Indiana Law Review


Judicial Opinions Involving Health
Insurance Coverage: Trompe L'Oeil or

Window on the World?

William M. Sage*

Introduction

In contrast to a few areas of health care law with strong traditions of
research, such as antitrust and medical malpractice,' the contractual relationship

between health insurer and insured has remained relatively untested empirically.

There has been considerable research on access to health insurance and its cost,

but relatively little on how insurance operates for those who have it. Studies
which have been performed tend to focus on one important subset of coverage
issues—disputes regarding the "medical necessity" of treatment or its
"experimental" or "investigational" character—and apply an even narrower
method: explaining the legal system by examining the written opinions of courts

in cases they have decided.

A recent example of this genre is an ambitious, methodologically
sophisticated, two-year investigation headed by Mark A. Hall and Gerard F.
Anderson, and funded by the Federal Agency for Health Care Policy and
Research (the "Hall study").^ The Hall study was designed to test several
hypotheses regarding judicial treatment of coverage decisions.^ These included

the effect on judicial outcomes of (i) the method of technology assessment
employed by insurers, (ii) the severity of the patient-plaintiffs illness, (iii) the
contractual language used in the policy, (iv) the presence of procedural

protections, (v) the substantive and procedural barriers to recovery under the

Employee Retirement Income Security Act of 1974 (ERISA),"* and (vi) the
changing perceptions of cost constraints in the health care system.

The authors ofthe Hall study performed multivariate analysis of data derived
from 203 published opinions between 1960 and 1994.^ They found that the
following factors were significantly associated (p<.05/ with patients prevailing

in coverage disputes: not being in federal appeals court, the insurance contract

not expressly reserving interpretive discretion to the insurer, and seeking

* Associate Professor of Law, Columbia University. The author thanks Max Mehlman,
Karen Jordan, Mark Hall and Karen Rotschafer Sage for comments and advice, and David

Anderson for research assistance.

1. See Stephen Zuckerman et al.. Information on Malpractice: A Review ofEmpirical

Research on Major Policy Issues, Law & CONTEMP. Probs., Spring 1986, at 85.
2. See Mark A. Hall et al.. Judicial Protection ofManaged Care Consumers: An Empirical

Study ofInsurance Coverage Disputes, 26 Seton Hall L. Rev. 1055 (1996).

3. Grant proposal on file with author.

4. 29 U.S.C. §§ 1001-1461 (1994 & Supp. I 1995).
5. Hall et al., supra note 2, at 1058.

6. A p-value of under .05 means that, statistically speaking, there is less than a one-in-
twenty likelihood that the observed association was due merely to chance.



50 INDIANA LAW REVIEW [Vol. 3 1 :49

treatment for a life-threatening clinical condition.^ The authors were themselves
surprised to report that the presence or absence of ERISA was not a significant
independent predictor of outcome.*

The crucial caveat for this type of research is less whether it yields valid
answers but whether it asks the right questions. Judicial decisions are visible, but

are they important? That is, do they indicate how well our largely private,
pluralistic system of health insurance accomplishes its public policy goals? If

not, research on judicial decisions risks the absurdity of looking for the lost coin

under the lamppost solely because the light is better.

There is also an inherent paradox about empirical research involving

outcomes of litigation. By the act of engaging in it, we are perhaps admitting to
ourselves that the legal system is dysfunctional.' Research on decided cases asks

why some parties "win" and others "lose." This is important information not
only to actual or potential litigants, but to policymakers concerned with the

efficiency of the courts as a forum for resolving disputes.

In our common law system, however, the results of litigated cases are
supposed to create or confirm law. If coverage decisions generated clear, binding

precedent, there would be little to analyze statistically (and few valid objections
to the results of such analysis), since the cases would state the law. An important
realization is, therefore, that we only "study" decisions empirically because the
cases do not state the law.^° In other words, empirical research on judicial

decisions looks for subtexts where the text is unreadable.

This essay offers a few thoughts about using judicial decisions as the dataset

for research into health insurance coverage. Part I offers a general overview of

insurance coverage law. Part II considers why students of health insurance
coverage gravitate toward studying published opinions. Part III then discusses

what is wrong with the approach, and suggests alternatives. Finally, Part IV
turns to what may be right with the approach, concluding that judicial opinions
in coverage litigation may reveal the functionality (or dysfunctionality) of the
coverage process in managed care. Although the basic critique which the essay
presents applies to areas other than litigation involving medical necessity or

experimental treatment, it offers special insights into issues like health insurance

coverage where legal doctrine and public policy may not be congruent.

7. Hall et al., supra note 2, at 1067.

8. Id.

9. As discussed below, this is one reason why medical malpractice has been an intuitively

appropriate subject for empirical research.

10. The grant proposal for the Hall study, see supra note 2, anticipated that most of the

decisions considered would be from appellate courts, and concluded that this would be

advantageous because such decisions were more likely to create binding legal precedent. At the

same time, the study's application of statistical methods of analysis to opinions implicitly

recognized that little binding precedent was being created.



1998] JUDICIAL OPINIONS IN HEALTH CARE 51

I. Why Coverage Law Matters

Insurance coverage, long a backwater ofhealth law, has come to the forefront
in recent years for very good reasons." First, more medical treatments are
available, many ofwhich may be exceedingly expensive. Marginally beneficial
treatments are now frequently recommended by mainstream members of the
academic medical community, where they once were the province of fringe

practitioners or outright charlatans. As a result, disputes are less often about
naturopathy or Laetrile, and more often about chemotherapy, immunotherapy,
bone marrow reconstitution and organ transplantation for life-threatening illness.

Second, and relatedly, cost has become of significantly greater concern to
sponsors of health insurance, notably employers and government. Not only has
this made insurers more likely to challenge proposed therapies, but it has led to
the development of a variety of prospective methods to control

expenditures—^techniques which fall under the general rubric of"managed care."
Finally, the nature of insurance in health care is ambiguous.'^ On one hand,

it can be viewed as the efficient diversification of unsystematic but similar risks.

On the other hand, it can be seen as a process of social pooling, and hence
redistribution. Regulatory interventions in health care sometimes follow the

former paradigm, sometimes the latter.

Coverage litigation has therefore become one of the American health
system's Crimeas: a designated battleground for opposing values. On one side
are arrayed individual patients with idiosyncratic needs, and the physicians and

hospitals who stand ready to serve them. On the other side can be found
employers, insurers and government—in each case claiming to represent the
interests ofbeneficiaries or taxpayers as a whole by denying reliefto one member
ofthe group. This is, of course, the core challenge of managed care: creating an
efficient system of population-based health management which nonetheless
accounts equitably for the interests of individuals.

Litigation resulting from opportunistic behavior by insurers and 20-20

hindsight by beneficiaries is not unique to health care, but affects the insurance

industry broadly. Unsurprisingly, courts considering individual controversies

arising under blanket policies have occasionally strayed from clear doctrine.

Professor Jeffrey Stempel lists the common elements of disputes over insurance
coverage language which "tend to bring results less doctrinaire and consistenf

than in other areas of law: standard form contracting, unequal bargaining power,

non-negotiated terms, ambiguity, and recurring equitable considerations.*^

1 1

.

See generally Mark A. Hall & Gerard F. Anderson, Health Insurers ' Assessment of
Medical Necessity, 140 U. Pa. L. Rev. 1637 (1992).

1 2. Deborah A. Stone, The Strugglefor the Soul ofHealth Insurance, in THE POLITICS OF

Health Care Reform: Lessons from the Past, Prospects for the Future 26 (James A.

Morone & Gary S. Belkin eds., Duke Univ. Press 1994).
13. Jeffrey W. Stempel, Interpretation of Insurance Contracts: Law anc

Strategy for Insurers AND Policyholders § 2.1 (Brown & Little 1994) (quoted in Peter Nash
Swisher, Judicial Interpretations of Insurance Contract Disputes: Toward a Realistic Middle



52 INDIANA LAW REVIEW [Vol. 3 1 :49

Managed care has intensified the problems facing courts asked to determine
coverage. These derive principally from three sources: the fact that health

insurance in the United States is governed by disparate bodies of law, the
convergence of coverage and care in prepaid systems, and the increasing risk of

conflicts of interest affecting payers and providers.

A. Same Problem, Different Rights

Well-insured patients with identical medical conditions seeking equivalent

treatments are far from equals in a court of law. The relatively small group of
individual policyholders and employees of state and local governments with

private insurance are protected by state insurance regulations and have available

a panoply of legal claims and remedies under state law if coverage disputes arise.

