Reconstructing Psychopathology: A Data-Driven Reorganization of the Symptoms in the Diagnostic and Statistical Manual of Mental Disorders, 2024,Forbes

Chandelier

Senior Member (Voting Rights)

Miriam K. Forbes,
Andrew Baillie,
Philip J. Batterham,
Alison Calear,
Roman Kotov,
Robert F. Krueger,
Kristian E. Markon,
Louise Mewton,
Elizabeth Pellicano,
Matthew Roberts,
Craig Rodriguez-Seijas,
Matthew Sunderland,
David Watson,
Ashley L. Watts,
Aidan G. C. Wright,
Lee Anna Clark

Abstract​

In this study, we reduced the fifth edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-5) to its constituent symptoms and reorganized them based on patterns of covariation in individuals’ (N = 14,762) self-reported experiences of the symptoms to form an empirically derived hierarchical framework of clinical phenomena.

Specifically, we used the points of agreement among hierarchical principal components analyses and hierarchical clustering as well as between the randomly split primary (n = 11,762) and hold-out (n = 3,000) samples to identify the robust constructs that emerged to form a hierarchy ranging from symptoms and syndromes up to very broad superspectra of psychopathology.

The resulting model had noteworthy convergence with the upper levels of the Hierarchical Taxonomy of Psychopathology (HiTOP) framework and substantially expands on HiTOP’s current coverage of dissociative, elimination, sleep–wake, trauma-related, neurodevelopmental, and neurocognitive disorder symptoms.
We also mapped some exemplar DSM-5 disorders onto our hierarchy; some formed coherent syndromes, whereas others were notably heterogeneous.
 

AI Summary
Summary of the Study "DSM Disorders Disappear in Statistical Clustering of Psychiatric Symptoms"

The study "Reconstructing Psychopathology: A data-driven reorganization of the symptoms in DSM-5," led by Miri Forbes and colleagues, explores the quantitative structure of psychiatric disorders using a statistical approach that challenges traditional DSM-5 classifications. The research is set in the context of the Hierarchical Taxonomy of Psychopathology (HiTOP), a model that emphasizes dimensional, hierarchical, and quantitative classification of mental disorders based on symptom patterns. The study's methodology is groundbreaking, utilizing a large, socio-demographically diverse sample of 14,800 participants who completed surveys based on DSM-5 symptoms. These symptoms were presented in a random order to minimize bias, and participants rated how true each symptom was for them in the past 12 months.

Through advanced statistical methods, the study identified 139 symptom clusters ("syndromes") and 81 solo symptoms. These clusters were subjected to hierarchical analysis to determine higher-order constructs, resulting in eight major spectra, such as Externalizing, Harmful Substance Use, Mania/Low Detachment, and Thought Disorder, with 27 subfactors. For instance, the "Internalizing" spectrum, which traditionally encompasses disorders like Major Depressive Disorder (MDD), was broken down into subfactors such as Distress, Social Withdrawal, and Trauma. Importantly, the results revealed that many classic DSM-5 disorders, including MDD, Generalized Anxiety Disorder (GAD), and Post-Traumatic Stress Disorder (PTSD), did not emerge as distinct, coherent syndromes in the analysis.

The Disappearance of Major Depression and Other Disorders

The study's findings challenge the validity of traditional diagnostic categories like MDD, GAD, and PTSD, which failed to emerge as distinct clusters when subjected to statistical analysis. These disorders, as defined by DSM-5, appear to be heterogeneous in nature, with varying symptom subsets that do not correspond to a stable, identifiable syndrome. For example, while MDD is characterized by a combination of symptoms such as depressed mood, anhedonia, and changes in sleep and appetite, these symptoms were found to dissolve into more fundamental, statistically homogeneous clusters like "Distress," "Neurocognitive Impairment," and "Dysregulated Sleep." This observation raises the question of whether MDD, as it is currently conceptualized in the DSM, truly represents a cohesive disorder or simply indexes a broad, overlapping range of symptoms that vary across individuals.

Participants in the study demonstrated a wide variety of symptom combinations, making it clear that there is no single, unified symptom pattern that can reliably define disorders like MDD. For example, some individuals with MDD might present with depressive mood and suicidality, while others might experience irritability and sleep disturbances. The findings suggest that traditional psychiatric categories like MDD, GAD, and PTSD are, in fact, variable collections of symptoms that share certain core features (e.g., depressed mood for MDD, pervasive anxiety for GAD), but do not reflect fixed, stable conditions.

Implications for DSM and Psychiatric Diagnosis

The study's results imply that the symptom heterogeneity inherent in DSM-5 constructs may be problematic for accurate diagnosis. The lack of statistical homogeneity within the criteria for MDD and other disorders challenges the assumption that these disorders represent stable, discrete entities. Instead, they may reflect varying and overlapping symptom clusters that do not map neatly onto traditional diagnostic labels. As noted by psychiatrist Ken Kendler, while DSM criteria serve as an index for clinical depression, they may not fully capture the underlying, variable nature of depressive symptoms, leading to potential gaps in diagnosis and treatment.

Limitations and Future Directions

While the study provides valuable insights into the complexity of psychiatric disorders, there are important limitations to consider. The research relies on self-reported symptoms, which may not capture the full complexity of each individual's experience. Additionally, the study does not account for clinician observations or differentiate between symptoms caused by different factors (e.g., insomnia due to anxiety versus insomnia due to substance withdrawal). The time scale of 12 months used for symptom assessment may also overlook shorter-term variations in symptom patterns.

