Indiana Law Review Empirical Studies of Judicial Decisions serve an important role in the cumulative Process OF Policy Making Comments on a Paper by Professor William Sage Karen A. Jordan* Introduction Professor Sage has presented a cogent analysis of the role that empirical studies ofjudicial opinions can play in the shaping of health law and policy.' For those of us who have an interest in empirical research, but who are not as well versed as we would like to be in statistics or the basics of social science research, the most beneficial aspect of Professor Sage's article is the section detailing the limitations associated with using judicial decisions as an empirical data set? My comments are largely directed to that aspect of his article. Overall, I agree with Professor Sage's conclusion that, despite the limitations of using decisions as a data set, empirical studies of judicial decisions can yield useful information. However, this commentary stresses that the limitations of studying judicial decisions are not necessarily as critical as suggested, and that empirical findings of such studies can play an important role in the cumulative process of policy making. I. Some Fundamentals of Empirical Research To assess the seriousness of the limitations associated with the empirical study of judicial decisions, it is useful to first review some fundamentals of empirical research designed to facilitate social policy.^ The function of policy research is to generate information that can render policy making more effective. Broadly speaking, then, the goal of empirical research in this context is to verify propositions about some aspect ofthe relationship between the objectives of the policy and the means available to achieve those ends."* However, effective policy making is a process. For example, policy making has been explained as a series * Associate Professor of Law, Louis D. Brandeis School of Law at the University of Louisville. The author formerly served as Assistant Director of the Center for Law and Health and congratulates the Center on its Tenth Anniversary. 1. William M. Sage, Judicial Opinions Involving Health Insurance Coverage: Trompe L Veil or Window on the World?, 31 IND. L. REV. 49 (1998). 2. Id at 57-64. 3. Several texts provide a detailed review of empirical research design and implementation. See generally OTTAR HELLEVIK, INTRODUCTION TO CAUSAL ANALYSIS: EXPLORING SURVEY DATA BY Crosstabulation (1 984); Robert J. Mutchnick & Bruce L. Berg, Research Methods for THE Social Sciences: Practice & Applications (1996); Carol H. Weiss, Using Social Research in Public Policy Making ( 1 977). 4. Robert R. Mayer & Ernest Greenwood, The Design of Social Policy Research 57 (1980) (describing interaction as a "means-end" relationship). 82 INDIANA LAW REVIEW [Vol. 31:81 of stages which includes the following: the determination of goals, needs assessment, specification of objectives, design of alternative courses of action, estimation of consequences of alternative action, selection of courses of action, implementation, and evaluation of outcomes.^ To put this into context, consider the problem faced by persons who receive their health coverage through plans governed by the Employee Retirement Income Security Act of 1974 (ERISA).^ Because of the interplay between ERISA's civil enforcement provisions and ERISA's preemption provisions,^ policyholders who sustain injury as a result of a denial of coverage by the plan administrator are often unsuccessful in challenging those denials in court.^ Those interested in health law and policy may therefore determine that some modification of the law ofERISA is warranted to ensure that claimants covered through ERISA plans receive equitable treatment when forced to bring their disputes into the judicial system.^ If so, the next stage in the policy-making process is to perform a needs assessment. That is, a policy maker should seek information confirming the need for a change in policy and about the amount of change in existing law required to reach the goal.'° Descriptive research provides information relevant to this stage of the process. Descriptive research studies are designed to yield quantitative measurements of the characteristics (also referred to as properties or variables) associated with the phenomenon under study." For example, to assess the need for change and the amount of change required in the ERISA context, a study might simply measure how "the law applied" affects the outcome of the case; in other words, did the claimant win less often if the court applied ERISA than when state contract law was applied. ^^ Descriptive studies yield information such as the 5. Mat 58. 6. 29 U.S.C. §§ 1001-1461 (1994 &. Supp. I 1995). 7. ERISA's civil enforcement provisions permit certain suits to be brought against benefit plans by plan participants and beneficiaries but limit the remedies available. See 29 U.S.C. § 1 132(a) (1994). ERISA's preemption provisions provide that ERISA supersedes state laws that "relate to" ERISA plans, unless exempted as a law that regulates insurance, banking or securities. See 29 U.S.C. § 1 144(a), (b)(2)(A) (1994 & Supp. I 1995). 