The much larger group of persons insured through their workplace under ERISA
have a limited set of rights and remedies, and the subset whose employers self-
insure are denied the benefit of state regulatory intervention as well.^'*

Beneficiaries of government programs such as Medicare and Medicaid are

subject to procedural restrictions on judicial review in addition to slightly

different substantive standards.'^ Federal workers receiving coverage through the

Federal Employees Health Benefit Program (FEHBP),*^ active duty military

under the Civilian Health and Medical Program of Uniformed Services

(CHAMPUS),^^ and veterans covered by the Veterans' Administration health
system are also treated somewhat differently than privately insured individuals.'^

Moreover, these financing systems increasingly deliver services through the

same corporate managed care entities. Consequently, an individually insured
person, a worker whose employer purchases insurance, an employee of a self-
insured firm, and a retiree who has enrolled in the Medicare product offered by
her insurer may think they are dealing with the same "health plan." In fact, they
may be subject to widely disparate rules. Because managed care organizations
operate under readily apparent cost constraints, this different legal treatment of

similarly situated individuals tends to offend notions of basic fairness.

B. Convergence ofCoverage and Care

Coverage litigation prior to managed care was perceived as tangential to
health care delivery. Disputes generally arose long after treatment had been

rendered, and focused on payment rather than survival. Not only did this allow

Ground Approach, 57 OHIO ST. L.J. 543, 560 ( 1 996)).

14. Pilot Life Ins. Co. v. Dedeaux, 481 U.S. 41 (1987); see also Margaret G. Farrell, ERISA

Preemption and Regulation ofManaged Health Care: The Casefor Managed Federalism, 23 Am.

J.L.& MED. 251(1997).

15. See Eleanor D. Kinney, The Medicare Appeals System for Coverage and Payment

Disputes: Achieving Fairness in a Time ofConstraint, 1 ADMIN. L.J. 1 (1987).

16. 6 U.S.C. §§ 8901-8914 (1994 & Supp. I 1995).
17. 10 U.S.C. §§ 1071-1 106 (1994).

18. See 44 AM. JUR. 2d Insurance § 2057 (1982).



1 998] JUDICIAL OPINIONS IN HEALTH CARE 53

reviewing courts some detachment from compelling human equities, but it meant
that health care policymakers did not connect coverage litigation with broader

themes of access, cost and quality.

As the cost of medical care has risen, and as managed care organizations
have combined financing with delivery of services, the relationship between

coverage denials and inability to receive care has become clearer. In particular,
precertification requirements for hospitalization and surgery have increased the

urgency of resolving disputes, as well as raising the stakes for plaintiffs. Similar

considerations apply when patients seek access to specialty services not readily
available within a managed care network. At the same time, insurance
organizations with selective physician networks, drug formularies, and strict

utilization review standards are more likely to be implicated in patient injury
resulting from coverage denied or improperly provided.'^

C. Conflicts ofInterest

By incorporating financial incentives for cost containment into provider
contracts or otherwise promoting physician compliance with organizational

goals, managed care has realigned the traditional parties to coverage disputes.
For example, the recommendations of the physicians affiliated with the health

plan and other providers consulted by the patient may differ. At the extreme,
patients in managed care systems not only may be unsure of their insurer's
financial obligations, they may be unaware that potentially beneficial treatment
exists.

Informational asymmetries between managed care organizations and
individual patients relating to coverage have taken on correspondingly greater

importance. Unethical marketing practices and other potential opportunities for

insurers to deceive beneficiaries have always influenced courts. Conventional

insurance, however, presented limited opportunities for mischief. By contrast,
coverage law in the era of managed care must monitor a broad array of
intermediaries, including not only traditional insurance agents but also employee

benefits personnel, claims administrators and utilization review entities, plus

affiliated providers and their contracting vehicles.

Taking these factors in combination, coverage disputes are increasingly about

providing fair and uniform access to medical treatment from competent agents

making full disclosure. The world of insurance coverage law thus begins to
resemble the more familiar arena of medical malpractice. In the discussion that
follows, we will come back to this comparison in connection with the pros and
cons of, and alternatives to, studying coverage law through reported cases.

II. Why We Study Coverage Law Using Reported Decisions

At first glance, studying coverage law from judicial decisions strikes one as

19. See, e.g., Wilson v. Blue Cross, 271 Cal. Rptr. 876 (Cal. Ct App. 1990). But see

Corcoran v. United Healthcare, Inc., 965 F.2d 1321, 1331 (5th Cir. 1992) (ERISA preempts

available remedies for "medical decisions incident to benefit determinations").



54 INDIANA LAW REVIEW [Vol. 3 1 :49

sensible. Contractual provisions designed to apply across-the-board seem
susceptible to legal interpretation apart from the facts of particular cases. For

example, in the case of clauses excluding coverage of "investigational" or

"experimental" treatment, the therapy, not the patient, is the apparent focus of

inquiry. Therefore, judicial interpretation of insurance provisions could have

systematic effects.
^°

These expectations are not borne out in practice. Despite the fact that
insurance is an aggregate endeavor, and the policy language interpreted by courts

affects many people simultaneously, reported cases generally reflect unique
needs and circumstances. As a result, coverage cases are seldom brought in class
action form. Moreover, although a ruling overturning an exclusion for

experimental treatment could theoretically change the policy for all patients

requesting the same treatment, this does not seem to occur.^^ Why, then, are
reported opinions an attractive database for empirical research?

A. Judicial Decisions are Abundant

A major reason we tend to examine judicial opinions in coverage cases is that
there seem to be a lot of them.^^ However, the apparent abundance of formal
judicial findings is largely artifactual.

Most personal injury cases are tried before juries.^^ However, compared with
medical malpractice disputes, there are relatively few jury verdicts in coverage

disputes, a phenomenon probably explainable by ERISA. Though the percentage

has declined in recent years, well over half of privately insured patients receive

coverage through employment.^"* Coverage claims involving employer-sponsored

20. By contrast, medical malpractice cases tend to depend on highly idiosyncratic facts and,

as discussed further below, are generally tried before juries whose reasoning is not revealed in

written opinions. Some empirical work in malpractice has therefore explored the factors that cause

plaintiffs to win or lose in jury verdicts and settlements. Moreover, many empirical studies of

malpractice have focused on non-judicial measures of system performance, assessing global cost

and efficiency from data such as malpractice insurance premiums, defensive medicine, correlation

between negligent injury and litigation, and adequacy of compensation. See infra note 1 14.

21

.

In addition, a defendant insurer could be collaterally estopped from challenging a prior

adverse determination regarding the meaning or legitimacy of a contractual exclusion. See

generally Parklane Hosiery Co. v. Shore, 439 U.S. 322 (1979). See also Clements v. Airport Auth.,

69 F.3d 321, 330 (9th Cir. 1995); Texas Employers' Ins. Ass'n v. Jackson, 862 F.2d 491, 500 (5th

Cir. 1988); 18 JAMES William Moore, Moore's Federal Practice § 132.01 (3d ed. 1997).

Nonetheless, courts are reluctant to extend rulings from one case to another except through the

more limited application of stare decisis. Consequently, although insurers may redraft contractual

provisions in response to judicial decisions, they are seldom compelled to.

22. This seeming abundance may be deceptive when samples are subjected to rigorous

statistical analysis. See infra Part III.A.

23. Kevin M. Clermont & Theodore Eisenberg, Trial by Jury or Judge: Transcending
Empiricism, 11 CORNELL L. REV. 1 124, 1 136-37 (1992).

24. Mark A. Hall, Rationing Health Care at the Bedside, 69 N.Y.U. L. REV. 693, 779



1 998] JUDICIAL OPINIONS IN HEALTH CARE 55

health plans must be brought under federal ERISA law, which broadly preempts
related state law claims.^^ Although the Seventh Amendment to the U.S.
Constitution guarantees the right to a federal jury trial upon request of a party in
"suits at common law,"^^ disputes involving equitable remedies may be
conducted as bench trials.^^ Claims for benefits under ERISA were traditionally
regarded as equitable because federal pension law incorporates large portions of

the lawoftrusts.^^

As a result, several federal appellate courts have denied the right to a jury
trial in ERISA cases, although there are indications this may change with respect
to claims for damages as opposed to injunctive relief.^^ Bench trials obligate the

judge to issue a Memorandum of Findings of Fact and Opinions of Law, which
may be published.^° By contrast, a jury verdict in a medical malpractice case
creates little official record, unless detailed rulings are issued on post-trial

motions.

Whether or not brought under ERISA, suits in preauthorization cases
requesting preliminary injunctions requiring insurers to pay for or provide

treatment are decided by judges. In addition, a subset of coverage disputes arises

under Medicare and Medicaid, both historically fee-for-service insurance

programs. Challenges to Medicare or Medicaid benefit determinations are

channeled through administrative adjudicatory mechanisms.^' Because

beneficiaries have a right to judicial review (although not to a jury trial), many
ofthese cases result in published opinions. Moreover, settlement opportunities

(1994).

