Assignment
Paris, J. (2015). The intelligent clinician's guide to the DSM-5 (2nd ed.). New York, NY: Oxford University Press.
The intelligent clinician's guide to the DSM-5
Chapter 6 Dimensionality
Page 84
Science can describe phenomena using names or numbers. Categories are qualitative, and dimensions are quantitative. Some things in nature, such as the periodic table of elements, neatly fit a categorical paradigm. Others either have fuzzy edges or do not fit into categories at all. This problem has led to debates about the number of planets in the solar system, the hierarchical classification of biological species, the difference between life and nonlife, and the nature of subatomic particles. Most of these disputes depend on where to draw a line to define a category. Many things are better described on a continuum. Yet cognitive science consistently shows that people prefer to think in categories (Rosch & Lloyd, 1978). Medicine has always classified illness in that way. When experienced physicians assess a new patient, they do not go through an extensive checklist of signs and symptoms to make a diagnosis. Instead, they take one look at a patient, ask a few key questions, and rapidly develop a hypothesis and test it through further history taking and examination procedures (Groopman, 2007). Thus, categorical diagnoses correspond to what happens in a clinical encounter. Categories describe typicality in a clinical picture, are the basis of differential diagnosis, are practical for communication, and help determine treatment planning. DSM-5 accepts categories but views them as artificial, proposing scoring procedures, wherever possible, to turn names into numbers. These scores are described as “dimensions,” a geometrical metaphor rooted in a quantitative approach to psychopathology. The uncertainty of psychiatric diagnosis supports a suspicion of categories, given that most consist of a set of symptoms that may or may not
Page 85
correspond to a coherent pathological entity. We sometimes forget that even if diagnoses are useful guideposts, they do not correspond to Platonic reality. Yet categories become reified with time (Hyman, 2010) and are seen as if they were just as real as medical conditions with known biological markers. Major depression is a prominent example—a set of symptoms that may or may not correspond to a coherent diagnostic entity associated with a unique endophenotype. Yet despite teaching psychiatry for many years, I have been unable to convince my students that this diagnosis fails to provide specific guidance for clinical management. For most trainees, major depression closes the case for writing a prescription. Major depression fades imperceptibly into the sadness that everyone feels from time to time (Horwitz & Wakefield, 2007). Using a cutoff, as in DSM’s category of major depression, of five out of nine criteria (as opposed to, for example, seven or eight out of nine) is an arbitrary procedure. That is why this category is so heterogeneous. In addition, patients with subclinical symptoms (less than five) still suffer distress. It is also not clear that patients who have more than five criteria are fundamentally different from those who only barely meet the cutoff. The number 5, chosen because it is more than half of 9, is arbitrary rather than being based on scientific data. Different levels of severity, or even separate disease processes, are obscured by a single category. Despite its ubiquity, major depression is one of the most problematic diagnoses in psychiatry. Categorical classification imposes an artificial order when boundaries between disorders (or between disorders and normality) are fuzzy. Diagnostic categories are always dichotomous, allowing only for a yes or no decision. Those who do not quite meet criteria will be excluded, and patients with different levels of severity can get the same diagnosis. This process leads to a loss of information (Krueger et al., 2011b). Categories are the basis of differential diagnosis, but that procedure makes sense only when disease mechanisms are known. When they are not understood, as is almost always the case in psychiatry, differential diagnosis is little but guesswork. It should be kept
Page 86
in mind that the most important illness categories in current use, including well-researched diagnoses such as schizophrenia, may or may not meet the test of time. The problems of diagnostic categories in psychiatry are many, including an inadequate scientific base, excessive comorbidity, inadequate coverage, an arbitrary boundary with normality, and heterogeneity among persons sharing the same diagnosis (Kraemer et al., 2004). Comorbidity, when massive, is a good sign that a category is not valid. Inadequate coverage is reflected by the fact many patients do not fit into the categories listed in the system. The “not otherwise specified” (NOS) option (now “other specified” or “unspecified”) introduced in DSM-III was necessary because so many patients do not meet criteria for the specified diagnoses. It is therefore used as a “wastebasket” for patients who do not meet criteria for any category within a group. In personality disorders, 50% of cases could only be fitted into a NOS category in DSM-IV (Zimmerman et al., 2005). Finally, categories obscure clinically important differences between patients meeting criteria for the same diagnosis. These are serious problems