Are Critiques of DSM/ICD as Devastating for Psychiatric Diagnosis as Some Critics Seem to Think?
George Ikkos — a psychiatrist in the UK and a good friend (see his Mixed Bag post on the social in psychiatry) — brought up an interesting question in personal correspondence that deserves addressing in some detail in this newsletter. Researchers in psychiatry and psychology are now accustomed to pointing out that scientific studies on etiological mechanisms or statistical factor analyses do not support the validity of DSM and ICD diagnoses. Ikkos mentioned two recent illustrative examples. Bruce Cohen and Dost Öngür write in a 2023 paper:
“There is growing evidence that these primarily categorical structures of both ICD and DSM do not fit either the clinical presentations of patients, recent discoveries from genetics and neurobiology, nor therapeutic choices well.”
“The current state of research on the underlying mechanisms and clinical expressions of psychiatric disorders does not support the validity of standard diagnoses, especially for psychotic and mood disorders.”
Another recent paper authorized by prominent researchers (A review of approaches and models in psychopathology conceptualization research) summarizes: “In other words, the dominant models of mental disorder classification (such as discrete DSM-5 diagnoses) do not fit the data.”
Ikkos writes that even as we accept the clinical utility of psychiatric diagnoses, “By any standards, these seem to me devastating statements and … come close to the positions of some of the harshest critics of psychiatric practice.” He goes on to say that he fears that the staunchest “defenders” of psychiatry online may be lagging behind the foremost academics and researchers in appreciating the magnitude of the challenge we face.
These are indeed devastating statements; in fact, I wholeheartedly agree with these statements. But they are also not devastating in the way in which some of the harshest external critics of psychiatry imagine them to be. The relationship of DSM/ICD with the larger diagnostic practice requires some unpacking. In speaking of such harsh critics, I’m thinking of the British psychology “criticals” here, who have been arguing for abandoning psychiatric diagnosis and replacing it with formulation-based alternatives such as the Power Threat Meaning Framework, as well as various writers for “antipsychiatry” websites who delight in making a mockery of diagnostic manuals, as if only fools could take them seriously to begin with. What unites these critics is that they take statements by scientific critics, such as those quoted above, about the validity of DSM/ICD as the basis for the rejection of the entire practice of psychiatric diagnosis. Such critics reveal their impoverished understanding of the nature of psychiatric diagnosis in the process. Furthermore, by arguing for non-diagnostic alternatives, which face even greater scientific challenges with regards to clinical reliability and mechanistic validity, they reveal the unscientific motivations behind these criticisms.
(To be clear, Ikkos is not one of the critics I’m referring to above. George and I share many of the same concerns, and his own ideas about radicalism and psychiatry are briefly set out in a recent paper in BJPsych Bulletin)
Clinicians typically rely on prototypical descriptions of psychiatric categories
How much do psychiatric clinicians rely on DSM and ICD? If DSM and ICD were to disappear, what would psychiatric clinicians fall back on?
External critics of psychiatry seem to imagine that the DSM and ICD form the backbone of psychiatric diagnostic practice. The reality is that most psychiatric clinicians don’t care that much for the manuals. Nearly everyone uses them for administrative documentation and billing/reimbursement. Only about half or so refer to specific diagnostic criteria with any regularity (see, for instance, the survey by First et al, 2018). So, what is happening here?
DSM uses operationalized diagnostic criteria. It specifies precise duration and severity thresholds that need to be met in order for someone to qualify for a diagnosis. This reliance on operationalization ends up being pseudo-precise and somewhat arbitrary, but it was intended to increase reliability (i.e. increase the chance that two different examiners will arrive at the same diagnosis), especially in research settings. Clinicians, however, don’t generally use operationalized criteria. Clinicians rarely, if ever, administer the Structured Clinical Interview for DSM Disorders (SCID), which is reserved for research settings. Instead, clinicians typically rely on prototypical descriptions of psychiatric categories (Huda, 2019).
Psychiatric classification has a long history, and the end result is syndromic categories that have demonstrated practical value to multiple generations of clinicians. DSM and ICD are formalizations and operationalizations of these categories, but the clinical descriptions of these syndromes exist independently in most cases (these descriptions often diverge from operationalized criteria in interesting ways). Furthermore, psychiatric diagnosis is not merely a label referring to a list of symptoms; it’s embedded in the process of a comprehensive psychiatric evaluation. Psychiatrists use the presenting complaints and initial history of the patient to create a differential diagnosis, use that differential diagnosis to guide their clinical assessment, and then refine or revise the differential diagnosis as the assessment proceeds (Huda, 2019).