8. See, e.g., Tolton v. American Biodyne, Inc., 48 F.3d 937 (6th Cir. 1995) (holding preempted a claim arising out of continued denial of request for inpatient mental care which led to the patient's suicide); Corcoran v. United Healthcare, Inc., 965 F.2d 1321 (5th Cir.), cert, denied, 113 S. Ct. 812 (1992) (holding preempted a claim for wrongful death arising out of a denial of coverage for inpatient care for high risk pregnancy). 9. See, e.g., Karen A. Jordan, Travelers Insurance: New Support for the Argument to Restrain ERISA Pre-emption, 13 YALE J. ON REG. 255 (1996). 1 0. Mayer & Greenwood, supra note 4, at 1 0. 11. /c/. at54. 1 2. See Mark A. Hall et al.. Judicial Protection ofManaged Care Consumers: An Empirical Study ofInsurance Coverage Disputes, 26 Seton Hall L. Rev. 1055, 1062 (1996). The study was analyzed by Professor Sage and included this type of inquiry. 1998] EMPIRICAL STUDIES 83 size, distribution, and interrelationships between variables. ^^ Next, the policy maker should identify the means available for achieving the goal of fair outcomes. This inherently involves an assessment of which factors lead to unjust results in ERISA cases. Exploratory research methods are appropriate for this stage of the policy-making process. Exploratory research involves the careful selection of a few units, which are studied comprehensively by means of a variety of unstructured and unrefined data collection techniques.^"* The data are analyzed qualitatively, rather than quantitatively, for certain inferences about, for example, what factors amenable to judicial intervention seem to be associated with unjust results in ERISA cases. In the ERISA context, exploratory research would likely lead to the inference that factors such as an abuse of discretion standard of review, exhaustion of administrative remedies, or seeking damages beyond the benefit itself are relevant for further study. '^ However, further descriptive research would be necessary to determine the size, distribution and interrelations between these factors or variables. Additionally, research designed to assess the causal relation between each variable and the outcome would help the policy maker ascertain what change in the law ofERISA would most likely lead to more just outcomes. ^^ Empirical research ofjudicial decisions would appear to be the obvious unit of study for each ofthe described stages—^the descriptive research performed at the needs assessment stage, the exploratory studies performed to identify the relevant factors or variables, and the further descriptive studies to quantitatively assess the size, distribution, interrelations and causal connections between the relevant variables. Moreover, due to the existence of a readily accessible data set, a researcher would likely combine all ofthese steps into one research study. The question is whether the limitations associated with the use of judicial decisions as a data set would compromise such a research study. 13. Mayer & Greenwood, supra note 4, at 54-55. Studies using the descriptive method are either univariate, describing the distribution of a single variable, or multivariate, describing the simultaneous distribution of two or more variables. Id. at 55. 14. /c/. at 52-53. 15. See, e.g., Bamett v. Kaiser Found. Health Plan, Inc., 32 F.3d 413 (9th Cir. 1994) (affirming denial of coverage for a liver transplant under the abuse of discretion standard applicable to the decision of the ERISA plan administrator); Corcoran v. United Healthcare, Inc., 965 F.2d 1321 (5th Cir. 1992) (dismissing the ERISA claim because plaintiffs sought damages for emotional distress and mental anguish, remedies not prescribed by section 502(a) of ERISA). See also Hall et al., supra note 12, at 1062. 16. The most sophisticated form of research, explanatory research, is generally not used in policy research until after a change in policy has been implemented. At this point, an evaluation of outcomes helps assess whether the policy change promoted the stated policy goal. The focus of a study undertaken for this purpose is whether the change in policy (the means) has a causal relationship with the objective (the ends). Mayer & Greenwood, supra note 4, at 59-60. 84 INDIANA LAW REVIEW [Vol. 31:81 II. Assessment OF THE Limitations Professor Sage discussed limitations stemming from small sample size, time lags, and selection bias.^^ As he noted, the time lag aspect does not affect the quality ofan empirical study of decisions, but only its effectiveness, or relevance, in a rapidly changing market.'^ However, small sample size and selection bias could impact the quality of research findings. The quality of findings in empirical studies is generally assessed with respect to four criteria: their generalizability, their validity, their reliability, and their practical significance.'^ The limitations pointed out by Professor Sage relate primarily to generalizability and validity. Accordingly, these concepts are explored before assessing whether the limitations of sample size and selection bias compromise empirical studies ofjudicial decisions. "Generalizability" refers to the extent to which empirical findings can be generalized as representative of cases other than those studied.^^ The term "external validity" is also sometimes used to describe this aspect of a study.