The Life and Death of Axis IV: Caught in the Quest for a Theory of Mental Disorder

A number of studies, most undertaken after introduction of the multiaxial system into DSM-III, examined the reliability and validity of axis IV as a measure of psychosocial stress. In a broad critique of these studies, Rey, Stewart, Plapp, Bashir, and Richards (1988) identified common methodological shortcomings leading to overestimation of both interrater reliability and generalizability of findings, and concluded that axis IV did not provide an adequate measure for either clinical or research purposes. Among the concerns noted were the absence of procedures or guidelines for eliciting information about the occurrence of particular stressors and their relative importance; patients’ subjective assessment of the impact of specific elements, including “retrospective falsification” by inflating rare or recent events while minimizing frequent yet minor ones; the difficulty of accounting for cumulative impact; and the difference in the role of specific stressors for various cohorts and cultural groups. This fundamental lack of reliability, the authors concluded, “is likely to be a major difficulty to be overcome before axis IV can make a solid contribution to the diagnostic system” (Rey et al., 1988, p. 290).

Another review of published reports empirically evaluating axis IV’s use, reliability, and validity was conducted by Skodol (1991). Examining whether axis IV ratings conformed to “clinical sense” and to previous research on the relationship between stress and mental disorder, Skodol found only “modest reliability and limited validity, beyond their value in identifying severe psychosocial stressors for the purpose of planning clinical interventions” (p. 503), though not necessarily for predicting treatment outcome. Another study by Mazure, Kincare, and Schaffer (1995) focused on comparing clinicians’ ratings of stressor severity with patients’ own appraisals. Their findings, unlike those of Rey or Skodol, suggest that axis IV can be used by clinicians to generate reliable or “objective ratings,” although Mazure et al. used interrater reliability to mean clinician–patient agreement rather than agreement among clinicians evaluating the same case.

More recent research has explored validity in a somewhat different way, as the relationship of axis IV (as conceptualized in DSM-IV) to diagnostic accuracy. Surveys and experimental studies by Hsieh and Kirk (2005) and Wakefield (2005) have examined whether identification of environmental stressors can help to rule out a “true” mental disorder (thus guarding against false positives) and, when disorder is present, to distinguish among plausible alternatives (thus guarding against misdiagnosis)—that is, whether the information captured in axis IV is helpful for increasing diagnostic accuracy or, if that cannot be definitively established, for decreasing diagnostic inaccuracy.

As Wakefield, Pottick, and Kirk (2002) are careful to point out, diagnosis of a mental disorder should only be given if symptoms represent an underlying dysfunction in the individual and not a reaction to a problematic environment. Behavior that is genuinely pathological, stemming from internal dysfunction, needs to be distinguishable from behavior representing a reaction to an adverse environment in order to prevent the false positives that occur when diagnostic criteria are incorrectly used to classify problems in living as mental disorder (Hsieh & Kirk, 2005; Wakefield, 2005). Distinguishing a true disorder from a “non-disordered reaction to a negative environment is an integral part of the quest started in DSM-III to formulate necessary and sufficient criteria for the valid diagnosis of mental disorders” (Wakefield, Pottick, & Kirk, 2002, p. 385).

How to achieve this has been the subject of intense debate. A recent example is the controversy over the bereavement exclusion for depressive disorder (Wakefield, 2012). In previous editions of the DSM, the exclusion was intended to differentiate symptoms due to a stressful event, such as loss of a loved one, from symptoms due to internal dysfunction, presumed to originate in neurobiology (e.g., a serotonin imbalance) and/or chronic psychodynamic factors; although symptomatology might be identical, a difference in cause could be used to distinguish “normal” from “abnormal” sadness. The bereavement exclusion has been deleted from DSM-5, however. Thus, for a diagnosis of depression, psychosocial context will no longer matter, although it is likely to retain its utility for providing a broader picture of a client’s life and its relevance for treatment decisions.

In general, the role assigned to context in formulating a psychiatric diagnosis has been inconsistent. As Thyer (2006) points out, criteria for assigning certain diagnoses such as conduct disorder include a clear directive to consider environmental factors, presumably by noting them under axis IV, and to refrain from giving the diagnosis if response to a disordered or deficient environment can account for the observed behavior. Quoting DSM-IV that a “conduct disorder diagnosis should be applied only when the behavior in question is symptomatic of an underlying dysfunction within the individual and not simply a reaction to the immediate social context … [thus] it may be helpful for the clinician to consider the social and economic context in which the undesirable behaviors have occurred,” Thyer suggests that this principle should be applied to allDSM disorders, since environmental circumstances are always relevant and indeed essential for assessment.

