An innovation in the third edition of the American Psychiatric Association’s Diagnostic and Statistical Manual of Mental Disorders (DSM-III), carried over into DSM-III-R and DSM-IV, was the presentation of diagnostic criteria for the identification of each of the specific disorders included in the manual. While the immediate goal of including such criteria was to improve reliability by minimizing criterion variance (1), the ultimate goal was to help clinicians and researchers make valid diagnoses by minimizing both false positives (instances in which individuals who do not have a particular mental disorder are mistakenly diagnosed as having the disorder) and false negatives (instances in which individuals with a particular mental disorder are mistakenly diagnosed as not having the disorder).
The use of specified diagnostic criteria, in both clinical and research settings, quickly gained wide acceptance. However, the high lifetime and current prevalence rates for mental disorders in the community, based on DSM-III criteria in the Epidemiologic Catchment Area study (2) and DSM-III-R criteria in the National Comorbidity Survey (3), raised concern that these criteria were overly inclusive (4). The concern was that even when correctly applied as stated, the criteria diagnosed many individuals who were exhibiting normal reactions to a difficult environment as having a mental disorder, whereas disorder implies an internal dysfunction of some psychological or behavioral mechanism (5–7). In other words, the concern was that the criteria resulted in many false positives, which we call the "false positives problem." This concern with the adequacy of the DSM criteria has not been limited to the mental health professions. Critics from outside psychiatry have been particularly vocal in arguing that the DSM criteria fail to adequately distinguish mental disorder from normal behavior and represent a dangerous medicalization of social problems (8–13).
The developers of DSM-IV attempted to deal with the false positives problem by adding the "clinical significance criterion" to the criteria sets for a large number of diagnostic categories. The clinical significance criterion requires that the individual exhibit "clinically significant distress or impairment." The clinical significance criterion is a version of a criterion that was used to deal with potential false positives in three DSM-III-R categories: social phobia, simple phobia, and obsessive-compulsive disorder. Because mild forms of the symptoms of these three disorders were thought to be common in the community, even in cases where there was no disorder, a criterion requiring marked distress or impairment in social functioning was added, although the term "clinically significant" was not used.
Comparing DSM-III-R to DSM-IV, we found that of the 281 axis I and axis II categories in DSM-IV that have diagnostic criteria (not including the "not otherwise specified" categories, which have no criteria sets), 130 (46%) include the clinical significance criterion as a new feature (either an addition to the criteria set for a category that existed in DSM-III-R or part of a criteria set for a new DSM-IV category). The clinical significance criterion was added to at least some categories in every major axis I diagnostic group except schizophrenia and other psychotic disorders, where it was felt that the nature of the symptoms made such an addition unnecessary. In some major groups of disorder, such as sexual dysfunctions and sexual paraphilias, the clinical significance criterion was added to every specific diagnosis.
In none of the many reviews of DSM-IV (14–17)—with the exception of some brief comments by Wakefield (18)—has this major innovation been discussed. In this article we critically examine the DSM-IV clinical significance criterion and assess its adequacy for addressing the false positives problem.
In this article "false positive" refers to classification of a nondisordered condition as a disorder when the DSM criteria themselves are applied correctly as stated. Obviously, false positives can also occur when the criteria are ignored or incorrectly applied, as when a clinician ignores the required 2-week duration of depressed mood when applying the criteria for major depressive disorder. False positives resulting from such misapplication of the DSM criteria have not been the concern of DSM critics and are not considered here, although the magnitude of this kind of false positives problem may be considerable. The phrase "false positives problem" considered here concerns only false positives due to the inadequacy of the diagnostic criteria themselves, not to their misapplication.
The magnitude of the false positives problem in various settings is unknown. There are studies that examine how often a diagnosis resulting from a structured diagnostic interview administered by a lay interviewer corresponds to clinical application of DSM criteria (19–23). These studies rely on the expert application of DSM criteria as the criterion for validity, so they do not address the question of whether correct application of DSM criteria yields false positives. We know of no study that attempts to empirically establish how often the correct application of the DSM criteria results in the diagnosis of an individual as having a mental disorder when in fact the individual—according to expert clinical judgment or some other criterion independent of DSM criteria—does not have the disorder. Although the magnitude of the problem has not been empirically studied, sufficient examples of hypothetical apparent false positives have been presented in the literature (24, 25) to suggest that the problem may be substantial. Clearly every effort should be made to address the problem as effectively as possible.