^' Generalizability is relevant when the unit selected to be studied (for example, judicial decisions) consists ofa large number ofthose units, and, because it is not possible to study all of them, a subset of the group is selected that hopefully represents the larger body. However, when the units selected for study do not correspond exactly to the target population (defined to mean the aggregation of units to which the study findings are hoped to apply),^^ generalizability is compromised.^^ Although some of the selection bias problems identified by Professor Sage as inherent in a data set comprised of judicial decisions would affect generalizability, some would not. Professor Sage pointed out that, among other things, reported decisions represent the tip of the iceberg—^they are not the average case and thus are not particularly representative of how most patients fare in their out-of-courtroom coverage disputes.^"^ However, this form of selection bias would not necessarily be detrimental to an empirical study of 1 7. Sage, supra note 1 , at 6 1 -65 . 18. Mat 62. 1 9. Mayer & Greenwood, supra note 4, at 257. 20. Id. 21. See, e.g., JOHN W. Creswell, RESEARCH DESIGN: Qualitative & Quantitative Approaches 158-59 (1994); Curtis D. Hardyck & Lewis F. Petrinovich, Understanding Research in the Social Sciences: A Practical Guide to Understanding Social and Behavioral Research 7 (1975). 22. Mayer & Greenwood, supra note 4, at 170-72. For example, consider a hypothetical ERISA study designed to determine whether, how, and to what extent the law of ERISA should be modified to ensure equitable treatment in the judicial system. The target population would include all persons covered through ERISA plans who look to the courts for some consumer protections. The researcher has access to the target population through judicial decisions. 23. The study findings can be applied only to those members of the target population which also fall into the study population. Id. at 258. 24. Sage, supra note I, at 58. 1998] EMPIRICAL STUDIES 85 judicial decisions. For example, the Hall Study was designed to assess the level of consumer protection available through judicial review of coverage disputes.^^ That is, the study focused on how courts enforce contractual entitlements to health benefits. Thus, in the Hall Study, the fact that the cases studied would not be representative of how patients fare in their out-of-court disputes would not have a negative impact on study results. The empirical findings would be generalizable to the whole target population—^those who resort to the judicial system. A second form of selection bias noted by Professor Sage similarly has a minimal effect on the generalizability of the Hall Study findings. Extrapolating from the low number ofcases involving self-insured plans. Professor Sage opined that ERISA cases may have been underrepresented and further suggested that the likelihood of preemption of the claim by ERISA is a powerful deterrent to suit.^^ Again, however, because the focus of the Hall Study was on judicial treatment of coverage disputes, this effect on the number ofjudicial decisions would not affect generalizability. Rather, the effect simply highlights the narrowness of the study. The Hall Study findings are indicative ofjudicial treatment of coverage disputes; they are not indicative of the way ERISA plan participants fare in the administrative treatment of coverage disputes. Although this aspect of selection bias highlights the narrowness of the data produced when judicial decisions are studied in empirical research, it has a minimal effect on generalizability. However, two other selection bias problems would appear to have a more substantial impact on generalizability. Professor Sage explained that publication bias exists because judges have discretion in deciding which of the opinions they write will be reported.^^ Similarly, Professor Sage pointed out that judges have discretion in deciding which of the many factors that may have influenced their decision will appear in the written opinion and that, "[b]ecause some stated rationales are fabrications intended to clothe otherwise naked truth, drawing empirical conclusions from them may be hazardous.'^^ These limitations could impact generalizability because they suggest that the findings would not accurately reflect how other patients would fare in cases taken to court. Perhaps more importantly, the limitations stemming from judicial discretion would also have an impact on the validity of a study's empirical findings. Validity (specifically, internal as opposed to external validity) concerns the accuracy of empirical findings and whether they match reality.^^ Thus, validity refers to the extent to which the study findings are applicable or relevant to the research objectives and, more specifically, to the extent to which the measures obtained reflect the variables specified in the research objectives.^^ 25. Hall et al., supra note 12, at 1056. 26. Sage, supra note 1, at 65. 27. Mat 65-66. 28. Id. at 67. 29. Creswell, 5Mpranote21,at 158. 30. Mayer & Greenwood, supra note 4, at 258. The principle limitation to validity is often the indicators selected to measure concepts intended to be studied. For example, if a study 86 INDIANA LAW REVIEW [Vol. 31:81 The problem of unstated rationales in judicial opinions would arguably impact the validity of a study of judicial opinions. For example, in our hypothetical ERISA study, one variable studied might be the use of the abuse of discretion standard ofreview. If a judge's opinion was written so that it appeared that the standard of review was key to the outcome but there were other unstated rationales, the measure of that variable would not match reality. Similarly, if the data set is underrepresentative of ERISA cases, as Professor Sage suspected in the Hall Study, this would impact the validity of the finding that ERISA does not have a statistically significant association with a claimant's outcome.