A paradigm shift in psychiatric classification: the Hierarchical Taxonomy Of Psychopathology (HiTOP)

Many have argued that a hierarchical dimensional approach to psychiatric classification would better align the nosology with data on the natural organization of psychopathology1. However, such proposals have often been resisted on the grounds that: a) consensus among dimensional models is lacking and b) categorical diagnoses are considered to be essential to clinical decision‐making.

The Hierarchical Taxonomy Of Psychopathology (HiTOP) consortium (see https://medicine.stonybrookmedicine.edu/HITOP) was formed by psychiatric nosologists to develop a consensus dimensional classification that is more clinically informative than the traditional diagnostic systems (DSM and ICD).

This group of scientists (now including 69 members) reviewed studies on the structure of psychopathology and developed a consensual model2. The resulting system offers to address problems of arbitrary disorder boundaries (consequences of which include subthreshold and not otherwise specified cases) and substantial unreliability of traditional diagnoses, by characterizing psychopathology in terms of dimensions rather than categories.

The system resolves the problem of within‐disorder heterogeneity by constructing dimensions on the basis of the observed covariation of symptoms, thus identifying coherent constructs. It deals with comorbidity by identifying higher‐order dimensions that reflect associations among lower‐order dimensions. This hierarchy summarizes patterns of comorbidity and enables practitioners to study and treat characteristics common to multiple conditions. Importantly, HiTOP encompasses both transient symptoms and stable maladaptive traits.

The HiTOP hierarchy has five levels. It combines symptoms, signs and maladaptive behaviors into tight‐knit symptom components (e.g., insomnia) and maladaptive traits (e.g., emotional lability). These, in turn, are combined with closely related components/traits into dimensional syndromes, such as vegetative depression (that includes insomnia, psychomotor retardation, lassitude and appetite loss)3. Similar syndromes are combined into subfactors, such as a distress dimension that includes depression, generalized anxiety, post‐traumatic stress and some borderline personality traits. Larger constellations of syndromes form broad spectra, such as an internalizing dimension that consists of distress, fear, eating pathology and sexual problems. Finally, spectra can be aggregated into extremely broad super‐spectra, such as the general factor of psychopathology that reflects characteristics shared by all mental disorders.

HiTOP organizes psychopathology according to evidence from statistical modeling and validation studies2, but it is a phenotypic model and does not directly incorporate etiology. Would such an approach perform substantially better than the traditional diagnostic systems? There are two reasons to expect that it will. First, dimensional phenotypes have been found to have greater reliability and stronger associations with validators than categorical diagnoses4, indicating that dimensional descriptions are more informative. Second, dimensions have been shown to be more useful in clinical research. HiTOP aligns much better than traditional diagnostic systems with the genetic architecture of mental disorders and with the effects of environmental risk factors, such as childhood maltreatment2, 5, 6. HiTOP dimensions can explain nearly all long‐term chronicity of psychopathology7. HiTOP also far outperforms traditional systems in accounting for functional impairment3. Moreover, HiTOP dimensions can help to explain why disorders from different classes respond to the same treatment (e.g., social anxiety disorder to antidepressants)5. Indeed, some spectra already have become useful targets for treatment development8.

Another response to shortcomings of traditional diagnostic systems is the Research Domain Criteria (RDoC) framework, a dimensional classification of basic psychological processes potentially relevant to psychiatric problems. The RDoC initiative aims to develop an etiologically‐based nosology, but its scope is largely limited to constructs conserved across species and linked empirically to neural circuitry. Also, the RDoC framework is focused primarily on basic levels of analysis, and its clinical translation lies well in the future. In contrast, HiTOP was designed to be immediately useful in clinical research and practice.

HiTOP can inform the RDoC initiative by identifying key clinical dimensions that need to be studied. Conversely, HiTOP is a descriptive system, and RDoC research can clarify the nature and validity of HiTOP dimensions. It is likely that some RDoC dimensions lack coherent phenotypes and that some HiTOP dimensions have intractable biology, but in areas of convergence these models may ultimately produce a unified nosology, achieving a comprehensive understanding of psychopathology.

Furthermore, HiTOP can help to improve clinical practice immediately. Clinicians often forego a formal diagnostic assessment, as many consider it to have little clinical utility9. Initial evidence suggests that dimensional models can be more informative than traditional diagnoses in clinical decision‐making¹⁰. Indeed, dimensional descriptors are indispensable in other areas of medicine (e.g., body mass index, blood pressure, laboratory test results). In psychiatry, dimensional measures have a long history of clinical use (e.g., personality inventories, symptom ratings, intelligence tests, neuropsychological tests).

To date, HiTOP has not been used clinically as a complete system, but it relies heavily on concepts and constructs embedded in widely‐used dimensional measures. In fact, available HiTOP‐aligned measures (see http://psychology.unt.edu/hitop) allow practitioners to implement many aspects of the system already.

HiTOP can be used most feasibly in a stepwise manner, beginning with a brief measure of the six spectra. If problems are detected in some spectra, lengthier measures can be administered to characterize dimensions within those domains (while the other domains do not require further assessment). Thus, a HiTOP diagnosis is a patient's profile on relevant dimensions. Although such profiles may include a large number of scales, they are often simpler than traditional manuals, with their hundreds of codes and numerous permutations necessitated by comorbidities¹⁰.

Clinical decisions require cut‐offs on dimensions to guide specific actions. The HiTOP consortium aims to develop such cut‐offs empirically, and cut‐offs based on statistical deviance already exist (e.g., two standard deviations above the mean indicate high severity).

Indeed, HiTOP is a work in progress. Ongoing efforts aim to extend the system to all forms of psychopathology, construct an integrated measure of all HiTOP dimensions, and develop detailed guidance for clinicians using the system. Much more needs to be done, but HiTOP already can be applied in a variety of contexts. At minimum, it provides a framework for conceptualizing research phenotypes and individual patients dimensionally. Ultimately, HiTOP is expected to offer a roadmap for researchers and clinicians that is much more informative than traditional diagnostic systems.

Roman Kotov1, Robert F. Krueger2, David Watson3
1Department of Psychiatry, Stony Brook University, Stony Brook, NY, USA; 2Department of Psychology, University of Minnesota, Minneapolis, MN, USA; 3Department of Psychology, University of Notre Dame, South Bend, IN, USA