25. 29 U.S.C. §§ 1 144(a) & 1003(a) (1994).
26. U.S. Const, amend. VII.

27. See Katchen v. Landy, 382 U.S. 323, 337 (1966) (citing Barton v. Barbour, 104 U.S.

126, 133-34(1891)).

28. See Coar v. Kazimar, 990 F.2d 1413, 1418 (3d Cir. 1993).

29. See, e.g., Katsaros v. Cody, 744 F.2d 270 (2d Cir. 1984). However, the Supreme Court

recently suggested that the nature of the desired remedy, not the underlying claim, may determine

the right to a jury trial. Mertens v. Hewitt Assoc, 508 U.S. 248 (1993) (denying claim for damages

as not within category of "appropriate equitable relief). Following Mertens, several federal district

courts have required jury trials for benefits claims requesting damages. See, e.g., Mullins v. Pfizer,

Inc., 889 F. Supp. 69, 76 (D. Conn. 1995); Hulcher v. United Behavioral Sys., Inc., 919 F. Supp.

879, 885 (E.D. Va. 1995); Algie v. RCA Global Communications, Inc., 891 F. Supp. 870, 875
(S.D.N.Y. 1994); Sullivan v. LTV Aerospace & Defense Co., 850 F. Supp. 202 (W.D.N.Y. 1994),
vacated in part by 82 F.3d 1251 (2d Cir. 1996).

30. According to the Bench Bookfor U.S. District Court Judges, which sets forth guidelines

for when trial courts are required to issue Findings of Fact and Conclusions of Law in civil cases

and motions, bench trials must result in written findings, as must granting or refusing interlocutory

injunctions (e.g., preliminary injunctions requiring coverage). A written opinion generally must
also accompany a grant or denial of summary judgment. FEDERAL JUDICIAL CENTER, BENCH BOOK

FOR U.S. District Court Judges, § 6.02 (4th ed. 1996).

3 1

.

See Eleanor D. Kinney, Procedural Protectionsfor Patients in Capitated Health Plans,

22 Am. J.L. & Med. 301, 314-18 (1996).



56 INDIANA LAW REVIEW [Vol. 31:49

are rare in public programs.

Motion practice also exaggerates the frequency ofjudicial intervention in
coverage cases. Even if a jury might theoretically be impanelled for an ERISA
benefits trial, courts are limited to an abuse of discretion review in situations in

which the ERISA plan document expressly reserves discretion to the plan
administrator.^^ Defendants therefore typically bring motions for summary
judgment, which often generate written rulings. Additionally, some jury
decisions in non-ERISA cases brought under state law may result in the
assessment of punitive damages (which are rare in malpractice cases),
occasioning post-trial motions to remit damages which must be decided as a
matter of law by the trial judge.

^^

Apart from identifying parties as plaintiffs, defendants, appellants or

appellees, the Hall study does not detail the procedural posture of the cases it

reviewed. Nonetheless, the distribution of cases between trial and appellate

levels in the Hall study is suggestive. State supreme court decisions accounted

for 9% of the opinions reviewed, state appeals courts 26%, state trial courts 3%,
federal appeals courts 22% and federal trial courts 39%.^'* The larger percentage
of federal trial court decisions may represent not only Medicare and Medicaid,
but bench trials or motions in ERISA cases. The low percentage of state trial
court opinions, on the other hand, may reflect non-ERISA cases decided by state
juries. Assuming that juries favor plaintiffs, this hypothesis is supported by the
fact that the insurer was the appellee in most federal appeals (66%) but was the
appellant in most state appeals (55%).''^

B. Contract Cases Appear Self-Contained

The contractual nature ofcoverage disputes may favor research usingjudicial
decisions. Parties to a contract form a voluntary relationship, the terms of which

are subject to judicial enforcement but little more. Compared with tort claims,
which convey intuitively a need to study broader social issues such as deterrence

of negligence and compensation for injury, contract cases seem less concerned
with factors beyond the agreement. In other words, tort analysis may predispose
to extrinsic research because public policy issues are explicit. By contrast, such
matters are implicit in contract analysis, which therefore tends to limit research

to intrinsic data such as judicial interpretation.

This is not to say that coverage language is written on a blank slate.

Insurance benefits provided by fee-for-service Medicare and Medicaid are not

32. Firestone Tire & Rubber Co. v. Bruch, 489 U.S. 101,115 (1989).
33. There is another possible reason why coverage litigation often results in a final judicial

determination. Lawsuits in coverage cases frequently represent last-ditch efforts by dying patients

and their families. Unlike malpractice cases, which center on monetary compensation for prior

injury, managed care litigation is all-or-nothing. The potential for compromise, and therefore for

settlement, may therefore be reduced.

34. Halletal., JM/?ranote2, at 1061.

35. /c/. at 1064.



1 998] JUDICIAL OPINIONS IN HEALTH CARE 57

contracts but legislative entitlements. In addition, the terms of coverage are

frequently constrained by state insurance regulation.^^ Nonetheless, strong

ERISA preemption of state regulation for self-insured employers, coupled with
the absence of federal substantive requirements, has created a period of

contractual free rein for many managed care agreements. Taking advantage of
this legislative laissez faire, managed care plans and employers have greatly
increased the range and significance of contractual limitations imposed on

beneficiaries.^^

As managed care becomes the dominant form of insurance, and coverage and
care converge, the current era of free contracting will probably come to an end.^^
Most importantly, federal legislators are beginning to amend ERISA to impose
on managed care plans substantive limitations typical of state insurance
regulation.^^ At the same time, both federal and state regulators seem more
willing to dictate medical practice in the context of managed care than was the
case in a fee-for-service environment.'*^

C Academics Understand Judicial Opinions
Quantitative and statistical work in law is a relatively recent phenomenon.

36. The Politics of Healthcare Reform: Lessons from the Past, Prospects for the

Future 51,210 (James A. Morone & Gary S. Belkin eds., 1994). Insurance regulation includes
limitations on contractual language as well as issues such as reserve requirements and mandated

benefits. This has two consequences forjudges evaluating insurance contracts. First, it reduces the

range of possibilities available to contracting parties. In addition, however, it may create situations

where compliance with regulatory requirements contradicts judicial principles of contract

interpretation. For example, a court might berate an insurer for failing to .state explicitly that a

particular treatment was excluded from coverage, despite the fact that "laundry list" exclusions are

disfavored by state regulators.

37. See, e.g., McGann v. H & H Music Co., 946 F.2d 401, 407-08 (5th Cir. 1991) (allowing
employer to adopt self-insured ERISA plan with greatly reduced benefits for AIDS).

38. One aspect of the convergence of coverage and care tending in the opposite direction

is the possibility that the professional practice standard to which physicians are held might be

specified in a managed care contract. In Dukes v. U.S. Healthcare, Inc., for example, the court left

open the possibility that a contractual standard of care would allow insurers to claim ERISA

preemption even in routine malpractice litigation. 57 F.3d 350 (3d. Cir.), cert, denied, 1 16 S. Ct.

564 (1995). It will be interesting to see if contractual standards of care persuade researchers on

medical malpractice to pay more attention to judicial opinions.

39. E.g., Health Insurance Portability and Accountability Act of 1996, Pub. L. No. 104-191,

1 10 Stat. 1936 (codified in scattered sections of 18 U.S.C, 26 U.S.C, 29 U.S.C, and 42 U.S.C);

see also Patient Access to Responsible Care Act of 1997, H.R. 1415, 105th Cong., 1st Sess.

40. Kg. , Newborns' and Mothers' Health Protection Act of 1 996, Pub, L. No. 1 04-204, 1 1

Stat. 2874, 2935-44 (codified at 29 U.S.C. § 1185, 42 U.S.C. §§ 300gg-4, 300gg-51 (1994))

(minimum hospital length of stay for postpartum care); Breast Cancer Patient Protection Act of

1997, H.R. 135, 105th Cong., 1st Sess. (minimum hospital stay for mastectomy and lymph node

dissection).



58 INDIANA LAW REVIEW [Vol. 3 1 :49

reflecting lawyers' increasing level of engagement with the world outside the
courtroom. "Law and empiricism" follows naturally from previous scholarly
movements—such as legal realism, law and society, law and economics, and
critical legal studies—^that drew upon extrinsic sources of information and
analysis to explain and inform legal doctrine. The enormous expansion of the
American health care industry, and its high degree of regulation, make empirical
work in health law especially attractive.