The DSM-IV clinical significance criterion appears as an additional required criterion, added to the symptom, duration, and exclusion criteria. Although the exact wording varies, the most common form of the clinical significance criterion is, "The symptoms cause clinically significant distress or impairment in social, occupational, or other important areas of functioning." In common variations, sometimes particular areas of impairment are noted.
Although the four volumes of the DSM-IV Sourcebook(26–29) have chapters that explain the rationale and empirical justification for many changes made in DSM-IV, there is no chapter that deals with the clinical significance criterion. Similarly, none of the DSM-IV field trials (29) that included the clinical significance criterion were examined to determine what effect, if any, their inclusion had on base rates or on minimizing false positive diagnoses. Thus, unlike the changes in DSM-IV that were based on empirical data, the decision to include the clinical significance criterion throughout the manual was made strictly on conceptual grounds. The DSM-IV Task Force considered the merits of including the clinical significance criterion for major depressive disorder, but the broader use of the clinical significance criterion throughout the manual was not discussed.
A paragraph explaining the reasons for introducing the clinical significance criterion into DSM-IV appears in the section Use of the Manual:
Criteria for Clinical Significance
The definition of mental disorder in the introduction to DSM-IV requires that there be clinically significant impairment or distress. To highlight the importance of considering this issue, the criteria sets for most disorders include a clinical significance criterion (usually worded "...causes clinically significant distress or impairment in social, occupational, or other important areas of functioning"). This criterion helps establish the threshold for the diagnosis of a disorder in those situations in which the symptomatic presentation by itself (particularly in its milder forms) is not inherently pathological and may be encountered in individuals for whom a diagnosis of "mental disorder" would be inappropriate. Assessing whether this criterion is met, especially in terms of role function, is an inherently difficult clinical judgment. Reliance on information from family members and other third parties (in addition to the individual) regarding the individual’s performance is often necessary. (DSM-IV, p. 7)
The essential rationale for the clinical significance criterion is that it "helps establish the threshold for the diagnosis of a disorder in those situations in which the symptomatic presentation by itself (particularly in its milder forms) is not inherently pathological." The target situations referred to in this passage appear to be those in which the symptoms satisfy DSM criteria but are so mild as to not be significantly harmful and, thus, in which the condition should not be considered a disorder. This kind of false positive might be called the "threshold problem." There is no reference to situations in which significantly distressing or impairing symptoms are present, but they are better understood as normal reactions to a difficult situation rather than a result of an internal dysfunction (analogous to DSM’s recognition that uncomplicated grief, despite its clearly significant symptoms, is not a disorder). Such false positives might be called the "dysfunction problem" (6, 25). Clearly, a solution to the threshold problem need not be the same as a solution to the dysfunction problem, and the solution to each of these problems could vary from diagnosis to diagnosis.
The key clinical significance criterion terms "clinically significant," "distress," and "impairment in social, occupational, or other important areas of functioning" are not defined or further clarified. Each of these terms has its own important ambiguities.
The critical phrase "clinically significant" is taken from DSM’s definition of mental disorder. That definition requires that a mental disorder be a "clinically significant" behavioral or psychological syndrome (DSM-IV, p. xxi). The intention—logically questionable, in retrospect, in a definition of mental disorder—was to exclude from the DSM listing those disorders that clinicians were unlikely to see or treat (e.g., caffeine dependence leading to occasional withdrawal symptoms) (30). In contrast, the clinical significance criterion’s use of the phrase "clinically significant" applies to distress or impairment. From the paragraph noted above (referring to a symptom threshold and to eliminating milder symptom presentations), it seems clear that "clinically significant" simply means a clinician’s judgment that the distress or impairment is significant, marked, or substantial in intensity or duration. Supporting this interpretation, Allen Frances, Chair of the DSM-IV Task Force, recently said of the clinical significance criterion, "This appeal to clinical judgment is a reminder to evaluate not only the presence of the symptoms in the criteria set, but also whether they are severe enough to constitute mental disorder" (31).
