Navigating Diagnostic Pluralism

Aftab et al. (2024) in J. Nerv. Ment.

Aug 02, 2024

I have a new article out in Journal of Nervous and Mental Disease, “Psychiatric Diagnosis: A Clinical Guide to Navigating Diagnostic Pluralism,” (see pdf here) with Allen Frances as senior author, and Konrad Banicki and Mark Ruffalo as co-authors. We provide a clinical overview of recent debates surrounding categorical and dimensional approaches to psychiatric diagnosis, offer a critical assessment of proposed alternatives (especially RDoC and HiTOP), and discuss how clinicians can navigate the current plurality of diagnostic frameworks.

I have seen prior discussions of these diagnostic alternatives, but the articles have tended to focus on research contexts, scientific considerations, or practitioner familiarity with these frameworks, and we wanted to focus on how clinicians can conceptualize and productively approach their concurrent use. We also draw on published literature on clinician attitudes and preferences to guide our discussion. This paper has had a long gestation period and spent a long time under review and in press. I am glad to see it published.

The emphasis is quite clinical and pragmatic. We are not trying to answer which framework best represents the statistical or mechanistic structure of psychopathology. We are trying to answer: we have these different classificatory frameworks available, what general principles can help clinicians use them in a complementary way? If you want to jump to the crux of the discussion, I recommend the sections on categorical vs dimensional approaches, clinician attitudes, and navigating diagnostic pluralism.

I’m posting some excerpts from the article here for your convenience, and hopefully these will entice you to read the full text.

The use of descriptive, operationalized criteria in DSM-III emerged in the context of concerns around the widespread lack of reliability of psychiatric diagnoses among psychiatrists. The development of operationalized diagnostic criteria was intended, in part, to remedy such discrepancies. When DSM-III was published in 1980, it was described as “atheoretical” with regards to etiology, a statement that was intended to emphasize its descriptive nature. However, descriptions always exist within a context of shared background theoretical assumptions (Zachar et al, 2023), and DSM-III was implicitly embedded within a neo-Kraepelinian approach to psychopathology (Blashfield, 1982; Klerman, 1978), with the assumption that studying psychopathology using descriptive, operationalized criteria will eventually lead to a convergence of validators and discovery of underlying disease entities (Rounsaville et al, 2002). 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, and 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.”

Categorical vs dimensional approaches

The choice between categorical and dimensional approaches can be conceived of as referring to an optimal strategy of making the complexity of clinical phenomena tractable for clinical and research purposes, either through reliance on a finite number of classes (categories) or by the employment of a finite set of qualitatively specified but continuous and quantifiable dimensions. Categorical approaches to classification have historically been dominant in psychiatry (and medicine), but recent decades have seen a burgeoning scientific development of dimensional models of classification. The choice between categorical and dimensional models is, however, not simply a pragmatic issue. Psychiatric categories may be conceptualized and measured either categorically or dimensionally, but whether the statistical latent structure is truly categorical or dimensional is a empirical issue to be discovered (Ruscio & Ruscio, 2008). The scientific question about the latent structure of psychopathology is therefore distinct from the question of the clinical use of categorical vs dimensional models.

Classifications in medicine and clinical psychology are pragmatic tools created to guide and structure clinical work. Since many clinical decisions are binary, categorical classifications are often conducive to clinical work in a manner that dimensions are not. In research contexts, categorization of dimensional measures entails some loss of data, and consequently, dimensional approaches often allow for greater statistical power, that is conducive to scientific research. The choice between dimensions and categories, however, can be deceptive as dimensions can easily be converted in categories based on practically relevant thresholds. Many general medical problems exist as dimensions but physicians treat them categorically. For example, blood pressure and weight are dimensions, but essential hypertension is diagnosed as a category using a threshold of 140/90 mmHg, and obesity is often diagnosed using cut-offs on the body mass index. Correspondingly, categorical classifications can incorporate elements of dimensionality, such as by conceptualizing the condition as existing on a spectrum and by specifying different levels of severity (subthreshold/mild/moderate/severe). In this sense, the goal of dimensional systems in psychiatry is not the elimination of categories, but rather the delineation and optimization of clinical cut-offs given the dimensional nature of the phenomena.