Forbes et al. themselves acknowledge the need for further research to validate these findings using alternative measures and time frames, as well as multi-method and multi-informant approaches. Additionally, future studies should explore how these results apply across diverse sociocultural contexts, given the potential for variations in symptom expression across different demographic groups.

In conclusion, the study by Forbes et al. offers a fresh perspective on psychiatric classification by challenging the stability and coherence of traditional DSM-5 disorders. It underscores the need for a more flexible, dimensional approach to understanding mental health that takes into account the complex, heterogeneous nature of psychiatric symptoms.
 
We also mapped some exemplar DSM-5 disorders onto our hierarchy; some formed coherent syndromes, whereas others were notably heterogeneous.

Maybe I am an idiot, but I completely fail to see how a mere consistent grouping of symptoms establishes a unique discrete entity with underlying heterogeneity.

In the same way that correlation is necessary but not sufficient to establish causation.
 
The study's methodology is groundbreaking, utilizing a large, socio-demographically diverse sample of 14,800 participants who completed surveys based on DSM-5 symptoms.
This reads too much like "we fed the output of a LLM to another LLM" to me, though. The DSM is mostly a collection of random unexplained bits, with some things that are probably mostly accurate but too superficial to be of much use. Good chances that most of the top LLMs hallucinate a lot less, the DSM is a truly wild document, filled with the odd neuroses of people who can't deal with problems they don't have a solution to.

I don't see how just symptoms is enough for this, this is something that has sunk us completely, because there is more data in the patterns. There is simply no depth to this kind of data. It's as hollow as asking the public questions like "is the country heading in the right direction?" without taking into account that a lot of people will give the same answer for completely different reasons, while others will give different answers but for the same reason. Some always more of the horrible thing that threatens others.

Using only symptoms is too superficial to work with. This is why there is a running joke about how every symptom you enter in WebMD eventually leads to cancer, which, unlike imagined magical powers of the mind, can and does create any and all symptoms. And one thing that makes the obsession with a single symptom so pointless, a fact that is emphasized by a complete lack of progress.

Which is one of many things that make the DSM invalid. So using it as a source to correct it seems mostly misguided. Although I am all in on the proposition that the DSM should be dismissed as a reliable source of anything, and defaulting to the proper scientific position: we don't know. This is the big change medicine needs to go through in order to truly mature into a real profession. It's everything.

Actually, since this very much overlaps with a recent badly misinterpreted study from OpenAI, I have to mention it. Yesterday, OpenAI published a study showing that with current methods, LLMs "hallucinations" are inevitable, simply because bullshit is more often rewarded than it is punished. LLMs "learn" entirely through reward feedback.

Psychiatry seems to produce the same mistakes, especially with the DSM. Bullshit is rewarded, and thus "hallucinations", aka bullshit, are inevitable, simply because you can never go wrong saying cruel nonsense like autism being caused by bad parents, or fatigue being caused by, uh, I don't even know at this point what they are even saying, and the DSM is basically a repository of such hallucinations, mixed in with things that are likely genuine but misinterpreted.

So none of this failure is inevitable, except for the fact that bullshit is rewarded more often than it is punished. As far as I can tell, making false allegations blaming the victims of illness for their own misery is never punished, in fact is more likely to be literally awarded and praised. So it doesn't need to always be rewarded to perpetuate itself, it just never gets corrected as long as the right answer isn't found, which a system that rewards bullshit systematically impedes.
The research relies on self-reported symptoms, which may not capture the full complexity of each individual's experience. Additionally, the study does not account for clinician observations or differentiate between symptoms caused by different factors (e.g., insomnia due to anxiety versus insomnia due to substance withdrawal)
This is ironically a strength in many cases, though not all, because if there is one pattern that is unmistakable when it comes to psychosomatic ideology is that clinical expertise adds nothing and removes most of the meaning, a reversal of a normal pattern that is never given due consideration.

Maybe mental illness is different, but even the "top" experts can't tell the difference anyway, so why bother with this charade? Other than protecting weak egos, of course.
 
Importantly, the results revealed that many classic DSM-5 disorders, including MDD, Generalized Anxiety Disorder (GAD), and Post-Traumatic Stress Disorder (PTSD), did not emerge as distinct, coherent syndromes in the analysis.
All of this could have been avoided with the simplest heuristic of all: we don't know. This is obvious to anyone who pays attention to this stuff, so it's very hard to describe why the entire profession sees things that don't exist, even in the process of describing this ver process in others. It reveals the very kind of lack of clarity of thought that is observed in mental illness, for causes unknown, except here the causes are not just known, they are obvious and intentional.
The study's findings challenge the validity of traditional diagnostic categories like MDD, GAD, and PTSD, which failed to emerge as distinct clusters when subjected to statistical analysis.
Some of the descriptions do make sense, but they are not the ones that are popular. Those aren't even coherent with themselves. Most of those are even common consequences of illness, and instead of making sense of it, they literally invented fake mental illnesses instead. Some sort of Inception "nightmare within a nightmare" Russian doll of misery and programmed death.
 
The lack of statistical homogeneity within the criteria for MDD and other disorders challenges the assumption that these disorders represent stable, discrete entities. Instead, they may reflect varying and overlapping symptom clusters that do not map neatly onto traditional diagnostic labels.
Unfortunately, the odds that this leads to "ah, more biopsychosocial is clearly the answer" is basically 99%. It was basically created to be this flexible "describes everything, explains nothing" fluid bit of random nonsense.

Not that it will change anything, this is already guaranteed to occupy at least the next decade.
 
Back
Top Bottom