requiring a serious solution. An overlap between categories leads patients to meet criteria for multiple diagnoses. This “comorbidity” reflects the absence of hierarchical rules: The DSM-IV-TR system did not generally allow clinicians to determine whether one diagnosis is primary and one secondary. Multiple diagnoses are usually a marker for severity. A community survey in the United Kingdom (Weich et al., 2011) found that those who meet criteria for several mental disorders have more severe levels of dysfunction. The problem with comorbidity is that we have no definite way to determine whether one of two overlapping categories (e.g., mood disorder vs. anxiety disorder, or conduct disorder vs. attention-deficit disorder) should take precedence over the other. If you follow the rules, you are forced to make more than one diagnosis. Comorbidity would be greatly reduced if DSM rewrote diagnostic criteria to minimize overlap. But lacking a clear justification, previous editions of the manual have been reluctant to do this. DSM-5’s view was that comorbidity might be reduced or eliminated by replacing categories with dimensions.
Page 87
A second problem concerns the boundary between pathology and normality. Many clinical symptoms blend into subclinical phenomena that can be seen in large sections of community populations. In other words, most people have periods when they are moody, unhappy, nervous, or show features of addiction. Many have suffered from some of the symptoms of common mental disorders. Even hallucinations sometimes occur in normal people (Stip & Letourneau, 2009). Again, there is no definite cutoff point at which mental disorder can be said to be clearly present. A third issue is that neo-Kraepelinian categories in psychiatry rarely have robust relationships to biological measures, such as genetic markers, hormonal variations, or neuroimaging. By and large, relationships are stronger with dimensional scales that measure traits or score signs and symptoms (Regier et al., 2011)— although even then correlations are not high. The problem with diagnostic categories is that they are dichotomous, allowing only for a yes or no decision. Those who do not quite meet criteria will be excluded, and patients with different levels of severity can receive the same diagnosis. In summary, psychiatric diagnosis has an inadequate scientific base, including massive comorbidity, inadequate coverage leading many patients to fit into only unspecified options, while categories obscure clinically important differences between patients meeting criteria for the same diagnosis (Kraemer et al., 2004). These difficulties have led some to the conclusion that categorical diagnosis in psychiatry should either be scrapped or kept only as a short-term expedient (Kupfer & Regier, 2011). If diagnosis would gradually become dimensional, instead of slotting patients into rigid categories, one could give them a score on one or more dimensions of psychopathology. Rather than packaging pathology within a single category, multiple scores could make overlap disappear. All mental disorders would be seen as lying on one or several spectra reflecting neurobiological variability (Insel et al., 2010). This is the basis of the “research domain criteria” (RDoCs) proposed by the National Institute of Mental Health (NIM
Page 88
(Insel, 2009). This model overlaps with the point of view of DSM-5, but it pushes the envelope. The basic assumption is that there is no absolute boundary between disorder and normality and that psychopathology is better described by a series of scores. But although RDoCs claims to be a more scientific system, it remains to be seen whether this model corresponds to empirical reality.
What Do Dimensions Measure?
Despite all their virtues, dimensions cannot untie the Gordian knot of validity in diagnostic classification. The reason is that dimensions suffer from most of the same limitations that afflict categorical systems. They are not based on a fundamental understanding of the etiology and pathogenesis of mental disorders. They are based just as much on clinical observation as on categorical diagnoses. Until we know more about psychopathology, no system can be further advanced than medical diagnosis was in the nineteenth century, prior to the development of blood tests and X-rays. Diagnosis must be rooted in independent markers, most of which will have to be biological. That would be true construct validity. At this point, psychiatry can only aspire to having such measures. Converting clinical observations into dimensional scores will not solve the problem. Dimensions may do a better job of measuring severity. It has been consistently shown that severe effects on functioning provide a better predictor of prognosis than categorization (Krueger & Bezdjian, 2009). However, the scoring of severity in psychiatry is not like the staging of tumors based on imaging and pathological findings. Rather, the procedure is entirely rooted in observation and/or self-report data. If severity ratings are little but a count of signs and symptoms, their introduction could be premature (First, 2005). In short, observable phenomena provide only indirect clues to underlying endophenotypes—the pathological processes that lie behind signs and symptoms. Until psychiatry understands mental illness better, the use of dimensions will only be a rough-and-ready expedient.