Thyer (2006) is arguing for psychosocial factors as catalysts for pathology he views as behavioral rather than mental, implying that “mental” disorder can only be diagnosed when there is no precipitating condition that might be noted on axis IV as playing a causal or contributing role. Without axis IV—that is, without a place to document environmental factors and thereby rule out internal disorder—the clinician has no option but to resort to the definition of mental disorder in the DSM: a pattern occurring in the individual, regardless of original cause (italics mine). Behavior that may represent a response or adaption to an oppressive, racist, abusive, or otherwise dysfunctional environment must therefore be conceptualized as neurobiological dysfunction. Eliminating axis IV thus promotes “lumping” and blurring, a potential threat to the validity of diagnostic categories.

Clearly, if people with dissimilar conditions are placed into the same category, reliability and validity of the category are undermined, since the purpose of a classification system is to structure decision making and enhance agreement. Use of axis IV can thus contribute to differential diagnosis by helping to determine which category an individual best fits (accuracy at the level of the individual) as well as which persons belong within a particular classification (accuracy at the level of the category). As Kirk and Hsieh (2004, p. 50) found, “the likelihood of experienced clinicians reaching a particular DSM diagnosis is significantly altered when the social context of the behavior changes” or when clinicians are provided with missing contextual information. If experienced experts display wide inconsistency in assigning people to categories, as they did in Kirk and Hsieh’s study, the idea of diagnoses as static criterion-based entities is seriously undermined, “shatter[ing] the illusion that diagnostic judgment can simply be a matter of matching presenting symptoms with DSM criteria, without the need to account for social context” (p. 51).

The utility of axis IV can be examined by looking at the frequency of use, subjective reports of its clinical usefulness, or studies of its predictive power for treatment outcome. In general, axis IV has been a “disappointment” because of its infrequent use in clinical and research settings although, as Skodol (1991) notes, statements about frequency are typically based on self-report rather than on systematic reviews of actual practice patterns such as written records. Nevertheless, surveys conducted during the 1980s indicate that a majority of practitioners did not routinely or consistently use axis IV, possibly because its complex severity ratings (in DSM-III) were considered too difficult to formulate and incorporate (Skodol, 1991). Other reasons may have included lack of training and support, the assumption that context was adequately taken into account through informal means or was too time consuming to assess formally, and perceptions about the unimportance of axis IV for billing within the dominant medical model.

Clinicians’ own assessment of axis IV’s utility has been inconsistent. Of the 274 clinicians participating in a 1978 study sponsored by the National Institute of Mental Health, 38% felt that axis IV was clinically useful in the form presented, 30% favored shifting to a purely descriptive approach, 14% thought the formulation was fine and probably useful for research but doubted they would use it themselves, and 10% favored eliminating axis IV entirely since “we do not yet know enough to measure stress in a way that would be useful clinically” (Spitzer & Forman, 1979, p. 819).

In another examination of practitioner use of axis IV, Schrader, Gordon, and Harcourt (1986) found that clinicians were able to make reliable criterion-based identification of psychosocial stressors, although determining the etiological significance of these stressors was more problematic. The difficulty, they found, came from both client and clinician variables. “It is not only the meaning of the stress to the patient that is problematic, it is also the meaning of the stress to the clinician” (p. 906), adding further ambiguity and making it nearly impossible to pinpoint generalizable relationships between type or degree of stress and particular disorders, as they had hoped to do.

Again, it is important to note that these were studies of axis IV as framed within DSM-III, not the simpler conceptualization found in the later edition. As with all research, how a question is asked will influence the response that is given. Had clinicians been asked whether identifying a client’s salient psychosocial stressors was useful—a formulation closer to the way axis IV was framed for DSM-IV—rather than about the importance of axis IV in formulating specific diagnoses, survey findings may have been quite different.