An alternative interpretation, based on the idea that the clinical significance criterion is supposed to distinguish what is inherently pathological from normality, is that "clinically significant" simply means "indicates disorder." Used this way, the clinical significance criterion is circular and offers no real guidance in deciding whether the level of distress or impairment is or is not sufficient to imply disorder. In effect, it just states that only disorders should be diagnosed as disorders. This problem has been acknowledged by Frances: "Unfortunately, this method of defining caseness contains the seeds of tautology; mental disorder is present only when there is ‘clinically significant’ impairment, but this determination is based on the clinician’s judgment" (31). Consequently, in the remainder of this article we interpret clinically significant distress or impairment to mean a level of severity in intensity, duration, or other relevant dimensions that is marked or substantial.
It is also unclear whether "distress" refers only to discomfort intrinsic to a symptom (e.g., the intrinsic discomfort of a panic attack) or also to distress about having a symptom (e.g., being upset about gaining weight). Apparently, both senses are intended, although it can be argued that only distress that is inherent in a condition is relevant to its status as a disorder—otherwise, marked distress about having a nondisordered condition may lead to a false positive diagnosis.
Finally, with respect to the impairment clause, what is meant by "other important areas of functioning"? If this is interpreted broadly, then any symptom might be construed as an impairment in some important area of functioning (e.g., insomnia being impairment in sleep regulation and weight loss being an impairment in appetite regulation). However, the apparently intended meaning—consistent with the reference in the clinical significance criterion paragraph to the difficulty of evaluating role functioning, and also consistent with examples provided in some criteria sets—is that the "other" functioning refers to the individual’s functioning in the environment (i.e., the performance of the individual) and not to the functioning of specific internal mechanisms, such as mood regulation and memory. (We note that even if "other functioning" is given a broader interpretation than the one apparently indicated by the text, it does not affect most of the problems with the clinical significance criterion that we mention below.)
The first sentence of the paragraph in DSM-IV explaining the clinical significance criterion states that "the definition of mental disorder in the introduction to DSM-IV requires that there be clinically significant impairment or distress." Thus, the clinical significance criterion is claimed merely to incorporate part of the definition of mental disorder into the diagnostic criteria sets. However, this is not the case. The relevant sentence from the DSM-IV definition of mental disorder is, "In DSM-IV, each of the mental disorders is conceptualized as a clinically significant behavioral or psychological syndrome or pattern that occurs in an individual and that is associated with present distress (e.g., a painful symptom) or disability (i.e., impairment in one or more important areas of functioning) or with a significantly increased risk of suffering death, pain, disability, or an important loss of freedom. (DSM-IV, p. xxi; emphasis added)
A careful reading of this long sentence indicates that the DSM definition recognizes, first, that there are harms other than distress and impairment that might be associated with having a mental disorder and, second, that disorder may be diagnosed in situations where a condition has not yet caused harm in the form of distress or impairment but is likely to do so in the future. Analogously, in physical medicine, a tumor may be asymptomatic but still be considered a disorder because with time it is likely to cause symptoms.
A second and more fundamental divergence between DSM’s definition of mental disorder and the clinical significance criterion is the different way in which disability (impairment) is conceptualized. In the definition of mental disorder, disability refers to impairment of any important areas of functioning, which could include either role functioning or biological functions (such as sleep, attention, or sexual arousal), whereas the clinical significance criterion, as noted above, seems to exclusively refer to social, occupational, or other role functioning. The clinical significance criterion thus imposes an impairment requirement that is much narrower than that required by the mental disorder definition. As we discuss below, this change leads to a problem of potential false negatives.
Most important, the heart of DSM’s definition is not addressed by the clinical significance criterion: "Whatever its original cause, it [the syndrome or pattern] must currently be considered a manifestation of a behavioral, psychological, or biological dysfunction in the individual" (DSM-IV, pp. xxi–xxii). As we will show, the failure of the clinical significance criterion to address the key concept of underlying dysfunction leads to its failure to eliminate false positives from several important categories of disorders.
The Clinical Significance Criterion Is Redundant With Symptom Criteria That Already Include Significant Impairment in Functioning
There are numerous DSM-IV diagnoses for which the symptom criteria already require that there be impairment in functioning at a level that is clearly clinically significant. For these categories, adding the clinical significance criterion has no effect on caseness and therefore eliminates no false positives. Its addition only makes the criteria more cumbersome and time consuming to apply. The following are examples.
