of the Harvard Medi-cal Practice Study III," New Englandjoumal of Medi-dne 325, no. 4 (1991): 245-251; and D.M. Studdertet al., "Negligent Care and Malpractice ClaimingBehavior in Utah and Colorado," Medical Care 38,no. 3 (2000): 250-260;

25. Hensler, "Compensation."26. Saks, "Do We Really Know Anything?"27 Kraman and Hamm, "Risk Management"

28. A. Kachalia et aL, "Does Full Disclosure of Medi-cal Errors Affect Malpractice Liability? The JuryIs Sdll Out," Joint Commission Journal on Quality and

Safety 29, no. 10 (2003): 503-511.29. Lamb et al., "Hospital Disclosure Pracdces."30. K.M. Mazor et al., "Health Plan Members' Views

about Disclosure of Medical Errors," Annals of In-ternal Medidne 140, no. 6 (2004): 409-418.

31. See, for example, G.B. Hickson et al., "FactorsThat Prompted Families to File Medical Mal-pracdce Claims following Perinatal Injuries,"Journal of the AmericanMedical Assodation 267, no. 10(1992): 1359-1363; C. Vincent, M. Young, and A.Phillips, "Why Do People Sue Doctors? A Studyof Padents and Reladves Taking Legal Acdon,"Lancet 343, no. 8913 (1994): 1609-1613; W.Levinson et al., "Physician-Padent Communica-don: The Reladonship with Malpracdce Claimsamong Primary Care Physicians and Surgeons,"Journal of the American Medical Assodation 277, no. 7(1997): 553-559; and H.B. Beckman et al., "TheDoctor-Padent Reladonship and Malpracdce:Lessons from Plaintiff Deposidons," Archives of In-ternal Medidne 154, no. 12 (1994): 1365-1370.

32. A.B. Witman, D.M. Park, and S.B. Hardin, "HowDo Padents Want Physicians to Handle Mis-takes? A Survey of Internal Medicine Padents inan Academic Setting," Archives of Internal Medidne156, no. 22 (1996): 2565-2569; and PJ. Moore,N.E. Adler, and PA. Robertson, "Medical Mal-pracdce: Tlie Effect of Doctor-Padent Reladonson Medical Padent Percepdons and MalpracdceIntendons," Western Journal of Medidne 173, no. 4(2000): 244-250.

33. Witman et al., "How Do Padents Want Physi-cians to Handle Mistakes?"; and Vincent et al.,"Why Do People Sue Doctors?"

34. Part E of the online Technical Appendix explainswhy this is the case; see Note 8.


marketwatch - atul gawande€¦ · of malpractice claims. the pool of medical in-juries (stage 1)...

Documents