It is important to highlight that a difference in quantity can be consistent with the emergence of a qualitative or categorical difference on that continuum. For instance, a normal person may experience hypnagogic or hypnopompic hallucinations and an anxious person might believe that their coworker is talking behind their back, but there are qualitative differences between such experiences and the psychotic states observed in schizophrenia spectrum disorders, even though these may be rated continuously on a symptom scale. Experiences of impairment and disability on symptom dimensions are one example – disability may only be experienced only if the symptoms are sufficiently severe. And from the perspective of network dynamics, symptoms may become self-sustaining via mutual feedback loops above a certain threshold of symptom activation but not below it. Dimensional approaches, therefore, in principle, allow for qualitative differences to exist on a quantitative continuum, and different neuropsychological mechanisms may be involved in their realization.

The scientific question of whether the latent structure of psychopathological conditions is discrete (categories) or continuous (dimensions) remains unsettled; there is no convincing evidence yet of discrete categories in psychopathology (Haslam et al, 2020). However, the notion that categories and continua are mutually exclusive has itself come under scrutiny (Borsboom et al, 2016). There are certain classical notions of what categories and dimensions look like (e.g., Kraepelinian disease entities and Intelligence Quotient respectively), but it is possible that psychopathology could be described in some ways as categorical (involving within-person dynamic phase transitions over time), in others as dimensional (symptom distribution in a population) (Eaton et al, 2023).

Research Literature on Clinician Attitudes Towards Classification

To summarize, these surveys reveal:

Clinicians prefer flexible diagnostic guidelines over strict criteria, suggesting a preference for a clinical prototype approach over strict operationalization. Clinicians generally support incorporating dimensional elements into categorical systems such as DSM and ICD.
Many mental health professionals, although they make use of diagnoses, do not routinely refer to specific diagnostic criteria in the manuals.
Clinicians want diagnostic guidelines to facilitate communication and guide treatment.
The majority use DSM/ICD for administrative/billing purposes, even when they are otherwise dissatisfied with the manual.
A simplified version of HiTOP appears to be acceptable to psychologists and offer some advantages over DSM when it comes to conceptualizing diagnostically complex cases with multiple comorbidities. It is unclear at the moment how medical professionals (including psychiatrists and non-psychiatric professionals such as primary care physicians) would assess HiTOP’s clinical utility compared to DSM and ICD.
Theoretical orientation of mental health professionals, especially psychologists, influences whether they find DSM to be useful and what alternatives they prefer. No current diagnostic system is highly endorsed by psychologists across all theoretical orientations.

Given the obvious limitations of psychological formulation in serving the pragmatic tasks that DSM and ICD currently serve in the system, attempts to eliminate psychiatric diagnosis altogether are unrealistic and harmful in our opinion. Diagnoses such as schizophrenia and bipolar disorder, while they may not reflect the existence of categorically distinct disease entities and do not provide etiological explanations, nonetheless capture important aggregate differences in clinical presentation and diagnostic validators; they allow for a convenient way to convey probabilistic information necessary for competent clinical care, such as risk factors, longitudinal course, prognosis, and available treatments. In due course, they will be replaced by diagnostic constructs that are able to capture such information more efficiently and powerfully. It is the case that psychiatric diagnoses often generate significant societal and professional discrimination and stigma, but these problems are not inherent to the classification itself. When utilized appropriately with clinical judgment, complemented by a comprehensive clinical characterization, and awareness of limitations, psychiatric diagnosis serves to enhance our understanding of the patient, informs treatment, and provides a useful language to communicate about similar problems across the population.

