Empirically Based Decision Making In Clinical Practice

Recent years have seen a proliferation of treatments for a variety of mental health and health conditions. In mental health alone, there are hundreds of different treatments, both psychological and psychopharmacological, for depression and anxiety and at least as many for other conditions that are manifest in emotional distress and cognitive impairment (Corsini, 1981; Herink, 1980). In recent years, there have been grave concerns expressed that some of these interventions are ill-advised and even harmful. Recovered memory therapy, for example, is only the most recent of a number of therapies (see Singer & Lalich, 1996) whose practitioners have been criticized for failing to account for the damage that their treatments might produce (Brown, Scheflin, & Hammond, 1992; Loftus, 1993).

Accumulating evidence (e.g., Beutler, Williams, & Wakefield, 1993; Beutler, Williams, Wakefield, & Entwistle, 1995) indicates that most of the theories and approaches that are used within the community of practitioners are unsupported by empirical evidence of effects. As a result of such observations, there have been several concerted efforts in recent years, usually by academic practitioners, to identify which of these many treatment approaches warrant the acclaim of being "empirically supported" (e.g., Chambless et al., 1996; Nathan & Gorman, 1998; Roth & Fonagy, 1996). Though somewhat fewer in number than the number of theories available, the specter of needing to learn as many as 70 different research-based models in order to treat the variety of people who present themselves for assistance is undoubtedly a daunting proposition to practitioners.

The disparity between what treatments are accepted as "proven" by scientists and those that are considered valid enough to form the basis of practice by clinicians may index a conflict between the culture of science, on one hand, and both the values of clinical flexibility and the value placed on clinical judgment, on the other. Certainly, psychotherapy—and mental health more broadly—is a field in which the belief in clinical observation and the creative expression of the theorist has always been valued more highly than results from scientific method, nomothetic research, and statistical analyses (Elliott & Morrow-Bradley, 1994; Havens, 1994). Garb (1998), for example, observed that when scientific and personal belief contradict, clinicians are highly prone to accept the latter and discount the former. Thus, there has always been and continues to be a significant distrust of those who would question, let alone test, the validity of the great theories and practices of the masters. This pattern is reflected in the conflict that is evident throughout the history of psychotherapy (Cushman, 1992; Karon & Widener, 1995; Kessler, 1992; VandenBos, Cummings, & DeLeon, 1992). In reviewing this history, it is apparent that among any group of like-minded practitioners, the standard of evidence for the validity of a clinical practice has often been whether it fits the theory, whether the advocates strongly believe in the truth of their theories, and whether they appear to be sincere in advocating the value of clinical experience in support of their beliefs. While rational theories have certainly benefited the field, we must not forget that strongly held beliefs and rational logic have also led to such widely varied practices as blood-letting, demonology, the inquisition, the holocaust, and the crusades as well as every other destructive political and religious philosophy of which one can think.

One thesis of this article is that a standard of evidence based on clinical experience and personal belief is inadequate, if not actually dangerous. To leave patient well-being and functioning to the dubious validity of sincere beliefs and appealing clinical theories is to risk patient well-being. A second and related thesis is that contemporary lists of "empirically supported" treatments (ESTs) are also ill-advised as well as unreasonable and infeasible. Finally, the third thesis of this article is that a new method of integrating scientific evidence and the realities of practice is called for. I illustrate one possible method that arises from Systematic Treatment Selection (Beutler & Clarkin, 1990) and that urges clinicians to apply a refined list of empirically established principles of treatment, rather than learning and applying whole treatment models.

Traditional Criteria of Effective Treatment

In our field, where what is true and valid is more dependent on the shared beliefs of a specific group of theorist-practitioners than on either an accepted social standard or an accepted procedure to which we look for establishing the validity and value of what we do, the law and social institutions have moved to establish cross-cutting criteria by which the public can evaluate our work. The resulting sets of criteria, at once, reflect political expediencies, the reluctance of our field to commit ourselves to empirical definitions of treatment effects, and the aforementioned tendency to value shared professional opinion over scientific evidence. Thus, none of the most widely accepted standards of evidence rely on the scientific method.