Page 89
Dimensions and Clinical Utility
A system designed for practitioners as well as researchers has to have clinical utility. If dimensional diagnosis ends up not being used, for lack of clinical utility, adopting it would be pointless (First, 2010). Although taking blood pressure is simple, that is an objective physical measurement using a standard machine. Physicians learn how to perform complex diagnostic procedures (e.g., electrocardiograms) that can make useful predictions about the nature and course of disease. But there is no equivalent procedure in psychiatry. The scoring of symptoms by practitioners, as recommended in DSM-5, has limited utility. Clinicians would need a fair amount of training to make these ratings valid. Busy practitioners are likely to produce unreliable results. This procedure is also different from administering self-report questionnaires, which have been developed over years to establish their psychometric properties and that are the basis of all dimensional research in psychopathology. Such scales do not require reliable rating and can be backed up by systematic testing. Research psychologists like these measures—they are used to deal with data in that way—and have long developed assessment instruments using dimensional approaches. But it is usually impractical to ask patients to fill out questionnaires, even in a waiting room. Even when they do, the answers may not correspond to what clinicians observe. Crucially, although dimensions give an impression of being more “scientific,” they have not been shown to be valid in relation to etiology, pathogenesis, outcome, or choice of treatment. Scores based on symptoms do not measure underlying processes behind disorders. At this point, the whole idea has to be taken on faith. Until more data become available, it seems futile to ask clinicians, who already found DSM-IV to be burdensome, to learn an even more complex system. Most will find dimensional scoring unwieldy. The scoring of symptoms by practitioners requires using a Likert scale (usually rated from 1 to 5) that clinicians need to be
Page 90
trained to use. But resistance on the part of clinicians to using dimensional scales is not based on habit or laziness (First, 2005). They know that scoring actually tells them no more than what they can readily observe. Why go through a complex procedure when an experienced clinician can know what is wrong in the first 5 minutes? Currently, only a few dimensional measures have been used widely. Although scoring systems have been recommended for personality disorders (Costa & Widiger, 2001), for psychoses (Rosenman et al., 2003), and for depression (Korszun et al., 2004), none of the existing scales has a strong clinical following. No one has been convinced that scoring tells you more about outcome and treatment than a categorical diagnosis. Again, dimensional diagnosis tends to deny the existence of a separation between pathology and normality. Although disorder can sometimes be a matter of degree, if one were to follow dimensionality to its logical conclusion, everyone would merit a score reflecting some level of active or potential condition. Physicians in general medicine do sometimes think in this way, as shown by the widespread measurement of blood lipids in asymptomatic patients. But one need not confuse risk factors with disease. Unless we want to give up diagnosis entirely, we still need to establish cutoff points to distinguish true cases of mental disorder from subclinical or normal variants. Medicine is the study of disease, not a description of normal variation.
Dimensions and Research
A former director of NIMH (Hyman, 2011) reported that when he was in charge, millions of dollars were spent on genetic studies of DSM categories, but that the money was almost entirely wasted. The reason was that psychiatric diagnoses are not true endophenotypes (Regier et al., 2009). In this way, dimensionalization seems more in accord with modern genetics (Hyman, 2010).
