The Explanatory Value of Descriptive Diagnosis

There is a legitimate sense in which depression can be said to affect how people think, feel, and act

On social media and blogosphere, a new round of the chronic debate about the errors of invoking diagnoses as causes of symptoms recently took place. This was partly driven by the publication of a recent paper by Kajanoja & Valtonen in Psychopathology (see text here) in which assertions of the form “depression causes a variety of symptoms, such as low mood and anhedonia” and “depression affects how people think, feel and act” on various medical websites are treated as misleading and inaccurate and are portrayed as instances of “circular reasoning” in which descriptions and causal explanations are conflated.

As is commonly my assessment of the state of popular discourse, the debate is needlessly polarized, partly because of ideological motivations and partly because people continue to fundamentally misunderstand the nature of psychiatric diagnosis.

There is a perfectly legitimate sense in which depression can be said to affect how people think, feel and act, and in my view, it is misleading and inaccurate to assert otherwise. I have written about this issue before, but this provides an opportunity to recapitulate where this debate goes wrong. (Also, see a response by Pies & Ruffalo in Psychiatric Times¹, and some general comments by Gipps in this thread on twitter)

The first mistake is treating operationalized criteria as constitutive of the condition itself.

The diagnostic criteria in official manuals such as the DSM and ICD are simply indices, as a way of pointing towards and recognizing the existence of a state, a condition, a syndrome, a property cluster, etc. They do not constitute the condition itself (see Kendler 2017). It is not the case that “depression” is nothing more than the symptom criteria in diagnostic manuals. Rather, the criteria are a way for us to recognize the condition of depression.

There are two ways in which “depression” as a condition is more than the symptom criteria:

• there are well-recognized symptoms and clinical features of depression that are not present in the diagnostic criteria

• there is a network of causes that generates the experienced and observed clinical features of depression but this network is not included in the criteria.

The second mistake is in thinking that because diagnostic criteria are descriptive, the referent can only be something descriptive.

Diagnostic terms can simultaneously have two different referents: the clinical presentation and the hypothesized mechanisms and processes that are responsible for the clinical manifestation (descriptive and causal respectively). There is nothing particularly problematic or inconsistent about that, as long as we are clear on how the terms switch back and forth between such referents (see Maung 2016). The diagnostic criteria can point towards not only the cluster of all recognizable and observable properties but also the poorly understood causal network. It is referring to the latter that allows us to say things like, “This person doesn’t have a history of mania or hypomania, but XYZ features lead me to suspect this is a case of bipolar depression.” That is, we are hypothesizing that some of the same mechanisms and processes at play in bipolar disorder are at play in this case of depression, even though the clinical presentation doesn’t meet official diagnostic criteria.

The third mistake is in failing to recognize form, structure, and organization as instances of non-trivial explanation.

These are situations where explanations are about the relationships between parts and wholes.

Kajanoja and Volten write: “… a medical condition can be the cause of symptoms only if the condition is conceptually independent from them. For example, a tumor in the lung can cause symptoms such as cough, fatigue, and shortness of breath, because a tumor is not synonymous with the symptom list itself.”

The problem here is that Kajanoja and Volten are thinking of “conceptually independent” only with reference to some process that is spatiotemporally external to the phenomenon in question. But part-whole relationships are not like that. A whole is not external to its parts, even though it is not identical to a mere itinerary of parts in isolation. This is because a whole also includes the interactions and relationships between parts that often generate novel properties (see Mental Causation and Mental Causation and Metaphysical Gloss). A state of depression as a clinical entity is a whole of a which individual features such as low mood and anhedonia are a part.

Descriptive diagnoses refer to an identifiable pattern of problems reported by the individuals and features observed by the clinician (symptoms and signs). When characterizing a cluster of symptoms as a syndrome, there is an assumption that the pattern is not random or a chance co-occurrence, and that the co-occurrence and co-variation is legitimate. This by itself is not a causal explanation but it is the beginning of one: it points towards the need for something to be explained. It is not merely a convenient shorthand for a list of symptoms, because there is a relationship between the symptoms. There has been a tendency to assume that the pattern exists because of a “common cause” or a “latent variable” which produces the symptoms, but the scientific community is much more aware now that explanations can take other forms (e.g., symptom networks in which symptoms reinforce each other to create a self-sustained feedback loop in the absence of a central mechanism).

It is often said that descriptive psychiatric diagnoses don’t explain anything… This is not quite true. The criticism conflates etiological or causal explanation with other forms of explanation. One such explanation can be described as pattern recognition and matching (“formal cause,” if you prefer thinking in Aristotelian terms). “Your presentation matches this well-recognized pattern of symptoms, and not these other patterns, and we can use this pattern to link your presentation with existing knowledge from clinical experience as well as research that has been conducted on this pattern and treatments that have been investigated for this pattern.” This may not be an etiological explanation, but it is not simply a tautological repacking of symptoms.²

Richard Gipps said in a tweet, “To contrast descriptions of X with explanations of X is not always helpful. After all, a frequent way we explain Xs is by redescribing them. Why did you move that piece of wood over there? I'm playing chess and this is me moving my queen out of danger.”

This is right on the mark, because a frequent way we use descriptive diagnoses as explanations in the clinic is to redescribe in clinical language what the person so far has described using everyday concepts of agency, discipline, and willpower. This is why accurate diagnoses of conditions such as autism and ADHD can offer powerful explanations for patients (for a good example of this, see this recent piece by the writer Mary H.K. Choi on her autism diagnosis at the age of 43, “What My Autism Diagnosis Finally Explained.”)

The fourth mistake is disregarding the fact that symptom patterns are connected with other information of interest to us.