The clinical significance criterion for conduct disorder reads as follows: "The disturbance in behavior causes clinically significant impairment in social, academic, or occupational functioning" (DSM-IV, p. 91). Given that the symptom criteria include items referring to aggression toward people, destruction of property, deceitfulness, theft, and serious violations of rules, it is clear that any youth whose behavior satisfies the symptom criteria for the disorder will also satisfy the clinical significance criterion by virtue of being significantly impaired in social functioning. False positives (e.g., antisocial behavior normative in a subculture ) would not be eliminated by the clinical significance criterion.
The symptom criterion is "consistent failure to speak in specific social situations (in which there is an expectation for speaking, e.g., at school) despite speaking in other situations" (DSM-IV, p. 115). To this DSM-IV adds the clinical significance criterion requirement that "the disturbance interferes with educational or occupational achievement or with social communication." However, consistent failure to speak in situations where speaking is expected constitutes an interference with social communication—rendering the clinical significance criterion redundant. Moreover, some potential false positives (e.g., a child who refuses to speak in school because she has been threatened by a bully ) are not eliminated.
The two primary symptom requirements for this disorder are "sudden, unexpected travel away from home or one’s customary place of work, with inability to recall one’s past" and "confusion about personal identity or assumption of a new identity (partial or complete)" (DSM-IV, p. 484). To this is added the clinical significance criterion requirement that the "symptoms cause clinically significant distress or impairment in social, occupational, or other important areas of functioning." It would certainly seem that wandering away from your social setting with loss of a sense of personal identity and memory of your past automatically constitutes impairment in social and other important areas of functioning.
The Clinical Significance Criterion Eliminates Potential False Positives That Could Have Been Addressed by Raising the Level of Distress Required by the Symptom Criteria
There are many DSM categories for which the symptom criteria imply considerable distress. One might think that in such cases, adding the clinical significance criterion is redundant because the distress requirement has already been satisfied. However, this is not so, because the clinical significance criterion requirement that the distress be clinically significant may set a higher threshold for diagnosis than the symptom criteria alone, thus possibly eliminating some false positives. However, this same improvement can be accomplished by adjusting the symptom criteria, thus avoiding the pitfalls of the undefined terms in the clinical significance criterion. The following are examples.
Separation anxiety disorder
The symptom criteria require at least three of eight symptoms, all but one of which are inherently distressing (e.g., "recurrent excessive distress" when one is separated from home or major attachment figures; "persistent and excessive worry" about losing attachment figures; "persistent and excessive worry" about being separated from attachment figures; and "repeated nightmares" about separation [DSM-IV, p. 113]). Although these criteria require some distress, the modifiers (e.g., "repeated," "recurrent," "persistent," "excessive") are consistent with a low level of distress (e.g., repeated nightmares that are not very upsetting or worry that is excessive relative to the actual degree of separation but is still moderate in absolute terms). Thus, it is conceivable that the symptom criteria might be met even in a case where the amount of distress does not warrant the attribution of disorder. Requiring clinically significant distress does eliminate such potential false positives. However, the same effect can be accomplished by slight modifications in the symptom criteria. For example, one might require "repeated and markedly distressing nightmares" and "recurrent excessive and marked distress" when separated from home or major attachment figures. Possibly, further improvement might be achieved if phrases such as "markedly distressing" were clarified or operationalized.
Generalized anxiety disorder
The symptom criteria require that there be "excessive anxiety and worry… about a number of events or activities" and that "the person finds it difficult to control the worry" (DSM-IV, p. 435). Again, worry about a number of events or activities might be excessive relative to the reality of the events and difficult to control, yet at such a low level of intensity that the minimal threshold for diagnosing disorder is not reached. The clinical significance criterion, by requiring clinically significant distress, does solve this problem. However, this threshold problem can equally be resolved simply by requiring that the anxiety or worry be markedly distressing.