^' The other major limitation identified by Professor Sage was sample size. He stated that there are far too few reported decisions pertaining to coverage disputes to draw statistically meaningful conclusions.^^ In the Hall Study, however, sample size became a problem because the study was limited to coverage disputes involving denials due to medical appropriateness, defined to include decisions turning on medical necessity, or on whether the treatment could be characterized as experimental or investigational.^^ This focus reduced the sample size by over 750 cases.^"* Thus, it is difficult to predict the extent to which small sample size may or may not be a problem in studies of other types of coverage disputes. For example, we are seeing more cases involving challenges to precertification procedures that delay care, or to denials stemming from unreasonable financial incentives to reduce care.^^ Studies of other coverage issues might well result in a sufficiently sized sample. But sample size is important. Professor Sage explained that small sample size can limit the choice of data analysis techniques.^^ For example, descriptive research studies designed to yield information such as the interrelationships between variables usually require a large number of units of study.^^ Further, size can impact validity, reliability and generalizability of empirical findings.^^ used age and level of education as indicators of a person's employability, validity would depend on degree of association between age and level of education and the length of time required to secure a job. Id. aX25^-59. 3 1 . See Hall et al., supra note 12, at 1066. 32. Sage, supra note 1 , at 6 1 . 33. See Hall et al., supra note 12, at 1057. 34. Id. at 1057-58. 35. See, e.g., Pappas v. Asbel, 675 A.2d 71 1 (Pa. Super. 1996), appeal granted, 686 A.2d 1312 (1996) (involving allegedly negligent delay in pre-authorization); Ouellette v. Christ Hosp., 942 F. Supp. 1160 (S.D. Ohio 1996) (involving an allegedly unreasonable system of financial incentives that caused premature discharge from the hospital). 36. Sage, supra note 1, at 61. 37. Mayer & Greenwood, supra note 4, at 54. 38. Reliability refers to the degree of confidence that can be accorded the research findings, generally considered to be the extent to which repeated applications of the research design under similar conditions would yield consistent findings. Id. at 259. Reliability may be hindered in two principle ways: if the study population is selected randomly, or if there is variability in the data collection techniques. However, in both instances, the degree of reliability may be measured. Id. 1998] EMPIRICAL STUDIES 87 This is because of the fundamental role that probability plays in statistical processes. Because researchers hope to prove that their findings represent more than a chance relationship between variables, probability is central to statistical significance and sampling.^^ For example, if the study population is selected randomly, there must be a sufficient number of units selected such that there is a high probability of reproducing the essential characteristics of the total population/^ However, there is not a generally recognized "requisite size" for an empirical study. Rather, a researcher generally must balance the need for reliability and validity against the costs and benefits of the study.'*' In my view, the need for balancing costs and benefits is the key to the question ofwhether judicial decisions should be used as a data set for empirical research. There are real limitations associated with an empirical study ofjudicial decisions. However, those limitations must be balanced against the benefits that can be gained from the information revealed. Indeed, some empirical researchers have concluded that validity is a concept "to be pursued, but not to be attained.'"*^ They have rejected the common view that if researchers "can acquire a sufficient amount of [validity], by applying appropriate techniques, one has somehow 'won' at the game called research."^^ Rather, they advocate that validity is to be assessed "relative to purposes and circumstances.'"*"* Thus, although there are limitations associated with studying judicial decisions, those limitations must be considered in light ofthe reasons researchers might want to study judicial decisions. This comment has already explained that empirical studies ofjudicial decisions may reveal information useful in various stages ofthe policy making process. For example, such studies have been found helpful in assessing the need for a change in policy or in formulating possible alternatives."*^ Further, Professor Sage pointed out that studies of judicial decisions can provide useful information. For example, studies can provide information about the increasing number ofcases ending up in court and possibly some insight into the reasons for the increase or information about the administrative processes which patients must exliaust prior to seeking redress in 39. Kenneth R. Hoover, The Elements of Social Scientific Thinking 97-98 (4th ed. 1988). 