These are known as validators, and include things like genetics, family history, personality traits, risk factors, course of illness, treatment response, etc. An accurate descriptive diagnosis connects an individual person in the clinic with a large body of clinical and scientific information. A difference between ordinary language descriptive explanations and clinical/scientific talk (e.g. “being an asshole,” vs having narcissistic personality disorder) lies in the iterative cycles of operationalization, validation, and refinement that occur in the process of clinical practice and research.

The fifth mistake is in overlooking that ordinary language retains the use of “cause” as a way of saying “this is how it best makes sense” or “this is how it is best described.”

This is relevant when it comes to communication in ordinary language, but it is also linked to the use of formal cause in clinical language, using “cause” in the sense of pattern recognition and matching. It is not tautological or circular to say that an economic recession is a cause of “high unemployment rate” even though high unemployment rate is one of the attributes of recession. This is because a high unemployment rate can be occur in a variety of contexts, and an economic recession is just one of them, and by saying that the cause is economic recession, we’ve zeroed in on the correct context. Similarly, anhedonia can occur in a variety of conditions, including schizophrenia, dementia, PTSD, adjustment disorder, and to say that, anhedonia is caused by major depression offers information not contained in the symptom itself. This becomes particularly relevant when we are talking about symptoms that are not a part of the diagnostic criteria. To say that irritability is caused by depression is to say that irritability in this case is best understood and treated as a part of the depressive symptom cluster, rather than mania, anxiety, or psychosis because irritability can be a feature of any of them.

The common idea that diagnosis offers no more information that the patient’s presenting complaints is also not true. This fails to appreciate that diagnosis is the end product of a diagnostic evaluation. A diagnostic evaluation is a careful consideration of the various possibilities relevant to a person’s complaints. A patient may present with depressed mood but the psychiatrist may arrive at a diagnosis of “Obsessive Compulsive Disorder” or “Alcohol Use Disorder” based on a careful history and evaluation, even when the patient may not have been thinking in those terms. A patient may present with anxiety, and get a diagnosis of a psychotic disorder. If we don’t take into account the process of diagnostic evaluation, we fundamentally fail to grasp what a diagnosis is and what it does.

The sixth mistake (and this one is usually made by defenders of psychiatric diagnosis) is an inadequate appreciation of the fact that casual use of causal language is not benign.

Talking about psychiatric diagnoses as causing things often encourages an essentializing interpretation such that people end up thinking that “depression” or “generalized anxiety” refer to some discrete physiological entity external to the symptoms — a dysfunctional neurobiological process, for example — that causes the symptoms. I have written previously about how the public misunderstands the medical model, and I won’t repeat it here, but a great contributor to that is the common use of causal language without appropriate clarification. Preventing such misunderstanding require being very explicit of the limitations of descriptive diagnosis, which is why I hammer this issue again and again on this blog.

People Are Stumbling From One Misguided Narrative About the Medical Model to Another

The seventh mistake (and this isn’t technically a mistake, more of an annoying tendency) is using critiques of descriptive diagnoses as a way of smuggling in one’s favored etiological explanations.

Kajanoja & Valtonen write: “An alternative non-circular approach is based on the fact that while no unified biological pathological process has been found that causes depression, the number of stressful life experiences is significantly associated with the likelihood of experiencing depression. Some authors therefore emphasize that depression can be described as an adaptive response or a functional signal to adverse circumstances. Contrary to the erroneous causal beliefs that circular claims promote, this approach underlines that low mood and/or loss of pleasure are often meaningful reactions to life events, and that they can be meaningfully understood. This approach is not only scientifically more accurate than claiming that symptoms are caused by their descriptive label, but it also has important beneficial effects for patients in practice… it would seem psychologically advantageous if health authorities, instead of promoting the circular causal claims identified in our study, either simply made clear that the diagnosis is merely a description, or, alternatively, advised people to construe their symptoms as a functional, adaptive reaction that can be meaningfully understood and responded to.”

The very reason why descriptive approaches to psychopathology remain essential is because there is no clinical or scientific consensus on the nature and etiology of psychopathology. It is odd to argue for “pure descriptions only!” and simultaneously say, “We should explicitly characterize depression as an adaptive response to adversity” (see my prior discussion of framing depression as a functional signal). Low mood possibly evolved as an adaptive response, just as pain evolved as an adaptation, but that doesn’t mean that it is accurate to characterize most clinical states of depression (i.e. depressive disorders) as adaptive. The paradigmatic cases of depressive disorders are anything but adaptive and functional.

This illustrates that some people are opposed to explanatory language around descriptive diagnoses because it encourages the kind of etiological interpretation they disfavor, and not because it violates descriptive purism and encourages etiological interpretation per se. This is also why many in the medical community have been complacent with regards to the use of explanatory language, because they hold neo-Kraepelinian views themselves; they don’t see it as a big deal if people are mistakenly left with the impression that diagnoses refer to discrete disease entities. From my perspective, the solution is neither a complacent reliance on ‘disorder’ language nor some uncritical acceptance of an ‘adaptive and functional response’ narrative; doing justice to the clinical consensus requires an explicit recognition of the dimensional nature of symptom distribution, the prototypical nature of clinical diagnosis, the necessity of explanatory pluralism, and the complex, multilevel, multifactorial etiology wherein mental health problems (and the processes that generate them) exist in a web of temperament, psychological development, behavioral learning, genetic vulnerability, and social adversity.

---

1. I disagree with the ‘volcano eruption caused Pompeii’s destruction’ analogy in the article by Pies & Ruffalo. Volcano eruption causing Pompeii’s destruction isn’t akin to depression causing low mood; it’s more akin to depression causing unemployment. Pompeii isn’t a descriptive component of the phenomenon of volcano eruption in the way “lava flow” is.

2. I have recycled some of this content from a previous post, Psychiatric Diagnosis: A Reintroduction