In contemporary society, clinical practices are often held hostage to three standards that are used to evaluate the effectiveness of our work (Beutler, 1998). The first standard comprises the conventional criteria used by large, service delivery systems. The other two are used by the court to decide on cases of malpractice. In these latter cases, effectiveness is implicitly assumed if malpractice is absent.

Cost-Effectiveness Criteria

Health care systems have constructed a measure of effectiveness based largely on two indices: (a) the number of people served and (b) the cost of the services (Aaron, 1996; Fraser, 1996). Thus, for most health care accounting systems, the best treatment is one that reaches the most people and costs the least. These criteria are insensitive to whether a treatment causes harm or benefit; it is sensitive only to costs and numbers (e.g., Wickizer, & Lesser, 1998).

The Standard of Common Practice

A second set of criteria has been established by the courts, which have used these criteria as a standard against which to decide issues of malpractice—an indirect, albeit an inverse, indicator of effectiveness. Malpractice is defined in terms of a "standard of care," which in turn is defined by reference to the usual practices that exist in a community (Bongar, 1991; Furrow, 1993; Simon, 1992a, 1992b. The implicit test of whether a treatment is "good" or "effective" rests on the frequency of endorsement of the treatment in the community—its popularity. In other words, other things being equal, a popular treatment is assumed by the courts to be a better treatment than one that is not popular. A harmful treatment, or just an unpopular one, may be defined as constituting malpractice, but conversely, a harmful treatment that is widely practiced cannot ordinarily be defined as constituting malpractice.

The Principle of the Respectable Minority

Because the courts recognized the importance of new developments that often are unknown or even unpopular, they constructed a third criterion of effectiveness. The doctrine of the "respectable minority" (Keeton, Dobbs, Keeton, & Owen, 1984) holds that a treatment cannot be held to constitute malpractice if (a) the treatment has an explicit theoretical foundation and a standard of delivery and (b) a respectable minority of those in the profession adhere to the underlying theory. While the definition of what constitutes a respectable minority has not been specified in the law, there is at least one case in which as few as six practitioners was sufficient to uphold the practice (summarized by Beutler, Clarkin, & Bongar, 2000). In the case law that established this criterion of its value, the treatment was not known to have any positive effects and was widely rejected by experts in the field. However, its theoretical framework had been described in writing and several (fewer than a dozen) people sincerely believed in its value. The court held that in such cases, effectiveness or value was assumed, without regard for any scientific evidence of effects.

The competition and frequent contradiction between the standard of common practice and the standard of the respectable minority is illustrated in the well-known Osheroff case (Klerman, 1990; Malcolm, 1986). In this case, the patient, Osheroff, was treated at Chestnut Lodge Hospital for a narcissistic personality disorder and bipolar depression with long-term, inpatient psychoanalysis, with no effect. After experiencing severe deterioration, he left the hospital and transferred to another psychiatric hospital where he was treated successfully with a brief course of psychotropic medication and supportive psychotherapy. He subsequently brought suit against Chestnut Lodge Hospital for malpractice, alleging that it failed to provide the most widely accepted treatment and neglected to inform him of this and that it failed to inform him of other treatment alternatives. He argued that the standard of care to which the hospital should be held includes psychotropic medication, since these are commonly given for depression.

Chestnut Lodge invoked the defense of the respectable minority rule, pointing out that psychoanalytic treatment was widely used for this type of problem, even though there were no research studies to support its effectiveness in such instances. The case was settled out of court and, thus, no case law evolved to resolve this conflict, but this case illustrates that neither of the invoked standards are sensitive to scientific evidence of what constitutes effective treatment.