40. Id. at 99. 41 . Mayer & Greenwood, supra note 4, at 1 78 (noting that the policy researcher tries to increase the number of observations in order to maximize reUability, but that the importance of reliability will vary from one type of study to another). 42. David Brinberg & Joseph E. McGrath, Validity and the Research Process 1 3 (1985). 43. Id. 44. Id. (emphasis omitted). The authors further note that rather than viewing validity as a necessary aspect of an empirical study, it would be more productive to pursue robustness analyses of empirical findings, i.e., further activities designed to assess the degree of certainty surrounding the findings. Id. at 1 19-38. They specify three sets of activities to assess robustness: replication, convergence analysis, and boundary search. Id. at 136. 45. See supra notes 1-17 and accompanying text. 88 INDIANA LAW REVIEW [Vol. 31:81 court.^^ These examples highlight the fact that empirical research is, by nature, progressive. Each study yields discrete pieces of information relevant to overarching policy concerns."*^ Thus, studies ofjudicial decisions yield useful, albeit narrow information, that moves us toward a greater understanding of the bigger policy questions. For example, the Hall study was designed to assess consumer protection provided by courts. It was therefore designed to determine, from an empirical perspective, what factors are associated with judicial enforcement of a contractual right to medically necessary benefits. Specific factors were selected for study, including, for example, the type of insurance (public programs or private or government employer); what law governs (state contract law, state statute, federal statute, ERISA); and discretion assigned to the insurer."*^ The study was designed therefore to ascertain small, but key points. For example, do those who obtain coverage through public programs prevail significantly less often than those covered by private insurance or government employees? The study showed that in fact they prevailed in seventy percent of the cases; and further, that government employees prevailed in only thirty-one percent ofthe cases."*^ This empirical finding could lead to research to determine why government employees prevail less often. The Hall study also showed that when the dispute is governed by ERISA, the patient/policyholder is less likely to win.^° Individuals familiar with ERISA cases would have guessed that this was the case. Nonetheless, this empirical finding is important because it substantiates and quantifies the information. Thus, the Hall study, as well as other studies ofjudicial decisions, are likely to produce important pieces of empirical information despite the fact that the data set is far from perfect. The information may be important in and of itself. However, it is also important because it provides direction for further research. Thus, to answer the question Professor Sage presented, empirical analysis of judicial decisions does not risk the absurdity of "looking for the lost coin under the lamppost solely because the light is better."^' Conclusion Two further comments are noteworthy. First, although I have concluded that empirical studies of judicial decisions are a worthy pursuit despite their limitations, I recognize that it is the additional empirical research that stems from 46. Sage, supra note 1, at 68. 47. Harris M. Cooper, Integrating Research: A Guide for Literature Reviews 1 1 (2d ed. 1989) (explaining that, because of the cumulative nature of science, trustworthy accounts of past research form a necessary condition for orderly knowledge building). 48. Other variables included coverage language (general or specific); jurisdiction (state appeals, federal trial, federal appeals); likelihood of death; and seriousness of patient's condition. See Hall et al., supra note 12, at 1066-67. 49. Id. at 1062. 50. Id. 5 1 . Sage, supra note 1 , at 50. 1998] EMPIRICAL STUDIES 89 studies ofjudicial decisions that will most likely shape health law and policy. For example, in the ERISA context, the Hall study revealed that patients covered through ERISA plans, and thus whose disputes arising from a denial of coverage must be pursued as an ERISA claim, are less likely to win in court.^^ Although this is important information, there is still a larger empirical issue. Namely, does that outcome promote or hinder the policy goals underlying ERISA. Further empirical studies are needed to assess this broader policy question. For example, the primary policy goal underlying ERISA, and specifically underlying ERISA's civil enforcement provisions which allow suits, but limit the available remedies,^^ is that ERISA plans should be shielded from thefinancial risk associated with a denial of coverage in order to protect the plan as a whole. Thus, ERISA plans are protected from punitive damages and from compensation beyond the "benefit" itself, in order to protect the plan as a whole.