Posttraumatic stress disorder
The symptom criteria require multiple symptoms from three areas: reexperiencing the traumatic event, avoidance of stimuli associated with the event and numbing of general responsiveness, and persistent symptoms of increased arousal (DSM-IV, pp. 427–429). Many of these symptoms explicitly require some level of distress (e.g., "recurrent and intrusive distressing recollections" and "recurrent distressing dreams"). Again, while the clinical significance criterion successfully eliminates some potential false positives due to minimal distress, the problem could be solved by merely requiring that the distress be marked or (as is actually used in criterion B4) "intense."
Although modifications in the symptom criteria can raise the threshold for diagnosis just as the clinical significance criterion does, it might be argued that there are some cases in which it is preferable to allow the clinician to judge whether the overall level of distress is significant, even if the distress associated with several individual symptoms is not marked. If such a two-stage diagnostic procedure (assessing first the presence of a number of symptoms and second the level of overall distress from the combined symptoms) has advantages over modifying distress thresholds of individual symptom criteria, this has not been demonstrated. The two-stage procedure used by the clinical significance criterion has the unfortunate effect of substituting for the hard work of improving the diagnostic criteria sets so that each criterion is at a level of clinical significance.
We have focused here on threshold problems with distress. Other kinds of symptom criteria (e.g., "difficulty falling or staying asleep" and "difficulty concentrating" from the posttraumatic stress disorder criteria) can run into the same kinds of threshold problems. If so, these problems can also be addressed by modifications in the symptom criteria (e.g., "marked difficulty falling or staying asleep" and "marked difficulty concentrating"). Other methods for raising diagnostic thresholds, such as duration, number, and kind of symptoms, should also be explored—recognizing the danger of increasing false negatives.
The Clinical Significance Criterion Increases False Negatives
A further problem is the potential of the clinical significance criterion for yielding false negative misdiagnoses. The following are examples.
The symptom criteria require both motor and vocal tics that occur many times a day, nearly every day, or intermittently for a longer period (DSM-IV, p. 103). With the addition of the clinical significance criterion, the diagnosis is now not given to a child who is neither markedly distressed nor significantly impaired in functioning by the tics. Problems of potential false negatives due to this change were noted by members of the Medical Advisory Board of the Tourette Syndrome Association in a letter to the Journal of the American Academy of Child and Adolescent Psychiatry:
"Marked distress" and "significant impairment" are not defined. Many young children with tics deny them most of the time; others (especially some teenagers) have no tolerance for even a few tics a day. Distress may be short-lived, based upon misunderstanding or novelty effects, or on the presence of one specific, more socially distressing tic. If distress is shown, for how long a period is this necessary to meet the criterion, and is this a once-and-for-all requirement that has been met?…The DSM-IV committee’s intention to clarify the diagnostic boundaries of Tourette’s Disorder was laudable, but the implications were inadequately thought through and there was no opportunity for discussion and feedback with the professional community. (33)
Admittedly, it is difficult to know how to draw a boundary between variations in normative sexual fantasies and behavior and clearly pathological sexual behavior. DSM-III-R attempted to do this by requiring, for all of the paraphilias, that "the person has acted on these urges, or is markedly distressed by them" (DSM-III-R, pp. 279–290). Thus, according to DSM-III-R, a paraphilic disorder is present if the individual acts on the paraphilic impulses even if not distressed by the impulses. (Whether this permits its own false positives is not considered here.) DSM-IV replaced this criterion in all of the paraphilia criterion sets with the clinical significance criterion, which states that "the fantasies, sexual urges, or behaviors cause clinically significant distress or impairment in social, occupational, or other important areas of functioning" (DSM-IV, pp. 522–532).
The addition of the clinical significance criterion to the criterion sets for the paraphilias appeared to some to exclude from diagnosis individuals who engage in repetitive paraphilic behavior (e.g., pedophilia, exhibitionism) as long as the behavior does not cause distress or role impairment, thus giving rise to false negatives. Not surprisingly, this perception triggered angry reaction from outside the profession. On March 22, 1995, the National Law Center for Children and Families (Fairfax, Va.) issued a press release claiming that the new criteria had "normalized" pedophilia and noted that the change had "provoked outrage among public interest groups that seek to end child sexual abuse." In response, the American Psychiatric Association (APA) issued a fact sheet on pedophilia attempting to explain the DSM-IV approach to diagnosing the disorder. It said of the clinical significance criterion, "Under this criterion, any sexual encounter with a child constitutes ‘clinical significance’ and warrants a diagnosis of pedophilia…. By definition, acting on pedophiliac urges is considered to be impairment in functioning" (34). APA’s interpretation of the application of the clinical significance criterion to the paraphilias remains controversial and may seem to stretch the meaning of terms in an ad hoc manner. The lesson of this controversy seems to be that some reasonable interpretations of the ambiguities in the clinical significance criterion would lead to important false negatives in the paraphilias.