Scientific Evidence as an Alternative Standard of Effectiveness

Why is there so much reluctance to consider demonstrable treatment outcomes as a key part of deciding whether a treatment is useful and beneficial? Adopting a standard based on effectiveness seems like such an obvious alternative to the current criteria, but there are processes that work against a scientific analysis of treatment effects from being incorporated into our definitions of "good treatments." For example, if standards of practice and of health care delivery were to consider the propositions that some treatments are effective and some cause harm, contemporary health care and service delivery programs probably would be faced with the demand that they ensure that all services they provide meet these standards. This would reduce the flexibility by which treatment is provided and would delay entry of new treatments into the arena of practice. Moreover, invoking such a standard would mean that clinicians would have to be trained and maintained at specified levels of performance; health care companies would have some responsibility for ensuring that this was done. That is, they may be called on to ensure that providers were trained in these scientific procedures and that they practiced them correctly. This would be costly and the question would then arise as to whether health care companies would be liable if this training was not done and if a given level of performance was not maintained.

Concomitantly, practitioners would face the possibility that they would be forced by these standards to give up favored positions and theories for which evidence of effectiveness is lacking. For the sake of financial survival, they would be required to adopt approaches with which they may either be unfamiliar or ineffective, in order to increase the scientific credibility of their work (Beutler, Kim, Davidson, Karno, & Fisher, 1996). Such a decision would place what may be a valuable premium on objective evidence and would accord relatively less credence to treatments that are based primarily on good intentions, strong beliefs, costs, popularity, and legal precedent.

Such a standard would seem, on the surface, to introduce a needed degree of objectivity into decisions of treatment assignment. However, this conclusion may be deceptively simple. Science is by no means clear about what constitutes adequate evidence of a treatment’s safety and benefit. Referencing research would introduce its own kind of ambiguity into treatment decisions. Research studies often reach very different conclusions, vary in quality, and have varying levels of relevance for clinical practice. Before implementing scientific standards, we must reach some consensus on how to evaluate this evidence and how strong it must be. A scientific standard must address both the positive and negative findings that emerge regarding a treatment and come to some conclusion by weighing the two lines of evidence.

Two methods have been developed and used to overcome these problems: (a) "consensus panels," in which identified experts debate and reach agreement on the nature of what is the best practice or treatment; or (b) "literature reviews" by single authors or groups, with a focus on the development of recommendations for practice guidelines. Both methods are replete with potential and real bias, however. For example, consensus panels are invariably chosen to represent the biases of the group that sponsors them. It is this group that defines the nature of "expertise," and frequently, the report is subject to a review and vote of the organizational membership. These things ensure that the bias of the group is represented in the findings.

For example, Barlow (1994) has aptly demonstrated how the guidelines recommended by the Agency for Health Care Policy and Research for treating depression are at dramatic variance from the literature reviewed by the panel in developing their recommendations. Likewise, the American Psychiatric Association (1993) guidelines for treatment depression reflect a common medical bias that is inconsistent with the strength of available literature (see Beutler et al., 2000). Consensus panels of experts reduce all questions about what constitutes good and effective practice to a popular vote, reducing it to a variant of the legal definition in which "popularity" (i.e., common practice) is the standard of practice.

Literature reviews are not immune from bias either. Reviews of research literature are dependent on the personal beliefs and biases of those who summarize the findings. Their presentation often reduces to a variant of the Principle of the Respectable Minority. The recommendations and conclusions have passed the test of one or a few authors whose opinions are not immune from bias either.

Efforts to employ truly reliable and objective standards to the definition of what treatments are and are not empirically supported have been few. The notable exception is the Task Force on the Identification and Dissemination of Empirically Validated Treatments, formed by the Division of Clinical Psychology (Division 12) of the American Psychological Association. Commissioned in 1993, this task force issued its first report (Task Force on Promotion and Dissemination of Psychological Procedures, 1995) in 1995 and subsequently issued addendum reports in 1996 (Chambless et al., 1996) and 1998 (Chambless et al., 1998). It now functions as a standing committee of the division, with the mandate to identify those treatments that have received sufficient empirical support to validly be transferred to practice.