Judicial decisions are a natural starting point for legal empiricism. Judge

Posner finds it significant that legal scholars focus on the written opinion rather

than the courtroom drama: "[Academics tend] to ascribe more importance to the
opinion, to its reasoning, its rhetoric, etc. than to the decision itself. Yet these

are secondary factors for most judges. For the judge, as for Hamlet, 'the play's

the thing.
""*^ Moreover, Posner notes that legal academics tend to study

appellate decisions more intensively than those of trial courts, and that "opinions
are virtually [the] only public product" of appellate judges."*^ Mixing trial and

appellate cases in empirical analyses therefore raises questions. According to

Posner, the two tiers ofjudging differ significantly: trial judges are both affected

and monitored by daily interaction with the litigants, while appellate judges play

a "game" according to intellectually satisfying but more formalistic rules."*^
Published decisions therefore represent an easy extension of traditional legal

scholarship to health care. With respect to coverage litigation specifically,

ERISA's broad preemptive effect has probably encouraged this focus by reducing
the number and influence of legislative initiatives. This is likely to change as the

"ERISA vacuum" begins to fill in response to recent judicial limits on
preemption and heightened interest in federal regulation of insurance.

"^"^

D. Additional Data Are Limited

Availability of information drives the direction of research. Reported cases

are the most easily available source of information about coverage disputes,

merely an electronic search away. Some data on a broader section of coverage
disputes are also available, at least for federal cases. The Administrative Office
of the United States Courts creates a record of each civil case terminated,

including the subject matter and jurisdictional basis, the amount demanded, the

dates of filing and termination, the procedural posture of the case at termination

and, if a judgment was reached, the prevailing party and the amount awarded."*^

41. Richard A. Posner, What Do Judges and Justices Maximize? (The Same Thing
Everybody Else Does), 3 SuP. Ct. EcoN. Rev. 1, 26 (1993).

42. Id at 7.

43. /^. at 7, 29-30.

44. E.g. , Health Insurance Portability and Accountability Act of 1 996, Pub. L. No. 1 04- 1 9 1

,

1 10 Stat. 1936 (codified in scattered sections of 18 U.S.C, 26 U.S.C., 29 U.S.C, and 42 U.S.C.)

(also known as the Kassebaum-Kennedy Act; applies to self-funded ERISA plans as well as

insurance arrangements).

45. Clermont & Eisenberg, supra note 23, at 1 133.



1 998] JUDICIAL OPINIONS IN HEALTH CARE 59

Whereas lawsuits are matters ofpublic record, relatively little is known about
how insurers reach coverage decisions in non-litigated cases."*^ Unlike medical
malpractice, where many liability insurers freely share data on their insureds,
health insurers often assert proprietary interests in their coverage standards."*^

Even coverage determinations under government programs are obscure, largely
a result ofthe low visibility ofthe private carriers and intermediaries with which
Medicare contracts."*^ This may change if constitutional due process
requirements are imposed on managed care organizations serving Medicare and
Medicaid beneficiaries."^^

By contrast, medical malpractice data are far more organized and accessible.
Research on malpractice tracks malpractice insurance premiums, claims and

awards. For example, data on nationwide jury awards are available from Jury

Verdict Research, although the database depends on voluntary submissions by
the parties and excludes settlements.^° Other important sources of information

are malpractice insurers' records of filed and closed claims, and state and

federally mandated reporting of settlements and judgments involving physicians.

The new regulatory focus on managed care should expand data availability
in the coverage arena. For example, several states now require managed care
organizations to disclose to beneficiaries information on grievance and appeals

procedures for challenging coverage determinations.^^ In addition, a few states
require standardized reporting to regulators and the public of patient satisfaction

with coverage, including disenrollment statistics and the number and outcome of
benefit disputes. This information, however, is not necessarily indexed to filed

or decided litigation.

E. Political Constituencies Are Underdeveloped

Politics explains some of these data differences. Medical malpractice has

46. See, e.g., William P. Peters & Mark C. Rogers, Variation in Approval by Insurance
Companies ofCoverage for Autologous Bone Marrow Transplantation for Breast Cancer, 330

New Eng. J. Med. 473 (1994). One exception is technology assessment, which has generated a rich
literature. See, e.g.. Institute of Medicine, Committee for Evaluating Medical

Technologies in Clinical Use, Assessing Medical Technologies (1985). However, the

science of technology assessment tends to be divorced from its utility in avoiding or resolving

coverage disputes.

47. Insurers may also have an interest in preserving ignorance, since the alternatives might

be to develop more rational standards or to admit that the emperor lacks clothes.

48. Medicare Part B: Inconsistent Denial Ratesfor Medical Necessity Across Six Carriers:

Before the Subcommittee on Regulation, Business Opportunities, and Technology, Committee on

Small Business, House ofRepresentatives, reprinted in GAO/T-PEMD-94-17 (statement of Eleanor

Chelimsky, Assistant Comptroller General).

49. See, e.g., Grijalva v. Shalala, 946 F. Supp. 747 (D. Ariz. 1996).

50. Stephen Zuckerman et al., supra note 1, at 90.

5 1

.

See William M. Sage & David Anderson, Health Care Disclosure Requirements, in 1 997
Health Law Handbook 185, 189-90 (Alice G. Gosfield ed., Clark Boardman Callaghan 1997).



60 INDIANA LAW REVIEW [Vol. 3 1 :49

clearly demarcated, well organized constituencies on both sides, which has
encouraged research to support desired legislative reforms and has also provided

a political outlet for studies which have been performed.^^ Trial lawyers and
consumer groups have stressed patient protection and the identification of "bad
doctors," while physicians and hospitals have emphasized the aggregate cost and

inefficiency of litigation.^^

Regarding private insurance coverage, at least, political constituencies have

thus far had less certain turf and less mature strategies. In particular, there is as

yet no group (save employers, who are anxious to downplay their potential
conflicts of interest as sponsors of health care benefits) arguing that there is "too

much" coverage litigation or that the threat or outcome of it unacceptably
increases health care costs. Insurers have kept a low profile, relying on ERISA
and other arcane legal safeguards, rather than engaging in a debate between

statistical and identified lives which they would surely lose.^"*

Although coverage research is still a character in search of an author, there

are now several likely candidates. For example, three of the ten largest jury
verdicts of 1995 involved managed care.^^ As a result, trial lawyers who had
long ignored medical malpractice cases because of legislative caps on damages
and the need for special medical expertise are looking twice at corporate

managed care defendants and the potential for lucrative claims of "bad faith"
insurance denial or infliction of emotional distress.^^ Academic health centers,
which feel increasingly vulnerable as both explicit government funding and

private cross-subsidies for clinical research are reduced, are another important

constituency for coverage research, as well as a significant source of intellectual

horsepower. In response, the insurance industry can be expected to commission

or encourage its own studies of the cost of mandated benefits, required appeals

52. Though not the intention of its authors, for example, the most recent attempt to quantify

defensive medicine immediately attracted the attention of tort reformers both within and outside of

the medical community. Daniel Kessler & Mark McClellan, Do Doctors Practice Defensive
Medicine?, Ill Q.J. ECON. 353 (1996). Obviously, widespread public interest is both an

opportunity and a risk. Hensler notes that "[t]he highly politicized world of policy research

challenges researchers to keep their political personae separate from their research analytic

personae." Deborah R. Hensler, Researching Civil Justice: Problems and Pitfalls, LAW «&
CONTEMP. Probs., Summer 1988, at 55, 65.

53. One should note that the former is deliberately anecdotal, while the latter generally

reduces the emotional appeal of severe patient injury by making dollars rather than lives the mode

of discourse.

54. At least in medical malpractice, the counterpoint to a severely injured victim is a single

physician making an individualized judgment—not a faceless corporation dictating aggregate
policies—although this may be changing now that incidents of malpractice can be gathered together
under the perceived responsibility of managed care organizations.

55. Verdict Trends in 1995 Disregard the Clamor: Congressional Tort Reform Has Had
Little Effect, Nat'l L.J., Feb. 5, 1996, at C2.

56. As in malpractice, moreover, the plaintiffs side generally prefers the sympathetic

anecdote to the bigger picture—except, of course, to say that "it could happen to you."



1 998] JUDICIAL OPINIONS IN HEALTH CARE 6

1

process, or unrestrained litigation, as has been the case with more actively
documented areas of insurance such as workers' compensation and automobile

no-fault. Moreover, with all fifty states and the federal Congress firmly

committed to consumer protection in managed care, the legislative can of worms
(despite ERISA) has been opened wide. This should greatly increase interest in
empirical studies of insurance contracting practices.

III. Limitations OF THE Judicial Dataset

Empirical studies ofjudicial decisions suffer from significant limitations.

Some of these pitfalls apply generally, while others take on special importance
in health insurance coverage. As a result, these studies may not prove what they
set out to prove—^who prevails in coverage disputes and why.

A. Small Sample Size

Although there may be enough reported decisions in insurance coverage to
attract attention, there are far too few to draw statistically meaningful

conclusions except for very general issues. In some situations, this small
numbers problems may prevent identification of trends that would reach
statistical significance given a larger sample. In other cases, aggregating data to

achieve statistical validity obscures important local variations, such as between

courts in different jurisdictions.