Navigating Diagnostic Pluralism

Diagnoses, whether categorical or dimensional, offer condensed information about a patient’s presentation, but they cannot substitute for a comprehensive clinical characterization.
Diagnosis should be based on a comprehensive clinical evaluation, and when we think of diagnosis, we should also think of diagnosis as a process (and not simply as a label).
While operationalized criteria are useful for research purposes, clinicians tend to use prototypical descriptions which allow for fuzzy boundaries and flexible application, and which do not make strict realist assumptions about the nature of the categories.
Just as DSM diagnoses can be made in a variety of ways in clinical and research settings (clinical interview, rating scales, and standardized interviews such as SCID), alternatives such as HiTOP should allow for flexible use to facilitate clinical uptake.
DSM, HiTOP, RDoC, PDM, and other frameworks are not mutually exclusive; they can be viewed and should be used in a complementary manner. In a manner of speaking, DSM is a coarse-grained categorical approximation of HiTOP’s dimensional and hierarchical schema. While DSM constructs are not officially included in HiTOP, clinicians can use the syndromic level, in between symptom/traits and subfactors, to translate between the two frameworks. HiTOP hierarchical elements can also be incorporated into DSM constructs, and it’s not uncommon, for instance, for clinicians and researchers to categorize DSM categories as belonging to internalizing and externalizing disorders even though internalizing and externalizing spectra are not officially part of the DSM.
Clinicians are generally agnostic on the issue of the “latent structure” of psychopathology. Categorical diagnoses need not correspond to the latent structure for them to have clinical utility or for them to carry clinically useful discriminatory information. Clinical utilization of categorical or dimensional classifications is a separate issue from our best scientific understanding of the nature of psychopathology. Clinicians care about what framework better allows them to assess the patient and match the patient to available treatment options. DSM/ICD have the historical advantage here given that nearly all available practice guidelines and regulatory approval of medications refer to DSM/ICD categories. Thus, even when clinicians are using HiTOP, they will have to refer at present to DSM/ICD diagnoses to ensure practice guidelines are being adhered to and to determine the appropriate pharmacological treatment. While it is true that pharmacological treatments such as SSRIs have broad effects across the internalizing spectrum, there are important differences in pharmacological treatment options for disorders within the internalizing spectrum that require being mindful of the diagnostic distinctions between, say, major depression, generalized anxiety, and obsessive-compulsive disorder. Furthermore, clinically important distinctions, such as being unipolar depression and bipolar depression, are currently not captured by HiTOP.
Dimensional approaches often look at symptom distribution in a population, however, clinicians are interested in individuals whose status as disordered changes dynamically with time. A depressed patient, for example, is not always depressed. Dimensional distribution of depression symptoms in a population has little direct relevance to that. Scientific approaches that take into account such “phase transitions” between depressed and non-depressed states are valuable and emphasize categoricity of dimensional phenomena at an individual level.
Fine-grained dimensional characterizations do not generally fit well with the flow of clinical work in contexts such as brief psychiatric appointments or in the emergency room evaluations. In such contexts the lesser cognitive and administrative burden of DSM/ICD (including the use of “not otherwise specified” categories) while guiding treatment is a distinct advantage.
One of the crucial tasks of DSM and ICD diagnoses is to rule out medical and neurological diseases as well as substance intoxication and withdrawal as causes of psychiatric symptoms. Dimensional systems and formulation-based alternatives have not so far formally included such considerations of differential diagnosis within the frameworks.
Developments in classification should be complemented in their conceptualization with theoretical developments in areas such as network theory, clinical staging, and phenomenological psychopathology.
At present, views regarding the clinical utility of diagnostic frameworks are heavily influenced by the theoretical orientations of the clinicians.

Conclusion

Advancing psychiatric knowledge and enriching clinical practice requires the profession to embrace a plurality of diagnostic and explanatory frameworks. Researchers have had to develop new frameworks to guide neuroscientific research efforts because it has become clear that limiting research studies to DSM and ICD categories has hindered scientific progress, but there nonetheless remains a need for an official classification—a role presently fulfilled by DSM and ICD—that provides a shared language for clinicians and researchers as well as permit other tasks such as administrative documentation, healthcare provision, insurance reimbursement, and legal and forensic use. Alternatives approaches such as RDoC and HiTOP are not yet able to serve these clinical needs across all relevant contexts in a manner that would make DSM and ICD redundant. However, frameworks such as HiTOP and PDM can be productively used in conjunction with DSM and ICD in clinical settings and they offer various advantages to clinicians. It is important for the psychiatric professions to make clear the limitations and many deficiencies of DSM/ICD approach, but to also recognize the difficulty of developing better alternatives that can completely replace DSM/ICD in clinical settings (Reed, 2018). For the near future at least, we are looking at a landscape of nosological pluralism, where multiple classification systems and frameworks will be employed for different settings and different purposes. In such a context, the need for a shared language to communicate and translate information remains essential; the wide acceptance of DSM and ICD has largely been due to their success at providing this common language (Pichot, 1994). A recurrent lesson in the history of psychopathology is one of humility, and it is better for nosologists of today to under-promise and over-deliver rather than vice versa.

See also: