A Dimensional-Spectrum Model of Psychopathology: Title and subTitle BreakProgress and OpportunitiesDimensional-Spectrum Model of Psychopathology

In this issue of the Archives, Kessler et al¹ provide a thorough account of the meaning of comorbidity among DSM-IV mental disorders. Their contribution involved comparing hypotheses regarding the developmental sequencing of comorbidity. One hypothesis was that specific disorders are involved in specific developmental comorbidity patterns. For example, obsessive-compulsive disorder might generally be a primary disorder, with that disorder then leading people to have few social contacts owing to the debilitating nature of their symptoms. Diminished social involvement might typically contribute to the development of a secondary major depressive episode in persons who have a primary diagnosis of obsessive-compulsive disorder. Numerous plausible pathways of this sort can be hypothesized.

A contrasting hypothesis is that specific patterns of association among disorders over time are specific instantiations of more general processes. In the Kessler et al¹ contribution, this generality hypothesis was supported, but the former specificity hypothesis was generally not supported. Eighteen different disorders were studied, yielding 306 possible pairwise, time-lagged associations between specific disorders. A model conceptualizing these time-lagged associations as functions of underlying internalizing (mood and anxiety) and externalizing (substance use and antisocial behavior) propensities fit better than a model positing highly specific associations among specific disorders. A handful of specific associations remained, but these were few and far between. The conclusion that internalizing and externalizing propensities underlie the development of comorbidity was based on extensive cross-cultural data. The conclusions of Kessler et al¹ were derived from representative community surveys in 14 countries around the world, as opposed to being derived from a specific locale.

The picture that emerges from the work of Kessler et al,¹ as well as other extensive work on modeling comorbidity, is clear. Mental disorders do not delineate highly discrete and easily distinguished categories. Rather, they delineate continuous underlying propensities to experience psychopathology.² The perspective that emerges from contemporary comorbidity research forms the prolegomenon to a quantitative dimensional-spectrum model of psychopathology.

The notion of a latent underlying dimensional spectrum may seem somewhat abstract on first reading. Consider, for example, part B in the Figure of Kessler et al.¹ This elegant model fits the data better than the more complicated model shown in part A of their Figure. But what is actually represented by the variables labeled “I” (the internalizing spectrum) and “E” (the externalizing spectrum) in part B?

The internalizing and externalizing variables are simply the extent to which the diagnoses in a specific group tend to go together. Simple counts of the number of diagnoses falling within each rubric are rough proxies for the I and E variables in part B of the Figure. Additional refinements of this straightforward understanding are also captured by the kinds of models used in this research. Consider, for example, Table 2 in Kessler et al.¹ In the total sample, bipolar disorder types I and II have the lowest loading of any of the disorders on the first factor (the internalizing spectrum). Bipolar disorder types I and II are related to the internalizing spectrum, but more weakly than the other internalizing disorders. This makes sense because some aspects of bipolar disorder types I and II pertain to internalizing disturbance, whereas other aspects are related to psychotic disturbance, and this hybrid nature of bipolar disorder types I and II explains why their placement in classification systems is often controversial.³

Kessler et al¹ model an impressive array of 18 specific diagnoses. This diversity of psychopathology can be well conceptualized in terms of the 2 broad internalizing and externalizing spectra. Many insights have been achieved by examining specific disorders and seeing how they fit within the internalizing-externalizing model. For example, borderline personality disorder is a complex form of psychopathology because it does not fit neatly into comparisons of internalizing and externalizing spectra and appears to be related to both underlying liabilities.⁴

Nevertheless, further development of a dimensional-spectrum model requires “unpacking” existing diagnoses and developing models at the symptom level. When this is done, the psychopathology encompassed by diagnoses defined such that they have low prevalence (eg, criterion A of DSM-IV –defined schizophrenia) can be unpacked, and additional spectrum constructs emerge as a result. For example, when a panoply of DSM-IV Axis I and II symptoms are modeled (as opposed to diagnoses), internalizing and externalizing spectra are complemented by 2 additional spectra, one pertaining to pathological forms of introversion (eg, social anxiety) and the other pertaining to pathological forms of thought disorder (eg, hallucinations and delusions).⁵ This “unpacking endeavor” suggests that the dimensional-spectrum approach need not be limited to common mental disorders but can also be extended to less prevalent disorders by working with their symptoms directly. The approach can be used to build an empirically based nosology from the symptom level up, as opposed to relying on clinical experts to assemble symptoms into diagnoses.

A dimensional-spectrum model can help improve the limited clinical utility of existing official nosologies. Rather than organizing disorders into chapters designed to promote differential diagnostic decisions that do not accurately map the natural tendencies for disorders to co-occur, diagnoses can be grouped on the basis of their empirical affinities. In this kind of arrangement, comorbidity is understood to be a natural function of shared underlying risk, and the dimensionality of psychopathology is directly accommodated.⁶ Indeed, accommodating the dimensionality of psychopathology is a major goal of the transition from DSM-IV to DSM-5.

A dimensional-spectrum model also provides a framework for understanding the specificity and generality of biomarkers, putative endophenotypes, and genetic factors that are linked to risk for psychopathology. For example, the genetic structure of psychopathology⁷ closely resembles the phenotypic structure delineated by Kessler et al.¹ In addition, a dimensional-spectrum model leads to testable hypotheses about the extent to which biological correlates of psychopathology are general vs specific. As an example, the p300 evoked potential was originally thought of as a candidate endophenotype for alcohol dependence but is actually better conceptualized as an endophenotype for the externalizing spectrum.⁸ Underlying dimensions of psychopathological variation, under the aegis of the Research Domain Criteria initiative, are becoming key phenotypic constructs for research supported by the National Institute of Mental Health.

Explaining the coherence of the internalizing and externalizing spectra will likely involve multiple levels of analysis, from the molecular to the systemic. Regardless of the chosen level of etiologic analysis, however, prevention and intervention efforts will almost certainly have the greatest impact when they successfully target these spectra, rather than their specific manifestations in specific forms of psychopathology. For example, the personality trait of neuroticism, reflecting the core of the internalizing spectrum, is arguably the single most important factor in behavioral public health⁹ and has documented economic costs that actually exceed the costs of diagnosable mental disorders.¹⁰

Thinking of broad-liability spectra as targets for prevention and intervention may seem daunting. Given that psychopathology manifests along these broad lines naturally, however, such fears are unwarranted. The dimensional personality traits at the core of psychopathology spectra can be treated and may, in fact, be the true targets of interventions aimed at common mental disorders.¹¹ A dimensional-spectrum model of psychopathology and personality has great potential to frame an empirically based approach to classification that, in turn, is generative of new insights.

Correspondence: Dr Krueger, Department of Psychology, University of Minnesota, 75 E River Rd, Minneapolis, MN 55455 (krueg038@umn.edu).

Financial Disclosure: None reported.