Context has always been a primary concern for social workers, and emphasis on the interaction between person and environment has traditionally been viewed as the hallmark of a “social work” approach to human problems. These interactions are recursive rather than linear, since psychosocial stressors such as relationship or occupational difficulties can represent the effects of a disorder rather than or in addition to their causes and, in turn, each can exacerbate the other. For social workers, assessment can be difficult to isolate as a stage preceding and distinct from intervention, unlike the medical approach where disease is first diagnosed and then treated. Attempts to apply a medical model of classification and treatment have thus placed social workers in a quandary.

Even with the inclusion of an environmental axis in DSM-III, many felt that a distinctly “social work” taxonomy was needed to classify problems in relationships, social functioning, and adaptation—issues that were often of primary concern to their clients, not simply extra information to be coded on a supplemental axis. As a result, the National Association of Social Workers funded the development of a classification system, later known as P.I.E. or Person in Environment (Williams, Karls, & Wandrei, 1989), conceived as a parallel or counterpoint to the increasingly dominant system of the DSM.

P.I.E. has never been widely used, however, in part because its categories are not linked to the types of assessment or treatment goals required for insurance billing (Probst, 2012), despite its authors’ optimistic prediction that “in the future, when the third party reimbursement system is less oriented toward the disease model, it is hoped that practitioners will be reimbursed for services provided to ameliorate social role and environmental problems” (Karls & Wandrei, 1992, p. 85). Other limitations of P.I.E. include the lack of empirical validation for its categories and its one-way conceptualization of the person–environment relationship, since individuals interact with their environments in complex, reciprocal ways. More than a simple interface, environment is “a web with hubs, linkages, intersections, patterns of communication, and pathways for the exchange of social capital” (Probst, 2012, p. 2). Exactly how elements of the environment contribute to the development or amelioration of problems remains elusive, however. The number of variables, combinations, intensities, interactions, and individual meanings is simply too vast for any model yet devised, and thus a true “environmental theory” of mental disorder has never been formulated.

Ironically, early versions of the DSMdid address the causal impact of environmental factors. The psychodynamic and psychosocial theories used in DSM-II viewed mental disorder as the result of “a mixture of a noxious environment and intrapsychic conflict” (First, 2010, p. 693), recognizing that mental dysfunction was a consequence of multiple, reciprocally interacting elements in the patient’s inner and outer worlds. The second manual sought to unravel the nature and origins of mental disorder, while the later editions sought to describe and classify. The shift in DSM-III to a neo-Kraepelinian perspective with its emphasis on observable criteria representing distinct categories—an attempt at a more “scientific” model, even if objective markers for each category had yet to be identified—served to undermine and eventually replace the psychoanalytic approach of the first two editions.

The abandonment of a psychodynamic orientation and its replacement with a purely descriptive model was not so much a repudiation of the idea that classification should be based on etiology as it was an attempt to formulate discrete, measurable syndromes resembling medical diagnoses and thus legitimizing “clinicians’ claims to be treating real diseases that deserved reimbursement from third party insurers” (First, 2010, p. 694). Eventually, it was hoped, the criterion-based definitions of DSM-III would “enable psychiatric researchers to make great strides toward understanding the underlying causes of mental disorders, with the eventual result that the descriptive approach would be replaced by a diagnostic system based on objective laboratory findings, as was the case with the rest of medicine” (First, 2010, p. 694), thus leading to an empirically based theory of mental dysfunction. As Paris (2013, p. 13) notes, “Kraepelin understood that psychiatric diagnosis must eventually be based on biological processes. But while waiting for specific markers to be discovered, categories could be provisionally based on signs and symptoms.”

Despite the hope of DSM-III’s creators that research would uncover specific relationships or causal links between environmental stress and psychiatric dysfunction, research has gone another direction, looking toward genetic and neurobiological factors to explain the cause of disorder and to differentiate one disorder from another. The Research Domain Criteria project of the National Institute of Mental Health, launched in 2009, may be the most ambitious effort to connect genetics, neural circuitry, and psychopathology. Its aim is to “translate rapid progress in basic neurobiological and behavioral research to an improved integrative understanding of psychopathology” (http://www.nimh.nih.gov/research-funding/rdoc/index.shtml)—a new kind of multiaxial system, perhaps, replacing psychosocial stressors with biomarkers and patterns of brain activity, and paving the way for a “future paradigm shift” (First, 2010, p. 693) that might, at last, establish a firm scientific basis for psychiatric diagnosis and provide a theory of who develops a mental disorder and why.