In the Hall study, for example, the authors were surprised that only 203 cases

relating to medical appropriateness and otherwise meeting their criteria for

inclusion resulted in published federal and state court decisions between 1960

and 1994.^^ Because of their small sample size, the investigators were forced to

abandon their original goal of performing a longitudinal analysis of cases to

measure judicial responses to improved practices by the insurance industry or

changing perceptions of the health care system.^^ Neither were they effectively

able to study narrow but important questions identified in their original grant

proposal such as the relationship between payer and outcome or the role of

specific methods of technology assessment.^^

B. Long Time Lags

Like observational astronomy, reported cases reveal the universe as it was,

not as it is today. Courts are necessarily reactive, weighing in only in

identifiable, fully developed controversies.^^ In the Hall study, the median time

to final disposition was 2.5 years, and many cases took much longer.^' In

57. Hall et al., supra note 2, at 1059.

58. Grant proposal on file with author.

59. Grant proposal on file with author.

60. Sheldon Goldn4an & Austin Sarat, American Court Systems: Readings in
Judicial Process and Behavior 8 (S. Goldman & A. Sarat eds., 1978).

61. Hall et al., supra note 2, at 1060.



62 INDIANA LAW REVIEW [Vol. 3 1 :49

addition, sample size constraints required pooling of cases that would have been
"old" no matter how quickly they had been resolved.^^

The time delays inherent in judicial decision making create special problems
for studies of industries like managed care which are in rapid transition.
Examining judicial decisions in coverage cases may therefore suffer from
irrelevance. For example, only six of the 203 cases studied by Hall and
Anderson involved HMOs or other managed care plans.^^

Given the inevitable time lag, the lack of cohesiveness to current coverage
law may even be a blessing in disguise. As discussed below, judges prefer to
articulate narrow justifications for their decisions.^ Because of their unusual

facts and compelling equities, health care coverage cases often present an

extreme example of this phenomenon. In managed care, changes in industry
practice are occurring so rapidly that judicial attempts to make sweeping law
would seldom synchronize with the state of the system at the time of the ruling.

C. Selection Bias

Just because there are a lot of coverage cases with reported opinions doesn't

mean that most cases generate reported opinions, that most disputes give rise to
litigation, that most coverage denials are disputed, or even that treatment options

which might be denied are proposed in the first place.^^ Reported decisions are

the tip of a very large iceberg. What occurs outside the purview of the courts is
probably far more significant to the average patient—and therefore to public
policymakers—^than the opinions ofjudges.

There are many reasons why a litigated case might result in a final decision.
It may indicate the failure of negotiation or an alternative mode of dispute
resolution to achieve settlement or determine rights.^^ It may reflect uncertainty
about the underlying law. It may indicate that the law, though clear, is

62. Id. at 1059-60.

63. Id. at 1056. The authors attributed this to another problem with collecting reported

decisions, selection bias from unlitigated cases, but it undoubtedly relates as well to the novelty of

many managed care arrangements.

64. See infra notes 86-96 and accompanying text.

65. In medical malpractice cases, for example, Danzon has examined the relationship

between court outcomes and settlements. She concluded that claims tried to verdict involve

atypically large dollar amounts, more uncertainty about liability, and weaker evidence for plaintiffs.

Patricia M. Danzon, Medical Malpractice: Theory, Evidence, and Public Policy 50-51

(1985). Studies of legal decisions in other areas have also had to address sample selection

problems. See, e.g., Philip D. Drake & Michael R. Vetsuypens, IPO Underpricing and Insurance
Against Legal Liability, FiN. Mgmt., March 22, 1993, at 64.

66. Gross and Syverud assert that despite liability insurance for defendants and contingent

fees for plaintiffs, trials are too expensive and risky for most parties. They conclude that "[t]he

main function of trials is not to resolve disputes but to deter other trials." Samuel R. Gross & Kent
D. Syverud, Don 't Try: Civil Jury Verdicts in a System Geared to Settlement, 44 UCLA L. REV.
1,63(1996).



1998] JUDICIAL OPINIONS IN HEALTH CARE 63

objectionable to one or both parties. It may mean that the underlying facts have
not been elucidated to the satisfaction of the litigants. Finally, it may suggest
that interest groups are actively pursuing an available avenue for legal change.

At the same time, trends in legal doctrine undoubtedly influence the way
parties behave when they make private ordering decisions.^^ This is what
Mnookin and Kornhauser refer to as "bargaining in the shadow of the law."^^ In
addition, as Mather describes, trial courts are "cumulative policy makers," with

the outcomes of earlier cases prompting or deterring additional, similar

litigation.^^ This is enhanced by the form of discourse in litigation, where parties

frequently argue similarities to or differences from previous cases in order to

obtain the desired result. Nonetheless, it is a leap of faith to conclude that cases

not yielding final opinions—most of which are not even "cases" in the technical
sense—mirror those that do.

Selection bias can significantly skew research findings. In the Hall study, for

example, 57% of the cases that had definitive outcomes were resolved for the
plaintiff.^° This may mean that coverage law favors patients over insurers.
However, it may equally signify the opposite—^that insurers are unwilling to
settle a large percentage of valid claims, perhaps because of advantages such as

ERISA's limitations on damages—or a range of intermediate positions.^' The
essential point is to recognize that decided cases are not necessarily

representative of the universe of actual and potential disputes.

Even the apparent direction of legal change may be misleading. Henderson

67. In a recent General Accounting Office survey ofHDC-ABMT for breast cancer, nine of
twelve insurers who decided to cover the procedure reported that litigation or the threat of it was

a factor in their decision, and five characterized legal concerns as among the most important reasons

for coverage. U.S. GAO, Health Insurance: Coverage of Autologous Bone Marrow
Transplantation for Breast Cancer, microformed on GA 1.13:HEHS 96-83, at 9 (GAO
Documents).

68. Robert H. Mnookin & Lewis Kornhauser, Bargaining in the Shadow ofthe Law: The
Case ofDivorce, 88 YALE L.J. 950 (1979).

69. Lynn Mather, The Fired Football Coach (Or, How Trial Courts Make Policy), in
Contemplating Courts 179 (Lee Epstein ed., 1995).

70. Hall et al., supra note 2, at 1062.

71. Priest and Klein argue that selection bias should result in an approximately 50-50

division among decided cases, which is open to misinterpretation as indicating unsettled law.

George L. Priest & Benjamin Klein, The Selection ofDisputesfor Litigation, 13 J. LEGAL STUD.
1, 5-6 (1984). Other researchers have extended this reasoning. For example, Clermont and

Eisenberg compared bench trials with jury trials, and concluded that plaintiffs in two areas, product

liability and medical malpractice, prevailed at a much higher rate before judges. They hypothesized

that when these types of personal injury cases come before judges, defendants are overly confident

and decline settlement opportunities, so that plaintiffs win a larger percentage of ultimate

judgments. Clermont & Eisenberg, supra note 23, at 1162. See also Robert H. Gertner,
Asymmetric Information, Uncertainty and Selection Bias in Litigation, 1993 U. Chi. L. SCH.

ROUNDTABLE 75 (concluding that information asymmetries can explain deviations from the 50-50

rule).



64 INDIANA LAW REVIEW [Vol. 3 1 :49

and Eisenberg point out that a change in the percentage of cases won or lost by
each side explains little; for example, plaintiffs may lose a larger fraction of
decisions over time because they are bringing more cases in response to a
favorable change in the law, while defendants are settling weaker cases more
frequently and trying only the stronger ones7^

Another important aspect of selection bias is its susceptibility to deliberate

manipulation. For example, lower-cost forms of alternative dispute resolution

such as mandatory, binding arbitration may be attractive to managed care
organizations. However, these methods exist at the sufferance of the legal
system, since submitting to binding arbitration implies a waiver of one's right of

access to the courts. Balance among outcomes is a superficial indication of
impartiality. Planning the organization's settlement strategy to produce an even

split in decided cases may convince a reviewing court or legislature that a biased
dispute resolution process is in fact fair, and may therefore discourage it from
tinkering with or overturning it.