The clinical significance criterion for all of the sexual dysfunctions in DSM-IV states that "the disturbance causes marked distress or interpersonal difficulty" (DSM-IV, pp. 493–518). This requirement potentially gives rise to false negatives for many sexual dysfunctions. For example, the symptom criteria for the diagnosis of male erectile disorder require "persistent or recurrent inability to attain, or to maintain until completion of the sexual activity, an adequate erection" (DSM-IV, p. 504). We would argue that this condition is a disorder even if the individual is not distressed by it and it causes no interpersonal difficulty. The clinical significance criterion implies that if a man who cannot sustain an erection sufficient to engage in intercourse can convince himself and his partner not to care about it, then he has no disorder.
To take one further example, vaginismus requires "recurrent or persistent involuntary spasm of the musculature of the outer third of the vagina that interferes with sexual intercourse" (DSM-IV, p. 515). Clearly, this ought to justify a diagnosis, but DSM-IV adds the additional requirement (clinical significance criterion) that "the disturbance causes marked distress or interpersonal difficulty."
Of course, the clinician may choose not to treat a sexual dysfunction unless it causes distress or interpersonal difficulties. However, we believe it makes no sense to claim that because such conditions do not meet the clinical significance criterion, they are not true disorders.
The clinical significance criterion for substance dependence appears not as a separate criterion but in the introduction to the symptom description: "A maladaptive pattern of substance use, leading to clinically significant impairment or distress, as manifested by…" (DSM-IV, p. 181). The requirement that the substance use cause clinically significant impairment or distress could yield false negatives. It is common to encounter individuals who have lost control over their drug use and are suffering various harms (e.g., threat to health) as a result (and who therefore, to us, have a disorder) but who are not distressed and who can carry on successful role functioning. Consider, for example, the case of a successful stockbroker who is addicted to cocaine at a level that is threatening his physical health but who has no distress and whose role performance has not suffered. Without the clinical significance criterion, the DSM-IV criteria correctly classify the individual’s condition as a substance dependence disorder. Applying the DSM-IV clinical significance criterion, it is not a disorder.
The Clinical Significance Criterion and Major Depression
We consider the clinical significance criterion for major depressive episode separately to illustrate how the problems described above converge in one diagnosis. The symptom criteria require that at least five of nine symptoms be present nearly every day for 2 weeks (DSM-IV, p. 327). Since most of these symptoms are either intrinsically distressing (e.g., depressed mood, psychomotor agitation, fatigue) or are almost invariably accompanied by distress about having the symptom (e.g., diminished interest or pleasure, weight loss or gain, hypersomnia, psychomotor retardation, thoughts of death), it is highly unlikely that one could satisfy the criteria and not be significantly distressed. Thus, if distress and impairment are interpreted broadly, the clinical significance criterion is pragmatically redundant. (Indeed, Dr. John Rush, chair of the DSM-IV committee on mood disorders, could not recall a case in which an individual met the symptom criteria for major depression and had neither clinically significant distress nor impairment [J. Rush, personal communication, June 5, 1998].) However, if the clinical significance criterion is interpreted more narrowly, false negatives become a problem. An individual who has marked loss of interest and pleasure (not depressed mood) with five of the nine symptoms (e.g., hypersomnia, weight gain) present to a marked degree nearly every day for 2 weeks might be neither significantly distressed nor role-impaired. Such an individual may still have a depressive disorder. Finally, the clinical significance criterion may eliminate some false positives in which symptoms are so minimal that no diagnosis is warranted. However, this could be accomplished by slightly modifying the symptom criteria.
A degree of interpretation is involved in deciding whether a condition causes significant distress or impairment (35). To the degree that the clinical significance criterion is interpreted broadly, it becomes redundant with the symptom criteria; to the degree that it is interpreted narrowly as requiring an additional test (such as role impairment or the patient being distressed by the condition), the criterion gives rise to potential false negatives.