Prototypes (fuzzy kinds, ideal types) are an alternative to a strictly categorical classification. This approach relies on the notion of a typical example, which the members of the category can approximate to various degrees. Clinicians learn prototypical, textbook descriptions of various psychiatric presentations such as depression, mania, panic disorder, dementia, etc. These prototypes are refined over the course of training and one’s professional career. Prototypes are characterized by vague and fuzzy boundaries that overlap with neighboring prototypes. Clinicians tend to utilize a matching process where key features of the clinical presentation encountered are compared to prototypes of diagnostic profiles in an attempt to determine the best fit, which is then subjected to subsequent clinical confirmation. Many clinicians also complement the diagnosis with a narrative diagnostic formulation that summarizes their conceptualization of predisposing, precipitating, and perpetuating factors.
Descriptive psychopathology is the backbone of diagnosis
A second thing to appreciate is that the backbone of clinical diagnosis is not classification but rather (phenomenological) descriptive psychopathology. Diagnostic categories are a way of organizing more fundamental psychopathological phenomena for the purposes of communication, treatment, and scientific inquiry, but when diagnostic categories are uncertain or doubtful, clinicians fall back on features of psychopathology, such as delusions, obsessions, anhedonia, mood lability, irritability, etc. Clinicians work with signs and symptoms. Clinicians appreciate the dimensionality of symptom severity. Psychopathological categories are not strictly binary when we look at symptom profiles. People can be “on the border,” “subthreshold,” or “atypical.” Clinicians constantly make judgements regarding how to categorize and treat such cases. Even clinicians who prefer rating scales are aware of the dimensionality of presentations. PHQ-9 or GAD-7 give a continuum of scores with no natural discontinuities.
To an external observer, the most obvious thing about the psychiatric diagnostic process might be the DSM/ICD label assigned, but that misses the bulk of what is important and vital in the diagnostic process. The craft of psychiatric evaluation has been refined over centuries; in a sense, it is Lindy proof.1
DSM/ICD categories are partially and non-trivially valid
DSM-III was implicitly embedded within a neo-Kraepelinian approach to psychopathology, with the assumption that studying psychopathology using descriptive, operationalized criteria will eventually lead to a convergence of validators and discovery of underlying disease entities. This hope guided scientific research over the next 3 decades, and the project to identify valid, categorical psychopathological disease entities was largely unsuccessful. Once the neo-Kraepelinian assumptions are made explicit and cast aside, it is understood that DSM and ICD are pragmatic manuals, offering operational definitions for purposes of reliable communication. While the categories do capture useful group differences on various validators, the manuals can make no claim with regards to “carving nature at its joints.”
The DSM/ICD categories are partially and non-trivially valid in the sense that they do capture differential information with regards to various external validators, such as genetics, neuroimaging, longitudinal course, clinical symptomatology, and treatment response. There are average differences at the group level between major psychiatric categories, such as major depression, bipolar disorder, schizophrenia, alcohol use disorder, etc. However, these categories are not valid in the sense that they don’t correspond to categorically distinct neurobiological entities with singular essences. There is no sharp delineation among the disorders when it comes to neurobiological mechanisms, etiological risk factors, or treatment.
Since scientific critics are driven by concerns about validity, their goal is to replace DSM/ICD frameworks with classifications that possess greater validity; this is in contrast to the anti-diagnosis critics, whose goal is to reject and delegitimize the entire diagnostic project. The invalidity of DSM/ICD categories is taken to be proof that the project of seeking valid descriptions of psychopathology is futile to begin with.
Since scientific critics are driven by concerns about validity, their goal is to replace DSM/ICD frameworks with classifications that possess greater validity; this is in contrast to the anti-diagnosis critics, whose goal is to reject and delegitimize the entire diagnostic project. The invalidity of DSM/ICD categories is taken to be proof that the project of seeking valid descriptions of psychopathology is futile to begin with.
HiTOP takes the same symptoms that DSM works with and uses statistical analysis of patterns of symptom covariations to describe a hierarchy of dimensions. Arguably, DSM and ICD are coarse-grained, flat, categorical approximations of HiTOP’s fine-grained dimensional and hierarchical schema. DSM categories are syndromes identified through an iterative process of clinical observation and clinical consensus. HiTOP is the result of quantitative psychological research.