The authors ofthe Hall study discuss case selection issues in connection with
the limited sample size. For example, they point to both long delays in resolution

and the fact that the median cost of treatment at issue was between $10,000 and
$50,000 as deterrents to litigating cases,^^ especially since ERISA generally
limits damages to the value of the benefit denied.^"* In addition, they speculate
that managed care gives rise to fewer litigated cases because many denials take
place at the treating physician or supervising physician level, reducing patients'

knowledge of their options.^^

Recall that the Hall study concluded that whether a case is governed by

ERISA is not a significant predictor of outcome.^^ This finding suggests that
another important selection bias may have escaped detection. Of the cases
studied, 17% were Medicare, 13% Medicaid, 34% commercial insurance, 18%
Blue Cross, 7% self-insured, 3% Taft-Hartley and 7% FEHBP or CHAMPUS.''
The authors do not indicate what percentage of the commercial insurance and
Blue Cross cases involved insured ERISA plans. Nonetheless, given that nearly
half ofemployers self-insure,^^ the small number of self-insurance cases indicates

72. James A. Henderson, Jr. & Theodore Eisenberg, The Quiet Revolution in Products
Liability: An Empirical Study ofLegal Change, 479 UCLA L. REV. 479, 502 (1990).

73. Halletal., ^Mpranote2, at 1060.

74. 29 U.S.C. § 1 132(a)(1)(B) (1994).

75. Hall et al., supra note 2, at 1061. The Hall study recognizes other selection biases as

well. For example, the authors interpret the low win rate for patients with life-threatening

conditions as evidence that insurers are more cautious about denying coverage to these patients.

Id at 1065.

76. On the other hand, the study found that federal appellate jurisdiction and contractually
reserved discretion by the insurance plan favor defendants. Id. at 1067. These factors are closely

linked to ERISA, making it problematic to consider them independent variables.

77. Mat 1061.
78. Managed Care: HMOs, PPOs, POs Now Cover Majority ofAmericans in Employer

Plans, 24 Pens. & Benefits Rep. (BNA) 316 (1997). Alternatively, the low percentage of self-



1998] JUDICIAL OPINIONS IN HEALTH CARE 65

that ERISA plan beneficiaries are underrepresented in cases resulting injudicial
decisions.

If most ERISA cases (or potential disputes) never reach decision, ERISA
becomes a very important factor regardless of the outcome of reported opinions.

As noted previously, ERISA restricts claims and damages, seldom confers a right
to a jury trial, and limits judicial review in many instances.^^ Anecdotal evidence
exists that many complaints are not pursued if defense counsel responds to claims
as being preempted by ERISA. This strongly suggests that ERISA is a powerful
deterrent to suit, and therefore a predictor of outcome as it should inform public

policy.

D. Publication Bias

In addition to selection bias, judicial decisions suffer from reporting bias.

Medical researchers are more likely to publish studies establishing causation or
clinical benefit than ones demonstrating its absence,^° Similarly, judges publish

only a fraction of the opinions they write.^' As a result, legal reporters and on-
line databases include a preponderance of rulings containing groundbreaking

legal analysis or novel conclusions. Among other things, this tendency can lead
legal researchers to overestimate the mutability and drama of the law.

Publication bias can also affect the geographic distribution of cases, which

can change modal conclusions regarding the law. State appellate courts vary

considerably in their publishing practices. Florida, for example, publishes more

than three times as many opinions as California despite its much smaller
population.^^ Overall, state courts are generally more predisposed to publish
their holdings than federal appeals courts, which have instituted fairly uniform

controls on publication.^^

Appellate opinions are most clearly biased in favor of novelty. For example.

Rule 53 of the Seventh Circuit Court of Appeals requires published opinions

when the decision (i) establishes a new, or changes an existing rule of law; (ii)
involves an issue of continuing public interest; (iii) criticizes or questions

insured cases may reflect dilution ofthe data set by cases from earlier decades when self-insurance

was rare.

79. See supra notes 14, 23-29, and accompanying text.

80. Publication bias is widely recognized in medical research, where the intellectual (and

sometimes financial) appeal of affirmative results leads to a high frequency of false positives in the

clinical literature. See Kay Dickersin, The Existence ofPublication Bias and Risk Factorsfor Its

Occurrence, 263 JAMA 1385 (1990). This induced bias is particularly worrisome in meta-analyses
which aggregate prior studies in order to draw statistically significant conclusions. Colin B. Begg

& Jesse A. Berlin, Publication Bias and Dissemination ofClinical Research, 81 J. Nat'L Cancer
iNST. 107(1989).

81. Professor Keeton's advice to new judges is simple: "Write opinions rarely." Robert

E. Keeton, Judging 139 (1990).

82. RUGGERO J. ALDISERT, OPINION WRITING 13, 13(1 990).
83. Mat 13-26.



66 INDIANA LAW REVIEW [Vol. 3 1 :49

existing law; (iv) constitutes a significant and non-duplicative contribution to

legal literature by a historical review of law, by describing legislative history, or

by resolving or creating a conflict in the law; (v) reverses a judgment or denies
enforcement of an order where the lower court has published an opinion; or (vi)

is pursuant to an order of remand from the Supreme Court which is not merely
ministerial.*"* Federal district judges also have discretion to request publication

of their opinions in the National Reporter System, which selects cases using
similar criteria, with the notable—^and similarly bias-inducing—^addition of cases
with unique or unusual fact pattems.^^

E. Unstated Rationales

Coverage cases are notorious for results-oriented reasoning. A famous quote
from Professor Keeton states that "[j]udicial opinions [in coverage litigation] are

less than ordinarily enlightening about principled bases for decision. Often . .

.

the favorite generalization advanced by outside observers to explain a judgment

against an insurance company at variance with policy provisions is the . . .
aphorism: 'It's an insurance case."*^ Spotting ambiguities in policy language

(or creating them) is a favorite pastime of judges in coverage cases, as is

questioning the impartiality of plan administrators or allowing hindsight to color

judgment. For example, an interjesting finding ofthe Hall study was that patients
for whom the treatment in question turned out to be effective were twice as likely
to prevail in suits to recover damages for the benefit denial as patients treated

unsuccessfully. Hall interprets this as indicating that courts are influenced by the

unique attributes of cases they consider.^^

Opinions are written with many audiences in mind.^* Why a court renders

84. Id. at 15-17.

85. West Publishing Co., Publication Guide for Judges of the United States

District Courts 2-3 (1994). In addition, on-line services (Lexis and Westlaw) make available

various unpublished trial opinions. Unlike unpublished appellate opinions, these may generally be

cited as precedent in subsequent litigation.

86. Robert E. Keeton, Basic Text on Insurance Law 34 1 ( 1 97 1 ), quoted in Peter Nash
Swisher, Judicial Rationales in Insurance Law: Dusting Ojfthe Formalfor the Function, 52 OHIO

St. L.J. 1037 (1991). Not all commentators are as cynical as Keeton. Swisher, for example,

invokes a middle ground between Legal Formalism and Legal Realism to explain judicial reasoning

in property and casualty insurance coverage litigation. Id. at 1045.

87. Hall et al., supra note 2, at 1067. This was true despite the fact that, unlike medical

malpractice litigation, causation is not an element of liability in coverage suits. On the other hand,

successful treatment implies lower damages, and is often limited to injunctive relief, perhaps

making it easier for courts to justify their holdings.

88. According to Leflar, these include posterity, the bar, future judges, the legislature,

current and future law students, newspaper readers, the judge himself or herself (to be satisfied with

the decision), the parties (especially the loser), and fellow judges (to obtain a majority). Robert A.

Leflar, Some Observations Concerning Judicial Opinions, 61 COLUM. L. REV. 810, 813-14 (1961).

See also Ronald A. Cass, Judging: Norms and Incentives ofRetrospective Decision-Making, 75



1998] JUDICIAL OPINIONS IN HEALTH CARE 67

a decision and how it explains that decision may therefore differ.^^ Preserving
the legitimacy of the judicial system compels reasoning from interpretive

principles, while discouraging results-oriented declarations. As Solan notes,
there is necessarily a "gap between decision-making and rhetoric in hard cases,"

although difficult decisions emphasize "seemingly scientific and neutral

justification[s]."^ Supporting this view, surveys of appellate judges frequently

yield admissions ofconflicts between individual equities or policy considerations

and rules of law, although commentators differ as to which holds greater sway.^^

Because some stated rationales are fabrications intended to clothe otherwise
naked truth, drawing empirical conclusions from them may be hazardous.

Judges can also avoid hard decisions by retreating into procedural devices.

These include mootness, lack of ripeness, lack of adversarialness, non
justiciability, lack of standing, failure to exhaust administrative remedies,

expiration of limitations periods, or non-compliance with filing requirements.^^

The Hall study deliberately excluded cases which had been resolved on grounds
other than the appropriateness of the treatment rendered.^^ However, it is

possible that some of the excluded decisions were in fact based on judgments as
to appropriateness, but were justified on technical or procedural grounds.

Even if the basis for a decision is not concealed, it may be framed
strategically. For example, it is often prudent for a judge to issue as narrow a

ruling as possible because the potential consequences of a broader statement are

B.U. L.Rev. 941 (1995) (examining the structural influences on judges' incentives and behaviors).