Page 91
Hyman is correct about the danger of reifying categories, but dimensions also lack empirical support. The decision of the current leaders of NIMH to direct funding away from existing DSM categories to a future “brain-based” system (RDoCs; Insel, 2009; Insel et al., 2010; Sanislow et al., 2010) will probably be a failure. Although scoring could be more closely related to endophenotypes, genes, and neural circuits than traditional categories, we do not know how to do this in a valid way. To be fair, RDoCs is an interesting idea that should be researched systematically. But at this point, it has to be considered as a bet that may or may not work out. NIMH has become a center of neuroscience research, and it downplays psychosocial research. The principle that “mental disorders are brain disorders” appeals to those who want to reduce stigma. However, this phrase has become a mantra. It is not a fact but, rather, an ideology used to validate a certain approach to psychiatry. It represents the hope that mental illness can be translated into neuroscience without considering psychology or any of the other social sciences. It has even been suggested that psychiatry should reunite with neurology into one specialty that treats brain diseases (Insel & Quirion, 2005). (A neurological colleague of mine once suggested to me: “We treat the axon and you treat the synapse.”) The question is whether psychiatry can be reduced to the clinical application of neuroscience and whether clinicians should give up talking for prescribing. Needless to say, neuroscience is a valuable tool for our specialty. But studying mental processes on their own terms is an equally valid strategy. Mind is an emergent property of the brain and will never be fully explained on a cellular or molecular level. In other words, complex systems yield phenomena that “emerge” from simpler components but are not fully determined by them. In the simplest example, the properties of water are not explained by the atomic structure of hydrogen or oxygen. Although reducing a system to its components sheds light on what can be directly observed, explanations
Page 92
of mental disorder should ideally require links between different levels of analysis, from genes and neurons to psychosocial processes (Gold, 2009). Thus, the holy grail of neuroscience may never provide a complete resolution of the problems raised by diagnostic classification. Great advances in research in recent decades have taught us much about the brain. Imaging studies have identified specific functions for many cortical and subcortical structures. We have learned about the alarm system in the amygdala, the memory system in the hippocampus, the reward system in the nucleus accumbens, and the behavioral control system in the prefrontal cortex. At the same time, biochemical and physiological studies have defined synaptic pathways by which neurons communicate, as well as cellular pathways by which proteins are constructed and used within a neuron. Research may eventually specify the genes and proteins that shape all these processes. One can only applaud these scientific advances. However, none of this research has thus far had any clinical application—either to the understanding of disease mechanisms or to the treatment of mental illness. Molecular genetics and neuroimaging have not explained why people become psychotic or severely depressed. It remains possible that future editions of DSM will be guided by helpful neurobiological markers—but not this time around. This gap between basic and clinical science is no accident. Mental processes are too complex to be readily reduced to neuroscience. Moreover, brain mechanisms are only one of several relevant levels of analysis, and a comprehensive theory also needs to include psychological mechanisms. Biological reductionism in psychiatry supports a practice based almost entirely on drug treatment. In the widely cited article by Insel and Quirion (2005), the word “psychological” does not even appear. This mindless approach to psychiatry, supported by both DSM-5 and NIMH, has real consequences for practice. It supports the current tendency for psychiatry to abandon psychosocial models and to restrict itself to neuroscience.
Page 93
Dimensionalizing by Scoring Symptoms
DSM-5 could be used to dimensionalize diagnoses depending on severity by scoring the number of symptoms present in any patient. Thus, if there are nine criteria for a diagnosis and each is rated simply “yes,” “maybe,” or “no,” one would immediately have a scale ranging from 0 to 27. Alternatively, clinicians could score each criterion on a 4-point scale (e.g., not at all, some, a fair amount, or a lot). These procedures could be applied to any diagnosis. Again, it should be kept in mind that severity scores do not produce dimensions in the same way as scores on a test. There is no way of knowing how to weigh each criterion. Psychometric analysis of questionnaires requires that a large number of items be analyzed to ensure that the instrument as a whole produces a reliable result. Scales have to be smoothly continuous, not bumpy. Reaching that goal usually requires years of research. For this reason, scoring existing diagnostic criteria creates a chimerical beast that is neither fully categorical nor properly dimensional. Although the criteria in the manual can be useful in the aggregate, hardly any of them have validity on their own. Many reflect symptoms that resemble each other and that often occur together. This is why even sophisticated statistical methods (e.g., factor analysis) cannot “carve nature at its joints.” Ideally, a valid dimensional scale would need to be created de novo and be based on sources of external validity, not on existing criteria. DSM-5 does not have the evidence to support such a procedure or the time to develop it.