We have noted the many ways in which the DSM-IV clinical significance criterion often does not do what it is designed to do—minimize false positives. The reason it often fails is that it is based on the assumption that the way to determine that a condition is pathological is to ensure that it causes sufficient distress or impairment in social or role functioning. In the rest of medicine, a harmful condition is considered pathological if there is evidence of a biological dysfunction in the organism. Neither distress nor role functioning failure is necessary to make most medical diagnoses, although both often accompany severe forms of disorder. For example, a diagnosis of pneumonia, heart disease, cancer, or innumerable other physical disorders can be made in the absence of subjective distress and even if the individual is successfully functioning in all social roles.
Of course, identifying what has gone wrong (the dysfunction) in mental disorders is often much more difficult than it is in the rest of medicine, given our lack of knowledge about underlying psychological and biological mechanisms. Moreover, it is true that mental mechanisms and functions are sometimes directly tied to social role functioning in a way that is often not true of physical mechanisms. For example, personality disorders, by their very nature, involve disturbances in interpersonal functioning, and in such cases role impairment may indeed be a required indicator of the pathology. However, as the examples above show, it makes no sense to impose such a requirement in the many cases in which the underlying dysfunction is not inherently tied to social role functioning. Mechanisms involving sleep regulation, mood regulation, sexual functioning, and many others can go wrong and be disordered without necessarily impairing role functioning.
Perhaps the most troubling aspect of the DSM-IV clinical significance criterion is that it misdiagnoses the underlying problem that leads to false positives, giving the misimpression that the problem has been addressed. Elsewhere (6, 25), one of us has argued that the false positive problem is often due to a failure of symptom criteria to indicate an underlying dysfunction. For example, extreme sadness and associated symptoms of depression can be due to a normal reaction to extreme loss (not a disorder) or to a malfunction of mood-regulating mechanisms (disorder). The same symptom picture may be present in either of these possibilities. Thus, focusing on distress and impairment in themselves is not sufficient in making a valid distinction between disorder and no disorder. What makes a condition a disorder is often not the symptoms alone but the fact that the symptoms are caused by a dysfunction, which can only be inferred from a careful assessment of the nature, course, and context of the symptoms.
DSM-IV acknowledges this line of thinking in the text discussion for one category, conduct disorder: "The 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" (DSM-IV, p. 88). However, this distinction between normal and disordered antisocial behavior in a child or adolescent is not reflected in the diagnostic criteria. The clinical significance criterion for conduct disorder is of no help in excluding a child from receiving the diagnosis when the child’s symptoms (e.g., running away from home, staying out at night despite parental prohibitions, and frequent lying) are clearly a normal response to an abusive and threatening family situation. In effect, DSM-IV’s textual cautionary note that conduct disorder should not be diagnosed if symptoms are not due to a dysfunction is an acknowledgment that the clinical significance criterion does not solve the false positives problem for this diagnosis.
Psychosocial stress can certainly trigger mental disorders. However, beginning with DSM-III and continuing through DSM-IV, the need to exclude normal responses to loss when diagnosing major depression is acknowledged, but only for bereavement. Parity of reasoning suggests that the same exclusion for major depression perhaps should apply to other nonpathological sadness responses to severe personal losses (e.g., diagnosis of terminal illness in oneself or a loved one, being abandoned by a love object). One might consider broadening the bereavement exclusion criterion with something like the following: "The symptoms are not better accounted for by a normal reaction to a psychosocial stressor (e.g., loss of a loved one, terminal medical illness in self or loved one, loss of relationship); i.e., the disturbance is judged to be markedly excessive in intensity or duration relative to the nature of the stressor." The same kind of exclusion criterion could be applied to many other DSM-IV categories where the symptomatic clinical picture may represent a normal reaction to psychosocial stress, such as dysthymic disorder, oppositional defiant disorder, generalized anxiety disorder, and primary insomnia.