89. One piece of evidence for this in coverage Htigation is the frequency with which judges

disclaim general applicability of their decisions. For example, the court in Pirozzi v. Blue Cross-

Blue Shield concluded as follows:

Worth noting here is the modest breadth of this decision. It is not a green light

signalling a general expansion of coverage under group health policies like the Plan.

Rather, this decision is narrowly, but firmly, anchored in the specific expert medical

testimony presented and in the terms and structure ofthe Plan's experimental exclusion

provision. Of course, a different experimental exclusion, or different expert testimony,

or a plan that conferred broad discretion on the administrator might well require a

different result.

741 F. Supp. 586, 594 (E.D. Va. 1990).

90. Lawrence M. Solan, The Language of Judges 1 1 , 1 77 ( 1 993).

91. Llewellyn regarded many judicial constructions as merely providing a means to an

already determined end, which he viewed as largely derived from "fireside equities." Karl

Llewellyn, The Case Law System in America 79 (Paul Gewirtz ed. & Michael Ansaldi trans.,
1989). Marvell, on the other hand, concluded that judges place more emphasis on policy

implications. Thomas B. Marvell, Appellate Courts and Lawyers: Information Gathering

IN THE Adversary System 144 (1978) ("If you can ac/zieve justice in that particular case and still

do no violence to the law, I'm willing to go along" was a representative comment.). Which set of

concerns prevails may differ between trial and appellate courts. Id. at 1 57-58.

92. Posner, supra note 4 1 , at 2 1

.

93. Halletal., .sw/7rcrnote2, at 1057.



68 INDIANA LAW REVIEW [Vol. 3 1 :49

not knowable at the time.^^ According to Posner, the distinction between holding

and dictum buttresses this practice by allowing judges to join opinions with
which they do not wholly agree while still not "mortgaging . . . future votes."^^
Similarly, judges frequently prefer to be perceived as constrained in their

discretion, and therefore write opinions which portray the court as having but a
single option.^^ These proclivities can confound empirical studies which seek to
understand the causes underlying judicial outcomes.

IV. What We Might Learn From Judicial Decisions

Despite these limitations, the study of judicial decisions has redeeming

qualities. This section describes two ways in which published opinions in
coverage cases can help us understand underlying policy issues. One approach
is to simplify the empiric inquiry from "what can decisions tell us from their

outcomes and reasoning" to "what can decisions tell us from their existence."

This avenue can yield information about disequilibrium and adversarialness in

the health care system. A second approach is to use judicial opinions to assess
the coverage system's ability to bring facts under consideration and to assure fair

process. These are essential contributors to the overall success of health

insurance, and happen to be things that courts do well.

A. Why Courts Get Involved

Even if we cannot learn as much as we might hope from the content and
outcome of reported cases, we can certainly glean information from their
incidence. One explanation of the fact that coverage decisions have attracted
attention is that the number of coverage disputes generating written opinions has
increased markedly over the last decade. The Hall study found that the number
of reported cases grew from 5 in the 1960s to 36 in the 1970s, 71 in the 1980s,
and 200 in the 1990s.^^ Determining why judicial activity is on the rise may yield
important insights into the health care system.

Courts may become active because circumstances are changing and a large
number of individuals are aggrieved by the changes. For example, although
Llewellyn regards most judicial outcomes as idiosyncratic, he admits that an

accumulation of cases favoring one side may induce a shift of the underlying
legal rule. He describes this as the result of "a newly emerging consortium of
interests pressing hard against an outdated, maladaptive legal norm."^^

94. MarYELL, supra note 9 1 , at 223-24.
95. Posner, supra note 41, at 20-21 . It is not clear how the Hall study treated dicta, or even

whether it identified them.

96. Solan, supra note 90, at 185. Posner calls this the "theory of power without

responsibility." Posner, supra note 41, at 20.

97. Hall et al., supra note 2, at 1060. The last figure was based on a linear extrapolation of

cases from 1990 to 1994. Of course, the number of reported opinions in other areas of law has also

increased during this period.

98. Llewellyn, supra note 91, at 100. One might ask whether these interests are



1 998] JUDICIAL OPINIONS IN HEALTH CARE 69

The rapidity of change in the health care system during the recent transition
to managed care is self-evident. Major factors include employer-driven cost
constraints, federal budgetary retrenchment, and the integration and

consolidation of insurance and provider organizations. Earlier transitional

periods and their effect on coverage litigation also may be identifiable. For
example, increases in reported decisions during the 1970s and 1980s may be
related to the impact of new technologies on established underwriting practices
and principles of insurance interpretation.

Another possibility is that the health care system is simply becoming more
adversarial. In this view, not just the existence of change but its direction

promotes litigation. In today's health care system, the erosion of trust produced

by the inversion of financial incentives from fee-for-service practice to managed
care happens to coincide with a general increase in the aggressiveness of medical

consumerism. However, ajudicial model of medical decision making is a radical

departure from professional traditions in health care, and may have important
implications for quality of care and patient satisfaction.

An increase injudicial decision making might also represent an alternative
to legislative change. Courts have a recognized role in public policy making.^^

In product liability law, for example, Eisenberg and Henderson speculate that tort

reformers failed in their legislative agenda, but still convinced individual judges

that reform was needed, as demonstrated by declining plaintiff success rates
through the 1980s.'°° In insurance coverage, one wonders whether concerns

about cost, or more recently about the excesses of managed care, may have
prompted judicial activism during periods of legislative inertia. For example, we
know that federal ERISA law has limited state legislative intervention. '°'

In keeping with the earlier discussion of selection and publication bias,

however, we should resist the temptation to assume that an increase in reported
cases necessarily equates with an increase in underlying disputes. Nonetheless,

because many state HMO and insurance regulators now require managed care
organizations to maintain records of complaints arid grievances, the hypothesis

should be verifiable. Moreover, describing the sources of selection, publication,

or other biases, should they exist, might be as revealing as confirming their

deliberately bringing cases to the attention of courts. Neither malpractice nor coverage has spurred

much "impact litigation," in large part because the rewards for individual litigants and their counsel

are usually sufficient to ensure frequent judicial review.

99. Mather has identified several aspects of policy making in which courts engage: agenda

setting, providing a forum for political argument, agenda building, mobilization of support or

opposition, definition of local legal norms, creation of new legal norms, political symbolism and

provision of political or legal resources. Mather, supra note 69, at 179.

1 00. Theodore Eisenberg & James A. Henderson, Jr., Inside the Quiet Revolution in Products
Liability, 39 UCLAL. Rev. 731, 751-54 (1992). The authors exclude shifts in accident trends, in
the propensity to file claims and in settlement behavior as causes of the decline.

101

.

ERJSA "supersede[s]-any and all State laws insofar as they may now or hereafter relate

to any employee benefit plan " 18 U.S.C. § 1 144 (1994).



70 INDIANA LAW REVIEW [Vol. 3 1 :49

absence.
^°^

B. How Courts Assess Non Judicial Processes

A supportable assertion about reported decisions in coverage cases is that
virtually all underwent other levels and forms of review or appeal prior to

litigation. Judicial proceedings might therefore shed light on the success or

failure of these nonjudicial processes.

A correlate of the infrequency with which legal precedent is established in
coverage litigation is the centrality of facts to the outcome of cases. Although

concern about facts is a defining feature of health insurance coverage cases, it is

also a staple of litigation in general. In the words of former U.S. Supreme Court
Associate Justice Robert H. Jackson,

It may sound paradoxical, but most contentions of law are won or lost on
the facts. The facts often incline a judge to one side or the other. A
large part of the time of conference is given to discussion of facts, to

determine under what rule of law they fall. Dissents are not usually

rooted in disagreement as to a rule of law but as to whether the facts

warrant its application.
'°^

An important lesson to be drawn from coverage decisions is that fact-finding
in modern health care is extremely difficult. For one thing, medical science is
generally complex and frequently uncertain. For another, the restructuring of

provider organizations in managed care, and the associated financial incentives,

have arguably diminished the availability and credibility of information.

Therefore, we may be able to learn from judicial decisions how information
regarding coverage and care is being shared—or withheld—in managed care
organizations. Because accurate, abundant information is central to the long-

term success of the health care system, understanding the judicial critique of the

mechanism by which information is generated and exchanged in the coverage
context could be valuable.

'°'*

A second lesson relates to procedural fairness. Resource allocation is a
critical subtext of insurance coverage litigation. Despite the contractual heritage

of health insurance, the litigants and the judicial system are fully cognizant of the

social implications of coverage determinations in terms of the cost and equitable

distribution of health care.*°^ Therapeutic health care is a difficult area for

102. These might include, for example, the tension between managed care organizations'

concern about the public relations effect of high-profile litigation and the deterrent to settlement

produced by limitations on damages under ERISA.