Far from casting doubt on the diagnostic project, HiTOP provides a convincing demonstration that symptom covariation and clustering are not illusions; they are a scientific reality. Symptoms cluster together in ways that can be hierarchically described, from dimensional syndromes to subfactors to spectra such as internalizing and externalizing. HiTOP spectra also appear to have improved performance on validators compared to DSM/ICD categories. HiTOP expands and enriches the psychiatric diagnostic project by offering a formalization of something good clinicians already do intuitively: think dimensionally and hierarchically. The “criticals” are either ignorant of HiTOP or they have failed to grasp the fundamental challenge HiTOP poses to the antipsychiatry worldview. It is easier for anti-diagnosis critics to keep bashing the DSM in 2023 as the ne plus ultra of psychiatric classification, and it is more convenient to offer wishy-washy formulations based on invocations of “power-threat-meaning” than it is to tackle the scientific developments in the field of nosology.
Far from casting doubt on the diagnostic project, HiTOP provides a convincing demonstration that symptom covariation and clustering are not illusions; they are a scientific reality… The “criticals” are either ignorant of HiTOP or they have failed to grasp the fundamental challenge HiTOP poses to the antipsychiatry worldview.
Other scientific developments, such as network approaches and clinical staging approaches, are also antithetical to the antipsychiatry worldview. Network approaches challenge reified and reductionist understandings of DSM/ICD syndromes, but they re-conceptualize them in terms of symptom networks and phase transitions in complex, dynamic systems. Again, rather than showing DSM syndromes to be illusory, network theories offer a more scientific way to understand their clinical reality. Clinical staging approaches highlight the pluripotential nature of psychopathology in earlier stages, but they do recognize classic psychiatric syndromes as phenotypes that stabilize later in the course of illness.
So, if scientific critiques of DSM/ICD are not devastating for the psychiatric diagnostic project, then what exactly do they demolish?
What has become obvious in the 21st century is the collapse of the neo-Kraepelinian worldview. The idea that psychiatric syndromes have essences consisting of neurobiological dysfunctions, to be identified by a convergence of validators, is dying, if not quite dead. Neuroscientific mechanisms of psychopathology crisscross, cluster, and overlap in complex ways, but they do not converge onto disease entities.
Cohen and Öngür write:
“Of particular note, unlike diseases that are heterogeneous in expression but share a single cause, or diseases that are heterogeneous in cause, but share a single expression, psychiatric disorders are heterogeneous in all ways and at all levels studied, in cause, mechanism, and expression of illness.”
How do you make something that is “heterogeneous in all ways and at all levels studied” tractable for clinical and scientific work? The answer is that we rely on practical kinds (see Peter Zachar’s work). Practical kinds are useful heuristic constructs that “carve up” the psychopathological space in ways that serve our clinical and scientific goals. The multi-level heterogeneity of causal mechanisms in psychopathology means that there is no single, privileged classification, no one, true categorization of conditions. Rather, different demarcations are optimal for different clinical and scientific goals. If we aim for statistically robust patterns of symptom covariation, we end up with HiTOP. If we aim for clinically recognizable syndromes based on clinical observation and consensus, we will have something like DSM and ICD.2 If we aim for an idiographic characterization of psychopathology in terms of a person’s personality and psychological dynamics, we have something like a psychodynamic formulation. If we aim for phenomenology, we get descriptive psychopathology. This creates a state of nosological pluralism in which DSM/ICD remain viable as pragmatic, clinical constructs, along with other approaches and frameworks, but they have no unique privilege when it comes to causal explanations or statistical characterizations of psychopathology.
The scientific critiques of DSM/ICD have revealed the neo-Kraepelinian dream to be a fantasy, but they have set the stage for new scientific characterizations of psychopathology in the form of transdiagnostic dimensional approaches, network approaches, and clinical staging approaches that are far richer in scope and potential. This is the reality that the harshest critics of psychiatry have to contend with.
See also:
Another thing many critics do is conflate notions of construct validity with notions of psychopathology and biological etiology. The lack of validity of DSM/ICD categories is taken to be an admission that judgments of psychopathology are spurious and that neurobiological causes or mechanisms of psychopathology do not exist.