When it was first developed in the 1980s, axis IV represented psychiatry’s hope for a theory of mental disorder; its downfall, perhaps, was because it could not fulfill that hope. As Rey et al. (1988) note, many of the problems associated with axis IV stemmed from tension between its two aims. On the one hand, its intent (in DSM-III) was to be clinically meaningful, capturing the impact of psychosocial stress on diverse individuals for whom particular stressors have particular meanings (an idiographic approach, applicable to existing cases). On the other hand, it sought to be scientifically reliable, to provide an objective measure of the severity and significance of psychosocial factors across individuals, with the ultimate aim of helping to explain and predict clinical trajectories for future cases (a nomothetic approach, leading to the development of theory). Unable to do both things at once, axis IV eventually did neither and began to seem expendable.

In May 2004, the General Assembly of the APA (authors and publishers of the DSM) approved an action paper, by a vote of 81 to 29, calling for elimination of the multiaxial system in the next diagnostic manual unless there were empirical studies showing that patients had benefited, since “there is no evidence that most practitioners make use of it when they are free not to” (First, 2010, p. 698). The issue was referred to the Committee on Psychiatric Diagnosis and Assessment, which established a Multiaxial System Evaluation Committee to review clinical care, training, research, and administrative usefulness of each axis and of the system as a whole.

In a presentation to the Assembly in May of the following year, and in the Committee’s full report, Chairperson Juan Mezzich presented a comprehensive overview of research support—which was substantial—for the validity, reliability, and clinical utility of each of the axes as well as for the multiaxial system itself. Among the points made in support of axis IV was Skodol’s (1991) report of the shift from stressor severity in DSM-III (with only modest reliability) to identification of psychosocial stressors in DSM-IV (greater clinical usefulness), indication of improvement in the conceptualization and utility of axis IV over time (Mezzich, 2005). Put simply, axis IV had gotten better.

Evidence of the multiaxial system’s overall usefulness was provided by field trials, surveys, case reports, hospitalization decisions, and prediction of service utilization; additional evidence was provided for the usefulness of other multiaxial systems in use internationally (Mezzich, 2010). The system’s low regular usage in the United States was also noted, rendering its potential unfulfilled and thus contributing to its undervaluation. Among the reasons suggested for underutilization were an overemphasis on symptom checklists, perceived dispensability, inadequate training, and limited availability of effective formats for a structured recording of ratings (Mezzich, 2005), much like the points made by Williams (1985) and Skodol (1991) about the system in earlier versions of the DSM.

Based on the evidence, the Committee recommended that future editions of the DSM (including DSM-5, already well into development) maintain the multiaxial system, citing its consistency with biopsychosocial and person-centered care, increasingly perceived as important in both psychiatry and general medicine. The Committee also suggested that use of the system, while encouraged, continue to be optional. Additional recommendations were to (1) rewrite the mutiaxial section of DSM-5 to clarify its role and offer an adequate format to facilitate effective use; (2) increase accessibility to instructional materials and competent professional training; and (3) consider structural improvements for future multilevel systems, including provision of an idiographic narrative component for “flexible attention” to specific circumstances and traits.

Despite the Committee’s recommendation, the multiaxial system was eliminated from DSM-5; in fact, its fate was not even among the topics posted for discussion on the DSM-5 website. It simply disappeared. As Michael First, a Committee member and editor of the previous DSM-IV-TR, commented, “there is virtually nothing about the multiaxial system or its fate on the DSM-5 website. These decisions about the multiaxial system have not been put into writing anywhere that I know” (Michael First, personal communication, January 27, 2013).

Clearly, something happened between 2005, when the committee presented its recommendations, and 2012, when the new manual was closed to further revision, but the decision process was not made public. The omission is odd, given the transparency of discourse around other, less far-reaching revisions such as folding Asperger’s syndrome into autism. It may be, of course, that the transparency around certain proposed changes was the result of advocacy from particular constituent groups, such as the parents of children with Asperger diagnoses who were concerned with potential loss of services (Paris, 2013), rather than a reflection of an overall policy of transparency.