Central to effort of this task force has been a process designed to define what constitutes an objective standard that can be used to mark the point at which a specified treatment has passed the test of effectiveness. The standard selected is modeled after that used by the Federal Drug Administration to define when a particular medication can be considered to be safe and effective. However, to apply this standard, only treatments that have been delivered via structured manuals have been considered to be appropriate. Only such manualized treatments have been considered to have sufficiently reduced therapist and theoretical variability to a point that the results can be interpreted validly (Chambless & Hollon, 1998).

Many authors (e.g., Garfield, 1996; Silverman, 1996) have criticized both the methodological basis and the criteria by which ESTs have been defined. While there is value in these arguments, a more important concern may well reside in the implicit assumptions that underlie such lists in the first place. Lists of ESTs contain a variety of different treatment models, embodying different theoretical frameworks, and each is assumed to be appropriate for different clusters of symptoms or diagnostic conditions. An unstated assumption of such a list, for example, is that clinicians would master several different models of treatment in order to apply them differentially to different types of patients and different symptoms. Given the variety of patients a given practitioner is likely to see, this could mean learning as many as 70 different manualized approaches, each to some (usually unstated) criteria of compliance and expertise. These expectations are unrealistic, both in view of the lengthy training that is usually required to achieve expertise in just one of these approaches and because many of the theory-based approaches embody theoretical constructs that are inconsistent with the theories already accepted by practicing clinicians (Beutler et al., 1995) and with one another as well. The premises of interpersonal psychotherapy (Klerman, Weissman, Rounsaville, & Chevron, 1984, for example, substantially differ from the premises of cognitive therapy (Beck, Rush, Shaw, & Emery, 1979), and both are inconsistent with the premises of experiential therapy; however, models of all of these treatments are considered to be empirically supported by these criteria.

Even beyond these constraints is the reality that most clinicians do not like structured manuals and tend to depart from them after training. In addition, many of them are not attracted to the particular behavioral and cognitive theories that have met the criteria used to assess effectiveness (Silverman, 1996). These concerns render the recommendations impractical for use by managed care companies, under most circumstances. Concomitantly, there is no substantial evidence that these companies are inclined to commit the resources necessary to ensure that there is a sufficiently large cadre of practitioners available to represent the various methods and manuals that are identified. Also, there is no evidence that they are inclined to devote resources to ensure that each approved provider is trained in and capable of reliably delivering several different models of treatment.

However, even if managed care companies were so inclined, there may be unexpected side effects of applying structured manuals. It may be that conventional manuals are inordinately rigid, even if managed health care were to accept the value of making treatment decisions from the basis of scientific findings (Caspar, in press). The requirement that clinicians delimit the use of their usual, cross-cutting, or preexisting interventions is distasteful to clinical practitioners, and some manuals may even result in a deterioration of clinicians’ general therapeutic skill (Henry, Strupp, Butler, Schacht, & Binder, 1993). Indeed, there is at least suggestive evidence that the most effective clinicians may be those who occasionally violate the manualized treatment guidelines (Anderson & Strupp, 1996).

A Combined Standard: Literature Review and Cross-Validation

Caspar (in press) wrote a series of papers that review issues in training psychotherapists, the majority of which conclude that conventional treatment manuals are indefensibly inflexible and too closely bound to theories that are not widely accepted by practitioners. Beutler et al. (2000) have attempted to address these needs by modeling a method of defining treatment guidelines that are empirically supported but sufficiently flexible to be integrated into a variety of theories that are preferred by clinicians. This approach is founded in the initial work of Beutler and Clarkin (1990) but is reminiscent of Goldfried and Padawar’s (1982) earlier assertion that integrated treatments might be based on empirically defined principles rather than on explanatory theories or specific techniques. Beutler and Clarkin proposed a model of Systematic Treatment Selection that takes a broad-ranging descriptive (rather than explanatory) theoretical perspective, where the focus is on defining empirical relationships between treatment acts that are under the control of the clinician, on one hand, and treatment outcomes, on the other, rather than on either theories of psychopathology or theories of how people change.