This approach to solving the dysfunction false positives problem has its own ambiguities and pitfalls. It requires a clinical judgment about whether the disturbance is better accounted for as a normal reaction to psychosocial stress—a judgment with unknown reliability and validity. Certainly, a better understanding of normal reactions to stress would be helpful in making such judgments. However, the idea that the clinician must judge whether the reaction is proportional to the nature of the contextual stressor is already incorporated into other DSM diagnoses, such as adjustment disorder and body dysmorphic disorder. How to appropriately include context in the diagnostic criteria, without increasing false negatives, is a challenge that requires further work. Whether it is ultimately the best approach to solving the false positives problem—and specifically the dysfunction false positives problem, which is not addressed by the clinical significance criterion—remains to be demonstrated.
DSM-III, by providing diagnostic criteria for all of its mental disorders, led to a revolution in psychiatric epidemiology because it allowed lay interviewers to administer structured diagnostic interviews that operationalized the DSM-III criteria. The DSM-IV clinical significance criterion calls into question the validity of this attempt and thus has unnoticed but dramatic implications. If diagnostic criteria cannot adequately operationalize the distinction between disorder and nondisorder without resort to the undefined concept of "clinically significant," then one can have little confidence in epidemiologic surveys conducted by lay interviewers. Frances, in a commentary on "Problems in Defining Clinical Significance in Epidemiological Studies" (31), has noted, "Lay interviewers do not have clinical experience and cannot be expected to make judgments about clinical significance." The clear implication of his remark is that if criteria sets include a requirement of clinical significance that is not further operationalized, then lay interview studies of community prevalence cannot be validly carried out. In the absence of any data supporting this claim, it seems to us to be premature to suggest that psychiatric epidemiology as we now know it should be abandoned.
Although we have argued against using impairment in role functioning as a general requirement for disorder, we acknowledge that for certain purposes, such as determining the need for services, adding impairment criteria to diagnostic criteria might be helpful. However, this should not be confused with identifying disorders (36). For example, it is one thing to identify the prevalence of low back syndrome (of varying degrees of severity), and another thing to identify the prevalence of the disorder when it is so severe that it interferes with social, occupational, or other role functioning. In a child psychiatric epidemiologic study (37), prevalence rates based on symptom criteria alone were so high that researchers added impairment criteria. As might be expected, inclusion of such criteria increased the association of the diagnoses with service utilization, but their validity as indicators of disorder remains undemonstrated.
We recognize that many current criteria sets may be inadequate and yield false positives. What is needed is serious work (both conceptual and empirical) to identify the specific inadequacies of each criteria set and to either formulate specific improvements in existing criteria (e.g., raising thresholds) or add new criteria (e.g., taking into account the psychosocial context as a way of distinguishing disorder from nondisorder). Adding a generic criterion with an undefined reference to clinical significance undermines the point of having operationalized criteria in the first place.
In conclusion, the DSM-IV clinical significance criterion was an attempt to deal with a real and difficult problem with the DSM diagnostic criteria—the potential for false positive diagnoses. However, its wholesale application in DSM-IV is problematic and has no empirical support. For many diagnoses it is redundant, since the symptom criteria already imply significant impairment in functioning. For those diagnoses in which it is potentially helpful by raising the threshold of required distress, the same result could be more directly obtained by modifying the wording of the symptom criteria. For several important diagnoses, the addition of the clinical significance criterion potentially increases false negatives. Most important, the clinical significance criterion does not address perhaps the major source of false positives—failure of the criteria to indicate a dysfunction. We hope that in the process of revising DSM-IV, the generic application of the clinical significance criterion will be reconsidered. We suggest that serious attention be given to alternative approaches to solving the false positives problem that tailor the solution to the nature of each diagnostic category.
Received Dec. 7, 1998; revision received May 7, 1999; accepted May 17, 1999. From the Department of Psychiatry, Columbia University and the New York State Psychiatric Institute; and the Institute for Health, Health Care Policy, and Aging Research and the School of Social Work, Rutgers—the State University of New Jersey, New Brunswick. Address reprint requests to Dr. Spitzer, Biometrics Research, New York State Psychiatric Institute, Unit 60, 1051 Riverside Dr., New York, NY 10032; firstname.lastname@example.org (e-mail). The authors thank Drs. Michael B. First, Randall Marshall, Thomas Widiger, Donald Klein, and Ronald Rieder for comments on an earlier version of this article.