^"^ The important empirical question, then, is whether the goal of protecting the plan is promoted. Accordingly, studies should be designed to assess, among other things, the impact of fewer patients winning on variables such as administrative expenses, premiums, health outcomes or patient satisfaction. A study designed to test whether the fact that patients prevail less often results in lower administrative expenses or health care premiums would need to compare administrative expenses and premiums in ERISA plans with expenses and premiums in non-ERISA plans. Such information is far beyond the information found in judicial decisions. More sophisticated empirical studies such as these will more readily facilitate the shaping of health law and policy. For example, as a legal scholar I have devoted substantial time and energy to developing doctrinal arguments that can be used to take cases outside the scope of ERISA,^^ because I believe the law of ERISA should be modified so that patients covered through ERISA plans are treated comparably to those in non-ERISA plans. However, an empirical analysis showing that, although fewer patients win their benefits claims under current ERISA regulations, the goal of greater benefit to the plan as a whole is advanced might convince me that the law of ERISA does not need to modified. Thus, although empirical studies ofjudicial decisions bring important information to light and are important initial steps that point out the direction of further research, it is the further research which is more likely to be influential in shaping health law and policy. However, the influence of empirical studies may be limited in the judicial 52. The Hall research team speculated that this was due to the standard ofjudicial review in ERISA cases when the insurer has been granted discretion. See Hall et al., supra note 12, at 1062-63. 53. See supra note 7. 54. See, e.g., Massachusetts Mutual Life Ins. Co. v. Russell, 473 U.S. 134 (1985) (holding that extra-contractual and punitive damages were not available in a claim under 29 U.S.C. § 502(a)(1)(B) (1994) of ERISA and the plaintiff was limited to the remedies set forth in § 502(a)). 55. See, e.g., Jordan, supra note 9, at 255; Karen A. Jordan, ERISA Pre-emption: Integrating Fabe into the Savings Clause Analysis, 27 RUTGERS L.J. 273 (1996). 90 INDIANA LAW REVIEW [Vol. 31:81 arena. This is because empirical research is, in some ways, at odds with legal analysis. That is, although empirical studies might influence health law scholars, they are less influential to lawyers in the position of advocate. Again, consider coverage disputes in the context of ERISA. An attorney in private practice faced with a client with a coverage dispute may discount empirical findings. For example, assume the following facts: 1) a client needed cardiac surgery immediately; 2) her HMO first required a second opinion; 3) her HMO then disagreed about what facility should be used; and 4) the client subsequently did not obtain the requisite preauthorization until her condition had deteriorated to the extent that surgery was no longer a viable option. In this case, an attorney would likely believe that the HMO was negligent and would want to find a way to hold the HMO responsible. This tactic would be the appropriate way to attain compensation for the client and her family and to prompt the HMO to develop expedited precertification procedures so that this unfortunate situation would not occur again. However, based on current case law, it would be difficult to hold the HMO liable if the client obtained her coverage through a plan governed by ERISA.^^ Nonetheless, to serve the client, the attorney would engage in a "legal analysis" of the judicial decisions, looking for ambiguities or inconsistencies in the case law, or distinguishing specific facts. The attorney would find a meritorious legal argument that would support imposing liability on the HMO, because the job of advocate requires it, and because the process of legal analysis permits it.^^ Moreover, an advocate would do this even though empirical studies may exist showing that holding the HMO liable would hinder public policy and would be worse for the ERISA plan than not pursuing the case at all. This type of scenario highlights the fact that legal analysis and empirical analysis may often be like ships that pass in the night. In the ERISA context, and perhaps in others as well, they seem to serve cross purposes. One possible consequence of this is that, although empirical studies should htlp shape health law and policy, it is less likely that they will have a great impact in the judicial arena. This lack of impact is because the judicial system is driven by individual disputes; arguments are advanced by attorneys in their role as advocate; and judges, although attentive to empirical data, will ultimately be striving to ensure justice in the individual case before the court. 56. Most courts readily find that cases involving allegedly negligent denials of coverage are preempted by ERISA. See, e.g., Jass v. Prudential Health Care Plan, Inc., 88 F.3d 1482 (7th Cir. 1996); Kuhl v. Lincoln Nat'l Health Plan of Kansas City, 999 F.2d 298 (8th Cir. 1993), cert, denied, 510 U.S. 1045 (1994); Corcoran v. United Healthcare, Inc., 965 F.2d 1321 (5th Cir.), cert, denied, 506 U.S. 1033 (1992). 57. See, e.g., LiNDA HoLDEMAN EDWARDS, Legal Writing: Process, Analysis and Organization (1996); Robert E. Rodes, Jr. & Howard Pospesel, Premises and Conclusions: Symbolic Logic for Legal Analysis (1997).