Diagnostic Spectra
Overlapping diagnoses that reflect different aspects of the same pathological process may fall within spectra, a range of disorders rooted in common mechanisms. The oldest and best supported is the schizophrenia spectrum (Siever & Davis, 1991). This concept
Page 94
includes conditions ranging from severe psychosis (schizophrenia itself) and milder psychoses (brief psychotic episodes and delusional disorder) to diagnoses in which formal delusions and hallucinations are absent (schizotypal personality disorder). The validity of this concept is supported by family studies, in which spectrum disorders show a stronger pattern of inheritance than schizophrenia alone (Kendler et al., 1994). It has also long been known that schizotypal personality shares biological markers with overt schizophrenia, even if they are not specific enough to be used for diagnosis (Siever, 2007). The concept of a mood spectrum also has a degree of validity. Decades ago, the American psychiatrist George Winokur suggested that depression and alcoholism might reflect the same disease process—with the former being more common in women and the latter being more common in men (Winokur et al., 1975). Later, the Swiss psychiatrist Jules Angst proposed a more restricted spectrum, including classical depression and its subclinical variants, as well as conditions sharing the same endophenotype, particularly anxiety disorders (Angst & Merikangas, 1997). Other spectra have been proposed for panic/agoraphobia, substance use, psychosis as a whole, anorexia–bulimia, an obsessive– compulsive domain, and social anxiety (Frank et al., 2011). A bipolar spectrum (discussed in Chapter 9) would bring together all disorders in psychiatry in which mood instability is prominent (Ghaemi et al., 2002). Another concept is an impulsive spectrum (Zanarini, 1993), based on evidence that impulsive traits—present in substance abuse, eating disorders, and personality disorders—run in the same families. Spectra could be used to search for endophenotypes (biological pathways) that lie behind symptoms (Gottesman & Gould, 2003). But like dimensions, they all still depend on phenomenological observation. In the absence of biological markers, disorders may not fall within the same spectrum just because they resemble each other. Also, disorders may be based on multiple endophenotypes, in which case basing spectra on clinical features alone oversimplifies a complex problem.
Page 95
How many diagnostic spectra would we need to describe most of the disorders listed in the DSM manual? A seminal study using factor analysis (Krueger, 1999) suggested that most of the territory could be covered by just two factors—internalizing and externalizing disorders. However, these dimensions do not describe cognitive impairment, a central issue for psychiatry. Kotov et al. (2011) suggested that five dimensions would do the trick: internalizing, externalizing, cognitive, somatoform, and antagonism. Thus, internalizing dimensions would describe most anxiety and mood disorders; externalizing would describe substance use and impulsive disorders; cognitive would describe schizophrenia, neurodevelopmental disorders, and neurocognitive disorders; somatoform would describe somatic symptom disorders; and antagonism would account for many personality disorders. There may be something useful to be garnered from these factor analyses. But keep in mind that they only describe symptomatic resemblances, and they tell you nothing about etiology or pathogenesis.
Dimensional Assessments in DSM-5
Several dimensional assessment measures can be found in Section III of the manual. One consists of a set of disorder-specific measures of severity, with details provided concerning “Clinician-Rated Dimensions of Psychosis Symptom Severity.” Finally, DSM-5 has a series of “cross-cutting” symptom assessments that describe psychopathology dimensionally. They have 13 domains, including standard clinical problems such as substance use, suicidal ideation, and psychosis, each scored by severity. These scales have not been the focus of much research. However, their reliability was tested in field trials, and kappas were better than most of those proposed for common categories (Regier et al., 2013), probably because it is easier to determine whether a patient falls within a symptom spectrum than to make a precise diagnosis. These spectra need much more work and should be thought of as a work in progress.