Ahmed Samei Huda writes in The Medical Model in Mental Health (p 49-50):
“The purpose of classification in clinical practice is clinical usefulness. If it is impossible to delineate clearly separate categories, either because they do not exist because the condition is a spectrum or because we lack the capability to demonstrate the boundaries between conditions, then the next best option is to use a clinically useful classification based on identifying points on a spectrum. As long as people are aware that they aren’t necessarily describing clearly separate categories—either from states considered healthy or from other conditions—then, if a diagnosis can be reliably distinguished from other diagnoses and carries clinically useful discriminatory information (such as likely range of outcomes or different responses to treatment), it is a ‘worthwhile’ diagnosis.”
Thank you, Awais, for this elegant take-down of psychiatry's most vociferous and least-informed critics (there are, of course, eloquent and well-informed critics who do not seek to trash the entire clinical enterprise of psychiatry and psychiatric diagnosis).
The etymology of the term "diagnosis" is critical to understanding what psychiatrists actually do, as contrasted with what anti-psychiatry critics mistakenly believe we do. The term means, literally, "knowing the difference between" (dia-across, between; gnosis, knowledge or knowing). The psychiatric diagnostician's goal, upon first meeting the patient (from the Latin “patiens,” from “patior,” to suffer or bear) is to determine the difference between certain broad types of human experience: for example, an existential crisis vs. a psychotic break; or normal confusion brought on by stress vs. an incipient dementia.
Contrary to the puerile cliche that the DSM is "Psychiatry's Bible," we make use of the patient's symptoms and signs in a holistic, gestalt manner to arrive, initially, at a general type of human experience. As you suggest, this is often done clinically through "prototype matching". And as you say, "Prototypes are characterized by vague and fuzzy boundaries that overlap with neighboring prototypes." And there is nothing wrong or unscientific about this "fuzziness"! As Ludwig Wittgenstein once noted, a fuzzy beam of light is just as real as a sharply focused one [ Wittgenstein L: The Blue and Brown Books, New York, Harper Torchbooks; 1965].
We then try as best we can to refine our initial dia-gnosis [knowing the difference between] by means of laboratory testing (to rule out, e.g., an endocrinopathy); imaging studies (to rule out, e.g, a brain tumor); and in some cases, neuropsychological testing (to ascertain, e.g., certain personality traits). Often, we consult family members or school reports to confirm initial diagnostic impressions.
The DSM diagnostic categories, we clearly recognize, are only rough-and-ready guides to dia-gnosis, and not the alpha and omega of understanding [cf. verstehen] the patient's situation or problems. Often, we do not arrive at a "specific and certain" disease entity as the driving factor in the patient's presenting illness (if illness it be). This is also true in much of general and family practice. As family physician Kirsti Malterud, MD, PhD has noted, it is rare, in family practice, that the physician can link specific observable signs to a specific localized lesion or pathological process. On the contrary,
“The professional norm that objective signs are supposed to confirm subjective symptoms and thereby reveal monocausal disease processes falls apart in the sea of medical complexities encountered by the family physician.”
Nevertheless, Dr. Malterud notes, "...the solution of the patient’s problem can often be achieved despite the impossibility of reaching an established medical diagnosis." [Malterud K: Diagnosis-A tool for rational action? A critical view from family medicine. Atrium, Winter, 2013, pp. 26-35.]
Finally, Awais, you note that "...Many clinicians also complement the diagnosis
with a narrative diagnostic formulation that summarizes their conceptualization of predisposing,
precipitating, and perpetuating factors." I am not sure how many psychiatrists actually do this, in addition to merely proffering a DSM diagnosis--but in fact, the DSM-5 Manual itself requires such a case formulation, without which a legitimate diagnosis cannot be made. As the Manual clearly states on p. 19, "...it is not sufficient to simply check off the symptoms in the diagnostic criteria to make a mental disorder diagnosis."
And though it hardly seems necessary to say it, clinical care without a meticulous process of diagnosis is like the empty shell of a building without its core.
Respectfully,
Ronald W. Pies, MD
For further reading:
1. https://www.psychiatrictimes.com/view/science-psychiatry-and-family-practice-positivism-vs-pluralism
2. https://peh-med.biomedcentral.com/articles/10.1186/1747-5341-7-9
3. https://awaisaftab.blogspot.com/2020/06/an-exchange-with-ruffalo-pies-more-on.html