While the Systematic Treatment Selection model asserts that a given clinician’s theory of mental health and psychopathology often guides the content of what a patient learns in successful psychotherapy, it has little differential or specific effect on the outcomes (see Beutler, 1983; Beutler, Machado, & Neufeldt, 1994; Beutler et al., 1991 for a review of some of the literature that supports this "do-do bird" viewpoint). From an extensive literature review, Beutler and Clarkin posited that employing a theory of how to assign the mix of treatment techniques to patients with varying needs may improve the efficiency of outcome predictions over the usual methods of assessing the effects of one theoretical model against another.

Beutler and Clarkin drew heavily on their own research (Beutler, 1983; Frances, Clarkin, & Perry, 1984) and on the theoretical perspectives of other writers who advocated therapeutic integration and technical eclecticism (e.g., Lazarus, 1976; Lazarus, 1993; Norcross, 1986; Prochaska, 1984). These theoretical perspectives were integrated with available research to formulate initial hypotheses of how different patient variables might guide treatment selection. The ensuing model proposed a process that includes four interactive and cascading levels of decision making in treatment planning:

Level 1—Identifying and measuring patient predisposing variables (severity, distress, resistance, problems and symptoms, social support, chronicity/complexity, and coping styles)
Level 2—Selecting a level or context of treatment (treatment intensity, group vs. individual modality, psychosocial vs. medical modality, etc.
Level 3—Selecting a therapist and applying therapist interventions activity/relationship enhancement variables (therapist skill, therapist activity, therapeutic procedures, therapeutic alliance)
Level 4—Matching a patient and class of treatment (fitting a particular patient quality with the selection of various classes of interventions)

Beutler and Clarkin’s (1990) review identified over 30 patient variables (Level 1) that appeared to affect subsequent decisions (Levels 2 through 4). In subsequent reviews of literature, which increasingly applied standards of reliability and prediction (Beutler & Berren, 1995; Beutler & Consoli, 1992; Beutler, Goodrich, Fisher, & Williams, 1999), the overlap among these specific variables was reduced, and the list was pared to fewer than a dozen on which a critical mass of research was judged to be present.

In a culmination of this latter effort, Beutler et al. (2000) initiated a three-phase effort to identify and then test hypotheses about how these variables could assist the clinician to select and apply different types of treatment. In the first phase of this investigation, the authors reviewed approximately 2,000 research studies on nonbipolar depressive disorders in an effort to identify studies in which a reliably identified patient variable had been found to relate to outcome either directly or as moderated by some aspect of treatment that was identified within one of the other levels of treatment selection (Levels 2 through 4).

From this review, and after applying procedures to ensure the reliability and specificity of the variables identified, just under 300 studies were identified that had found a reliable patient quality or a specific class of treatment procedures that was related to outcome. Hypotheses were developed about what qualities of treatment procedures, at Levels 2 through 4, best fit with each identified patient characteristic defined at Level 1. From iterative reviews of the final studies, we identified both the patient qualities that were most frequently represented in outcome research (and the most common relationships or patterns of fit between these patient qualities) and specific aspects of treatment itself. These relationships were framed as hypotheses regarding the conditions under which one class of intervention would work better than another. These were framed as both prognostic (predictors of outcome) and matching (patient–therapy compatibility) hypotheses.

The second step in this process was to cross-check these hypotheses through a second review of literature. In this instance, the literature review was extended to conditions and problems that have been found to be closely related to depression and that are assumed to have an indirect but meaningful relationship to disorders in this latter spectrum. These dominantly included some generalized anxiety disorders, alcohol abuse, and drug abuse disorders. The result of these two steps was a distillation of 15 hypotheses in which a particular patient variable or patient–treatment fit was thought to relate to one of the following treatment variations (across patient diagnosis): (a) intensity (a Level 2 variable), (b) the use of psychosocial versus psychopharmacological options (a Level 2 variable), (c) multiperson versus individually delivered interventions (a Level 2 variable), (d) directive versus nondirective interventions (a Level 3 and 4 variable), (e) symptom-focused versus insight-focused interventions (a Level 3 and 4 variable), and (f) abreactive versus supportive interventions (a Level 3 and 4 variable).