Page 96
At one point, it was proposed that DSM-5 should have its own scale for predicting suicide. For some very good reasons, this cannot be found in the final version of the manual. Every clinician is trained to assess suicide risk using a melange of indicators, such as stated intent, past suicidal behavior, psychosocial status, the presence of substance abuse, family history, and clinical diagnosis (Bongar, 1992). However, the prediction of suicide has mostly been marked by failure (Paris, 2006). Empirical studies show, at best, a small relationship between predictors and outcome. Large-scale research using algorithms based on commonly applied clinical judgment failed to predict a single case (Goldstein et al., 1991). The sober fact is that psychiatrists are unable to determine which patients are at risk for taking their own lives. If they could, they might be in a position to prevent suicide. Why is this so? The reason is that although suicidal ideation is very common, and attempts are not infrequent, completed suicide is relatively rare. Because of this “base rate” problem, most methods of predicting suicide turn up an enormous number of false positives. Even the most successful predictors, such as the Suicide Intent Scale (Beck et al., 1974), can only predict completion with statistical significance (Suominen et al., 2004). Even if one sees relationships in large samples, the scale will be wrong most of the time in practice. This is why scores based on statistical data are not accurate enough to be useful in making clinical decisions. Although long-term follow-up studies have found that scales can have some predictive value, most people who score high on them never commit suicide (Goldstein et al., 1991). Thus, in the vast majority of cases, completion is not predictable. The failure of prediction has an important corollary. Most people who are admitted to the hospital as suicide risks are unlikely to kill themselves (Paris, 2006). Also, those who carry the most risk may never present clinically. The vast majority of completed suicides occur on the very first attempt and involve guns or hanging (Beautrais, 2001). The proposal for DSM-5, responding understandably to clinical need, was a scale assessing some standard risk factors (previous
Page 97
attempt, aggressivity, social isolation, recent losses, chronic pain, diagnosis of severe mental disorder, substance abuse, suicidal plans, and hopelessness). Yet it only had face validity, mirroring what psychiatrists have long been taught about the assessment of suicide risk and not providing scientific justification for clinical prediction. It did not belong in a diagnostic manual.
Functioning and Impairment
Diagnosis is a meaningful cluster of symptoms but does not always reflect functioning. We need a separate measure. Axis V in DSM-III and DSM-IV was a failure. It mixed apples and oranges. It tried to take into account symptom severity, ability to work, and the quality of intimate relationships—all in one number! The results were predictably unreliable and misleading. Thus, the history of Axis V is an instructive failure. Introduced in DSM-III, it was based on a measure developed years before by Luborsky (1962)—the Health–Sickness Rating Scale (HSRS)— and later adapted into a Global Assessment Scale (GAS; Endicott et al., 1976). The GAS was renamed in DSM-III-R as the Global Assessment of Functioning (GAF). The concept was to score functioning on a scale from 0 to 100. In case of a discrepancy between different areas of functioning (as one often sees), the score should reflect the lowest common denominator so that the most dysfunctional area would determine the final score. DSM-III also asked clinicians, as a way of assessing change, to record a score for the highest level in the past year, which could then be compared to the current GAF. For more than 30 years, I taught residents to write clinical reports that include a GAF score. They almost never got it right. The main reason is that the number is a composite. A patient could be unemployed and isolated but only have mild symptoms, whereas another patient could have severe symptoms despite a good job and a loving family: Both could receive the same GAF score. In practice, the GAF functions not as a 100-point scale but, rather, a 30-point
Page 98
scale. Any score greater than 80 would be Utopia, and I do not think I have ever felt that good for more than a week. Patients who score between 70 and 80 are having expectable reactions to situations, a level that defines them as mentally healthy. We do not see patients like that either. Those with real mental disorders always score less than 70. Thus, patients with only mild dysfunction would fall between 60 and 70, those with moderate problems between 50 and 60, and those with severe dysfunction less than 50. This leaves a very narrow range. Finally, the GAF scale lacked clinical utility. Almost none of my colleagues (including those who used to be my students) could follow its complex procedures. So I am not unhappy to see it go. But I am not yet convinced that DSM-5 has come up with anything better. One scale was intended to follow the guidelines of a National Institutes of Health initiative called Patient-Reported Outcomes Measurement Information System (PROMIS; Anatchkova & Bjorner, 2010). It might be used as a measure of both functioning and disability (Narrow et al., 2009; Narrow & Kuhl, 2011), but its scientific basis is uncertain. It is much more complicated than GAF, virtually guaranteeing that it will never be used in clinical practice. A much better choice, now formally recommended by DSM-5, is the World Health Organization Disability Assessment Schedule or WHO-DAS 2.0. Along with a guide to its use, this instrument is available online at http://www.who.int/classifications/icf/whodasii/en.