1 03

.

MarYELL, supra note 9 1 , at 1 39.
104. There is an important relationship between disclosure in the context of coverage for

experimental treatment and medical informed consent. See Nancy M.P. King, Experimental

Treatment: Oxymoron or Aspiration, HASTINGS CENTER REP., July-Aug. 1995, at 6.

105. This sets coverage cases apart from medical malpractice cases. Except for concerns

about the cost and efficiency of litigation as a method of dispute resolution, medical malpractice



1 998] JUDICIAL OPINIONS IN HEALTH CARE 7

1

regulation in large part because lives seem more "identified" than "statistical."
In such situations, a finding in favor of coverage allows the question of the

marginal value of life to be neatly avoided.
*°^

Fairness is a prerequisite to resource allocation, and legal process is the

principal guardian of fairness in democratic society. Another important reason

to look at judicial decisions as a benchmark for the health care system is
therefore that courts are well equipped to evaluate procedural fairness. '°^ Judges

in coverage cases are suspicious of decisions rendered without due process, and

respond favorably to adequate procedural protections for patients and

policyholders. *°^ Judges' thresholds for procedural fairness are especially high

in cases involving preauthorization of services for severe disease, with denial of

coverage unlikely unless due process has been scrupulously observed.

We can therefore learn from judicial decisions how well private processes
are operating, notably the manner in which coverage standards are developed and
the conduct of individual inquiries and appeals. The increase in managed care
enrollment by Medicare and Medicaid patients will add to courts' involvement
in procedural review, '^^ because government programs are subject to more
extensive due process requirements than are private parties.

^'°

Conclusion: Charting a Research Agenda

A systematic way to approach empirical research is to identify the policy
implications ofcoverage standards and formulate testable hypotheses. In medical

malpractice, for example, the objectives of the tort regime are generally

is not generally viewed as an issue of resource constraints. Either a treatment was delivered in

accordance with the professional standard of care or it was not, and cases are independent of one

another.

1 06. Coverage litigation highlights the distinction between statistical lives and identified lives

in our approach to valuing risks. A significant subset of reported coverage decisions involve high-
cost therapies for life-threatening conditions. When risk estimated ex ante is converted into loss
incurred ex post, and the loss involves human life, it is easy to second guess the earlier valuation.

107. Keeton observes that "good" judges are good because they are skilled at making hard

decisions, not because their reasoning is always a model of logic. Keeton, supra note 86, at 2.

108. The Hall study did not identify a statistically significant correlation between internal

process and case outcome, possibly because of small sample size. See Hall et al., supra note 2, at

1065-66. The authors have reported elsewhere, however, that courts' objections to internal

technology assessment by insurers tended to focus on insufficient or poorly matched sources of

information, lack of expert review, concern about financially motivated bias, and lack of current

information about clinical benefit. Mark A. Hall et al.. When Courts Review Clinical Practice

Guidelines, MED. CARE (forthcoming 1998).

109. See Grijalva v. Shalala, 946 F. Supp. 747 (D. Ariz. 1996).

1 10. See Shelley v. Kraemer, 334 U.S. 1, 13 (1948) (The Fourteenth Amendment "erects no
shield against merely private conduct."). However, when a private party's conduct has sufficiently

received the imprimatur of the State, it may be deemed state action for purposes of the Fourteenth
Amendment. See, e.g., Flagg Bros. v. Brooks, 436 U.S. 149 (1978).



72 INDIANA LAW REVIEW [Vol. 31:49

characterized as victim compensation and injury reduction. Health insurance

presents a different set of policy concerns. Useful studies will assess the impact

of contractual coverage standards (and legislative interventions regarding

coverage) on measures such as administrative expenses, health care premiums,

medical innovation, patient satisfaction, and health outcomes. For example, the

Institute of Medicine estimates that insurers deny only one or two percent of
claims, while a much greater amount of care is unnecessary and diverts resources
from other areas.^^^ Because of the social as well as individual implications of

insurance, these are weighty issues.

A threshold question with analogies to medical malpractice is whether courts
are reaching efficient and accurate results.

'^^
For example, Sykes criticizes laws

generally allowing "bad faith" claims against insurers because he believes that

courts are seldom able to accurately distinguish opportunistic behavior from

genuine and reasonable disputes.^ *^ In health insurance cases, it will be

important to assess the degree of correlation or mismatch between valid claims

and coverage cases filed, and between valid claims and relief granted.'^"* At least

for cases involving "medical necessity," some objective scientific determination
should be possible. Further studies will explore the cost of grievances and

appeals, whether conducted internally to the health plan, through an independent

review organization or via the courts.

The pace and direction of research will respond to political constituencies as
well as to the interests of academics, and grantmaking bodies will undoubtedly

react to both policy and political imperatives. For example, the medical

malpractice research agenda of physicians has generally been more focused than

that of consumers or attorneys.
^^^

Consequently, more study has been devoted

to the cost of malpractice litigation (groundless claims, administrative expense

and defensive medicine), than to quality of care (the amount of substandard

111. See also Julie A. Jacob, Managed Care Denials Less Frequent Than Expected, AM.

Med. N., Dec. 15, 1997, at 5 (describing recent studies).

112. There have been simple studies reviewing contractual exclusions and assessing the

consistency oftreatment of similarly situated individuals, but none has aggregated a large amount

of data or drawn statistical rather than anecdotal conclusions. The Hall study included as an explicit

goal measuring the potential for inconsistency in judicial decisions, but was hampered by small

sample size. Hall et al., supra note 2, at 1056, 1058.

113. Alan O. Sykes, "Bad Faith " Breach ofContract by First-Party Insurers, 25 J. LEGAL

Stud. 405 (1996).

114. Cf. Troyen A. Brennan et al.. Relation Between Negligent Adverse Events and the

Outcomes of Medical-Malpractice Litigation, 335 N. Eng. J. MED. 1963 (1996) (finding that

severity of disability, not occurrence of adverse events during medical care or related negligence,

is predictive of plaintiff recovery for medical malpractice); A. Russell Localio et al.. Relation

Between Malpractice Claims and Adverse Events Due to Negligence, 325 N. Eng. J. MED. 245

(1991) (finding gross mismatch between negligent care and filing of medical malpractice claims).

115. One reason malpractice reform is so important to physicians is because they suffer large

psychic damages from litigation which are not compensated by insurance. Kessler & McClellan,
supra note 52, at 357.



1 998] JUDICIAL OPINIONS IN HEALTH CARE 73

practice and the deterrent effect of litigation) or access to compensation for

negligent injury. Political interests in health insurance will probably be

dominated by taxpayers concerned about government expenditures under

Medicare, large employers seeking to reduce benefit costs and, more likely than
not, trial lawyers hoping to exploit the vulnerability of corporate organizations

to legal claims. Recent recommendations of the President's Advisory

Commission on Consumer Protection and Quality in managed care—including
comprehensive procedures for internal and external review of coverage

denials—have already prompted advocacy-based research on their likely cost."^
Managed care organizations will sponsor technology assessments and cost-

effectiveness studies, as will academic health centers and pharmaceutical

companies. Employers and other group purchasers will demand statistical proof
from insurers and risk-bearing providers that they are receiving value for money,
and will probably be compelled under ERISA to communicate this information
to beneficiaries. Most importantly, government will mandate reporting by the
full range of regulated entities, and will make that information available to
researchers. Notably, the expansion of Medicare and Medicaid managed care
will expand federal data collection to monitor cost, access and quality in insured

systems, and to detect and deter fraud. All of this information will shed light on

coverage standards and the processes for making coverage decisions and

resolving disputes.

A caveat is that much ofthis research may not be a planned element of health
insurance regulatory design so much as a by-product of data produced for other
purposes. This is certainly true in other areas of health law. For example, the

existence of comprehensive Medicare data allowed researchers to estimate

defensive medicine by linking restrictions on medical malpractice litigation to

service utilization.
^^^

Although no one can predict exactly how the research agenda will evolve, it
is virtually certain that we will witness an extraordinary expansion of empirical
work on health insurance coverage over the next few years, much of it based on
information extending well beyond judicial opinions. The simple reason is that
the stakes—^for identifiable constituencies and for society as a whole—are higher
than ever before. Higher stakes provoke interest in promoting or resisting

change, and greater interest generates data."^ Our challenge is to interpret those
data correctly, and to apply the results responsibly.

1 16. See Gauging Quality Regulation 's Impact on Premium Costs, MED. & HEALTH, Nov.
24, 1997, at 1; see also Consumer Bill of Rights and Responsibilities (visited Nov. 1997)

<http://hcqualitycommission.gov>.

1 1 7. Kessler & McClellan, supra note 52.
118. Of course, there may be political issues, such as abortion, in which the stakes are too

high and positions too polarized to admit research.