The next step in the process was to initiate an independent and systematic test of these hypotheses. In this process, we classified the hypotheses generated in the first two steps into two classes (see Appendix). One class of hypotheses addressed what we called "basic guidelines." These referred to hypotheses regarding relationships that can be assessed without direct observation of the clinical treatment session. Basic guidelines apply to predictions about the patient qualities that predispose the effective use of different treatment modalities, formats, and intensities (Level 2) but that cut (or generalize) across diagnoses.

A common cadre of trained clinicians rated all patients with a standard rating form that allowed the definition of the patient variables defined in our earlier review (Fisher, Beutler, & Williams, 1999). Completion of this rating form was based on each patient’s videotaped intake session, an extensive background description and intake note, and a variety of psychological trait and state measures that were used by the various research projects at intake. The clinician ratings were supplemented further, by standardized measures of pre-to-post treatment changes and by ratings of treatment processes based on early- and late-treatment videotapes, and completed by a separate set of trained raters. These latter ratings provided standard and reliable estimates of what was actually done in treatment sessions, as well as of the quality of therapeutic relationship that developed. We were then positioned to explore the chain of effects among patient predisposing variables (Level 1), aspects of subsequently assigned treatment type and intensity (Level 2), the direct impact of both the interventions used and the quality of the therapeutic relationship (Level 3), and the role of matching patient and treatment qualities (Level 4).

The treatments provided to patients in this study varied widely as a function of the objectives guiding the research program from which we drew our samples. However, all treatments were closely monitored for fidelity and purity. Included among the treatments were: individual, group, and couples treatments, representing different formats; psychosocial and pharmacotherapy modalities; face-to-face and telephone interventions to represent differences in treatment intensity; and cognitive, experiential, systems, and psychodynamic models to represent variations both in how emotional experience was addressed and the relative value placed on behavioral versus insight-based changes. These various interventions were conducted according to established treatment manuals, but we also included a "therapy as usual" condition that was not conducted according to a manual.

Using this relatively large data set, we undertook direct tests of the hypotheses that had been extracted from the earlier literature reviews. Thirteen of the original 15 hypotheses were supported. These hypotheses were restated as principles for guiding the application of treatment. To these, we added 5 guidelines that were derived from available research on patients who pose major risks of violence and self-harm. These latter guidelines drew from ethics and extant standards of practice rather than directly from the cross-validation phase of our study. Together, the supported hypotheses and the 5 additional principles comprised 18 guidelines or principles for treating patients who present with significant depression. Ten of these principles reflected what we have previously referred to as "basic" guidelines, and 8 reflected what we have called, "optimal" guidelines.

Two of the guidelines predicted general prognosis based on patient chronicity, impairment, levels of social support, and so on; one was related to predicting when psychotropic medication would be helpful; five related to assessment of risk; two applied to conditions for establishing a working alliance; one applied to the role of exposure; and seven applied to the identification of patient indicators for the differential application of various styles, formats, and methods of intervention.

These 18 principles constitute testable, empirically informed guidelines for selecting and applying treatment programs that cut across theoretical viewpoints. They fit our initial intention of not being so technique specific that they can only be applied within a certain theoretical model. Instead, they direct the therapist/clinician to select and employ his or her own favored procedures for accomplishing the objectives defined within the principles.