It has self-report and clinician-report versions. The WHO-DAS is a generic assessment instrument for health and disability that can be used for all diseases, including mental, neurological, and addictive disorders. It has been applied in both clinical and general population settings, produces standardized disability levels and profiles, and is applicable across cultures and for all adult populations. The WHO-DAS is a self-report measure that is short, simple, and can be administered in 5–20 minutes. It is commonly used for insurance assessments that require a formal and quantitative measure of disability. It is also likely to be used in research. However, it is unlikely to become a routine part of
Page 99
diagnostic procedures in practice. Busy clinicians will not want to take the time to do complex scoring when they can directly observe severity. The point to remember is that functioning is only partly driven by categorical diagnosis. Some people with serious mental disorders manage surprisingly well, whereas some patients with mild to moderate mental disorders function at a low level. Even in schizophrenia, 10–20% of patients are employed, depending on the job market (Marwaha & Johnson, 2004). Nearly one-fourth of men with schizophrenia (and approximately half of women) will eventually marry and have families (Saugstad, 1989). At the same time, patients with common mental disorders (anxiety and depression) can be unemployed and socially isolated. Rating psychosocial functioning is a complex business. Un packing the mixed bag of Axis V requires clinicians to make multiple ratings, which they will not have time to do. In any case, any scale used to determine whether patients will receive money from insurance companies or governments is not going to be fully objective.
Abolishing the Five-Axis System
Axis V was part of a larger assessment procedure that was considered innovative when DSM-III was introduced. It offered a way to be “biopsychosocial” by considering personality, medical illness, stressors, and functional levels on top of diagnosis. Some of my colleagues became very attached to the five axes and continue to use them even though they are not in DSM-5. My own conclusion is that the system never worked properly. Almost everyone focused on Axis I—the traditional way of making categorical diagnoses. Most clinical reports stopped there. Axis II (personality disorders and intellectual disability) became a second-class citizen, mostly ignored or, at best, “deferred.” Including these diagnoses in the same axis as every other category avoids this kind of stigmatization. Regarding Axis III, medical diagnoses are
Page 100
of interest, but they can usually be found in a full report. Axis IV depended on a vague rating stressor that was not validated and difficult to score. As already noted, Axis V was an amalgam of symptoms and psychosocial functioning. My view is that the five axes will not be missed, and they will gradually die out. What was cutting-edge 30 years ago turned out to be problematic in practice. One can only wonder whether the same judgment will be made about some of the innovations of DSM-5 when DSM-6 is ready for publication.
Dimensionalization as a Vision
The American journalist Lincoln Steffens, who visited the Soviet Union soon after the Russian Revolution, famously concluded, “I have been over into the future, and it works.” It took decades for everyone to realize just how wrong he was. Similarly, the editors of DSM-5 had a visionary view of psychiatric diagnosis. They knew where psychiatry was going, and they wanted to help it get there more rapidly. Dimensionalization was a major factor in their ideology. Only time will tell whether it works for or against clinical practice. First (2010) nicely summarizes these issues:
For there to be any chance that the DSM-5 dimensions will fare better than their DSM-IV predecessors, significant efforts must be made to establish their reliability, sensitivity to change, and clinical utility. Fundamental to this effort is empirical evidence establishing not only that clinicians find such measures “feasible” or “acceptable,” but also that the use of such measures improve clinical outcomes. Otherwise it is unlikely that clinicians will be motivated to spend the time and effort required to put the measures into routine clinical use. (p. 698)
The situation would be different if dimensions had the scientific status of blood pressure. As it stands, the only practical issue for
Page 101