One reason the multiaxial system was quietly eliminated, despite evidence for its utility, may have been that there was no specific Work Group to review the system for DSM-5 as there was for DSM-IV, and thus no one to draw attention on the topic and shepherd it through the developmental process (Juan Mezzich, personal communication, January 30, 2013). The paucity of APA resources assigned to the issue both reflected and determined its perceived unimportance; something that did not even merit its own Work Group could not inspire dialogue, support, or space in the new manual. William Narrow, research director of the DSM-5 Task Force, states that the decision was made by the Task Force based on recommendations from two “ad hoc study groups” (William Narrow, personal communication, January 29, 2013); presumably these were internal, informal groups. Interestingly, Narrow was a member of the APA Committee that recommended retention of the system in 2005.

Another reason for the system’s removal, according to Narrow, was that “the multiaxial system was felt by some to have been treated as a required aspect of the diagnostic assessment, thereby segregating the mental health professions from other health professions whose diagnoses were not required to carry the additional information of the multiaxial system to justify interventions” (William Narrow, personal communication, January 29, 2013). In other words, since a pulmonologist is not required to account for psychological factors when recommending treatment for a pulmonary disorder, a psychiatrist should not have to account for medical conditions (axis III), environmental issues (axis IV), or anything outside the psychiatric domain when determining an intervention. At the same time, Narrow confirms that use of the multiaxial system was always optional, so it is unclear how a nonrequired system could have enough force to create a “false segregation” of psychiatry from other branches of medicine.

In addition, Narrow explains, “it was felt that the separation of an individual’s health conditions into discrete axes reflected a false brain/body dualism” (William Narrow, personal communication, January 29, 2013), perhaps referring to the placement of medical problems under a separate axis III rather than viewing all forms of disorder as a mixture of mental and somatic elements. Narrow may also be referring to the role of neurobiology in mental disorder, the focus of much of current government-funded research. Nevertheless, it is difficult to understand how elimination of the multiaxial system fosters brain/body integration. There are many examples of conditions within axes III and IV that impact psychiatric diagnosis, especially for children. For instance, medical conditions on axis III such as eczema or allergies, and ecological conditions on axis IV such as living in crowded conditions and witnessing domestic violence, can cause behavior that is indistinguishable from symptoms of attention deficit/hyperactivity disorder, conceptualized as a brain-based condition belonging to axis I. Deleting these factors from assessment omits both body and setting, leaving only the brain and thus reinforcing, through circular reasoning, that mental disorder “must be” fundamentally a matter of neurochemistry.

The question now remains of how information formerly captured within axis IV will be represented in DSM-5. According to Narrow, “Clinicians using DSM-5 will be urged to document psychosocial and environmental factors relevant to the patient/client’s diagnosis, clinical course, prognosis, or treatment, as well as relational problems, and problems related to abuse and neglect. These conditions and factors will be based on what was provided in DSM-IV, as well as the listings in the Z-code chapter of ICD-10-CM … Using the Z codes as an organizing classification is also anticipated to harmonize the documentation of these factors across disciplines in mental health, medicine, and allied professions. The listings were developed with participation from individuals with backgrounds in child psychiatry, community psychiatry, social work, classification, and the mental health/primary care interface” (William Narrow, personal communication, January, 29, 2013).

How this “urging” will work is unclear. Providing a list of possible factors does not mean the list will be used by clinicians or taken into account by third-party payers who decide if a client’s suffering merits reimbursable treatment. As First notes, “DSM-5 will contain an extensive list of codes—so extensive that I suspect it will not be used by clinicians. But those interested in noting such problems will now be able to do so in a much more detailed way” (Michael First, personal communication, January 27, 2013).

Over the years, many authors have voiced concern about the inadequacy of a purely descriptive system for capturing the information needed for sound clinical care, favoring a more etiologically based system (e.g., First, 2010; Mezzich, 2010), although they acknowledge that such a system is not yet available and would be rife with problems including the near impossibility of isolating the causal influence of gender, culture, poverty, family dynamics, and so on. Interestingly, one stressor that has been addressed in the new DSM is trauma—an environmental event without a brain-based origin. As Friedman et al. (2011) suggest, it may make sense to cluster all disorders stemming from trauma regardless of severity, “emphasizing common etiology over common phenomenology” (p. 737). Despite variation in clinical presentation, the conditions emanate from a known, high-impact adverse event and are distinct from other kinds of disorders in that a causal mechanism—linking person and environment—is part of the diagnostic criteria.