We are now in the process of undertaking an independent, prospective, randomized clinical trial of a treatment built on these principles. This research effort, supported by the National Institute of Drug Abuse (NIDA) compares our "prescriptive psychotherapy" with manualized varieties of cognitive therapy and a relationship-oriented therapy. The prescriptive therapy being tested is based on the 10 principles from the list of 18 that apply specifically to the application of individual psychotherapy. At present, we are trying to demonstrate that we can train therapists to use their own favorite procedures within the context of a "prescriptive" model that is both flexible and discriminating. Our goal is to define what constitutes an artful clinician who applies empirically informed procedures in a maximally effective and reliable way. We have come to believe that the use of descriptive principles, rather than either global explanatory theories or lists of techniques, to guide treatment selection, as originally suggested by Goldfried and Padawar (1982), is what separates the skilled clinician from the technician. By following principles, the artful clinician can transcend the draw of canned explanations in favor of creating novel interventions that fit new demands and unfamiliar complexities. The technician, on the other hand, is limited by the techniques endorsed by his or her particular explanatory theory of psychopathology and psychotherapy, making it difficult to flexibly adapt to new situations and unusual patients. The flexibility that is inherent in this principle-driven model is expected to make it easier for clinicians to learn the procedures and to increase the likelihood that they will continue to use them, once learned.

Conclusions

I believe that the current system of managed care and the current approach to defining empirically supported treatments are shortsighted. On one hand, managed care programs overemphasize the cost, the access to untested services, and the acceptance of various treatments among peers at the expense of scientific evidence of effectiveness and efficacy. On the other hand, efforts to construct a finite list of empirically supported treatments ignore the unrealistic assumption that practitioners will willingly and easily forsake their own experience and preferences, that they will efficiently learn many different and contradictory methods in order to treat the variety of problems with which their patients present, and that once these skills and techniques have been learned, they will continue, unaffected, indefinitely.

While the scientific method offers the greatest hope of guiding the field to identify effective processes and standards, I believe that the obstacles mentioned above warrant a new approach. This new approach must, at once, identify the nature of effective treatments, the conditions of their application, maintain respect for the role of therapist investment in their own clinical experience and preferred theoretical models, and encourage managed care programs to reflect on the scientific status of the treatments advocated and used. Identifying and then learning to apply a treatment that draws from empirically established principles of intervention rather than from specific, theory-specific models promises to ease the process of instituting empirically supported interventions in actual clinical settings.

My colleagues and I (Beutler et al., 2000) have presented one effort to construct a set of treatment guidelines that are both cross-cutting (i.e., can be applied from a number of theoretical models) and empirically derived. We have done this by identifying cross-theory classes of interventions, based on similarities in the demand characteristics required to employ them, and then defining the patient and problem qualities for which these classes of intervention have been demonstrated to work.

When one applies these guidelines, the result is a treatment that is comprised of procedures whose application is, at once, guided by general descriptive principles rather than either overly general theories or overly restrictive techniques and is consistent with the interventions with which a given practitioner is familiar and skilled. By avoiding the traps of being either too narrow or too broad, the resulting treatment holds promise for helping clinicians perform in a way that is flexible, skilled, and specific as they employ techniques from their own models in ways that best fit particular patient needs and response dispositions.

To apply this type of model, practitioners must become better artists. This does not mean that they ignore scientific principles, but that they apply them in creative ways, consistent with the wide variations and permutations of problems and characteristics presented by their patients. The distinction between the role of a technician and that of an artist can be illustrated by a personal note. I am an instructor in "natural horsemanship," a form of training horses that was made popular in the movie, "Horse Whisperer."

A central tenet of this form of horsemanship is that horses are prey animals and that humans, in order to establish a true partnership, humans (as predators) must overcome the horse’s natural fear of being killed by predators. In natural horsemanship, we do not believe in the use of pain and dominance as training procedures to overcome this barrier. These are the methods of the "old school"—the school of "horse breaking" rather than "horse training"—and are thought to increase rather than reduce the natural enmity that exists between horse and human. Our methods are ones of cooperation and communication. In "normal" or old school training methods, one broke a young horse by dominating it (i.e., tying the animal down, snubbing its nose to a post, putting a blanket over its head, throwing a saddle on its back, and then riding the animal until it quit bucking—a process that takes about 2 weeks of daily work.

