Psychiatric diagnostics 8

This is a continuation of my series of posts on psychiatric diagnostics. I have previously described both operationalized and prototypical diagnostics as well as network-based approaches. In this section, I will write about an alternative model to the DSM: HiTOP (The Hierarchical Taxonomy of Psychopathology).

As previously mentioned, DSM-III was created as a response to the disorder that existed in psychiatric diagnostics. In a psychodynamic paradigm, diagnoses are not particularly relevant since the focus is on individuals’ problems and internal conflicts. Diagnoses are too superficial and describe people at a group level. As a result, for example, schizophrenia was more commonly diagnosed in the U.S. than in Europe. To counteract this, DSM-III was constructed with lists of criteria for each included diagnosis. The goal was to increase reliability and bring order. However, this would not necessarily improve validity, and it was assumed that diagnoses might change in the future as more knowledge about pathophysiology emerged. The nosology (classification) itself was partially based on studies but also relied on already known conditions.

Obvious problems with the DSM—and, for that matter, the ICD—include:

• Frequent comorbidity of questionable relevance.

• Some diagnoses are extremely heterogeneous (especially depression, which mixes melancholic depression with reactive conditions).

• Due to the categorical structure rather than a spectrum-based approach, the boundaries are arbitrary, creating strange distinctions.

• The same symptoms appear in a wide range of different diagnoses (e.g., anhedonia).

One solution to the above problems is to allow data to guide the classification of diagnoses instead of systematizing them a priori based on symptoms. This can be achieved by using statistical methods to examine how symptoms naturally cluster in individuals. In this way, a taxonomy is constructed with a hierarchy where each unit is dimensional. The system itself is intended to be updated as new data is incorporated, and the currently accepted model is as follows.

At the lowest level of the hierarchy are symptoms, and below them, corresponding DSM diagnoses (though HiTOP is not intended to simply reorganize DSM so that every diagnosis corresponds to a cluster). Above the symptoms are traits, which are then grouped into syndromes, subfactors, and spectra. At the top of the hierarchy is a super-spectrum corresponding to the p-factor and externalization.

This approach is not new—it resembles the model used for personality in the Five-Factor Model, or for that matter, the p-factor itself. The p-factor is intended to represent a general susceptibility to psychological distress that everyone possesses to varying degrees. It is analogous to the g-factor, which describes general intelligence and quantifies the correlation between individuals' results across different cognitive ability tests.

As an example, we can take a closer look at the externalizing super-spectrum. This includes antagonism and disinhibition as spectra, which then lead to harmful substance use and antisociality as subfactors. The former extends downward in the hierarchy to various substance use disorders, while the latter corresponds in the DSM to ADHD, oppositional defiant disorder, and antisocial personality disorder. Close to this cluster, symptoms related to dominance, manipulation, attention-seeking, etc., are also found, which aligns with the fact that borderline personality disorder is positioned nearby in the corresponding DSM diagnoses.

The hope for HiTOP is that it will be empirically supported and thus enable better research, as well as be useful for clinicians. However, I am skeptical about the latter, as competent clinicians rarely focus solely on a specific DSM diagnosis but instead assess a patient's strengths, weaknesses, and subphenotypes. In fact, earlier versions of the DSM used an axis system, where, in addition to a diagnosis like bipolar disorder, functional level, somatic comorbidities, and personality factors were also specified, providing a more comprehensive picture of the patient.

Moreover, the HiTOP model is based on clustering individuals’ DSM symptoms. As a result, it loses important phenomenological information—that is, how the patient experiences their situation. In terms of validity, it is quite possible that HiTOP will prove valuable. The last 10–20 years of investment in psychiatric genetics have been successful, with data collected from tens of thousands of patients revealing hundreds of robust genetic variants associated with different diagnoses. However, genetics has not followed the DSM structure; rather, there is substantial overlap.

A good illustration of this comes from an article by the Dutch geneticist Abdel Abdellaoui.

Here, strong correlations can be seen between different psychiatric diagnoses, as well as with cognitive phenotypes such as intelligence and educational level. I have personally participated in studies examining how different subphenotypes in bipolar patients are linked to genetic risk (polygenic risk scores) for various conditions. For example, functional level measured by GAF and the degree of inter-episodic remission are positively correlated with genetic risk for bipolar disorder but negatively correlated with the risk for schizophrenia and depression (see here). Furthermore, the level of sick leave and unemployment among bipolar patients is positively correlated with genetic risk for schizophrenia, depression, and ADHD, but not for bipolar disorder (see here).

Whether genetics and neurophysiology align better with HiTOP than with DSM is an empirical question that requires further research. Regardless of the model chosen, it is important to deepen diagnostics and include many alternative phenotypes in studies to later determine what is associated with what.