In natural horsemanship, we apply principles of positive and negative reinforcement to teach the animal that it can find comfort and safety by complying with the directives of the herd leader (the trainer). Thus, rather than being an adversary and an object of fear, the trainer attempts to become a partner and guide. Within 30 minutes, I can teach a student rider how to use the principles of positive and negative reinforcement, within a 50-ft pen, and can guide the rider in the development of techniques that will allow him or her to take a young, untrained horse from being wild to being "hooked" or "joined" with a trainer. Within 2 hours, the horse will usually accept a rider. The student can then perform these techniques on another horse in the small pen, usually with equally good results. The true test of the student’s learning, however, comes when he or she is asked to teach a horse to do a very complex task in a new environment (e.g., to come to come to a trainer at his or her command, across a 40-acre pasture rather than a small pen, and in this process to navigate rivers, obstacles, and other predatorial dangers). If the student has become an artist-practitioner, he or she will know that the same principles apply in this new setting and to this new task as they did in the small pen. The student will know how to break the tasks down, to build complex behaviors from simple ones, and to construct creative methods for evoking the wanted behaviors. If the student has not achieved the goal of artist-practitioner, he or she will be unable to engage the task. The techniques will get in the way.

Appendix

Principles for Systematic Treatment Selection

Reasonable and Basic Principles

Prognosis

1. The likelihood of improvement (prognosis) is a positive function of social support level and a negative function of functional impairment.

2. Prognosis is attenuated by patient complexity/chronicity, and by an absence of patient distress. Facilitating social support enhances the likelihood of good outcome among patients with complex/chronic problems.

Level and Intensity of Care

3. Psychoactive medication exerts its best effects among those patients with high functional impairment and high complexity/chronicity.

4. Likelihood and magnitude of improvement are increased among patients with complex/chronic problems by the application of multi person therapy.

5. Benefits correspond to treatment intensity among functionally impaired patients.

Risk Reduction

6. Risk is reduced by careful assessment of risk situations in the course of establishing a diagnosis and history.

7. Risk is reduced and patient compliance is increased when the treatment includes family intervention.

8. Risk and retention are optimized if the patient is realistically informed about the probable length and effectiveness of the treatment and has a clear understanding of the roles and activities that are expected of him or her during the course of the treatment.

9. Risk is reduced if the clinician routinely questions patients about suicidal feelings, intent, and plans.

10. Ethical and legal principles suggest that documentation and consultation are advisable.

Optimal Principles

Note: The original order of the principles have been rearranged to reflect some commonalities.

Relationship Principles

1. Therapeutic change is greatest when the therapist is skillful and provides trust, acceptance, acknowledgment, collaboration, and respect for the patient within an environment that both supports risk and provides maximal safety.

2. Therapeutic change is most likely when the therapeutic procedures do not evoke patient resistance.

Principle of Exposure and Extinction

3. Therapeutic change is most likely when the patient is exposed to objects or targets of behavioral and emotional avoidance.

4. Therapeutic change is greatest when a patient is stimulated to emotional arousal in a safe environment until problematic responses diminish or extinguish.

Principle of Treatment Sequencing

5. Therapeutic change is most likely if the initial focus of change efforts is to build new skills and alter disruptive symptoms.

Differential Treatment Principles

6. Therapeutic change is greatest when the relative balance of interventions either favors the use of skill building and symptom removal procedures among patients who externalize or favors the use of insight and relationship-focused procedures among patients who internalize.

7. Therapeutic change is greatest when the directiveness of the intervention is either inversely correspondent with the patient's current level of resistance or authoritatively prescribes a continuation of the symptomatic behavior.

8. The likelihood of therapeutic change is greatest when the patient's level of emotional stress is moderate, neither being excessively high nor excessively low.

Portions of this article, under the same title, were presented as a keynote address at the Canadian Psychological Association meetings, Halifax, Nova Scotia, June 1999. This article was partially supported by National Institute of Drug Abuse Grant RO1DA09394.

Correspondence concerning this article should be addressed to Larry E. Beutler, Counseling/Clinical/School Psychology Program, Department of Education, University of California, Santa Barbara, California 93106.
E-mail: beutler@education.ucsb.edu