Theoretical issues

Feinstein (Reference Feinstein1970, p. 457) offered several examples to illustrate his concept of distinct additional clinical entity – e.g. ‘If the index disease is primary cancer of the lung, the co-morbid diseases can be such entities as: episodes of mumps or bleeding duodenal ulcer that occurred antecedently; coronary artery disease or pneumococcal pneumonia detected when the cancer is discovered’. In brief, Feinstein's (Reference Feinstein1970; Kaplan & Feinstein, Reference Kaplan and Feinstein1974) discussion of comorbidity gravitates around the case of cancer with a pragmatic focus on identifying comorbid diseases or conditions that may affect diagnosis, prognosis, pathogenicity, or therapeutic outcome. Since the concept of comorbidity invariably raises differential diagnostic issues, Feinstein proposed differential diagnostic principles for assessing comorbidity, emphasizing toponymy (i.e. considering an event's anatomic location in relation to that of the index disease) and chronometric reasoning (i.e. events or conditions, prior or presently, that impact the natural course of the index disease). From Feinstein's original definition, the many examples he discussed in the article, and the character of the differential diagnostic principles, it is evident that his concept of comorbidity hinges on the notion of distinct additional clinical entities, and, most importantly, that to qualify as a distinct additional clinical entity a medical condition must have either known etiology and/or circumscribed pathology (Lilienfeld et al., Reference Lilienfeld, Waldman and Israel1994; Maser & Cloninger, Reference Maser, Cloninger, Maser and Cloninger1990; Meehl, Reference Meehl2001; Vella et al., Reference Vella, Aragona and Alliani2000; Wyrsch, Reference Wyrsch1930).

In the 1980s, Feinstein's concept of comorbidity found its way into psychiatry, where it was also used to diagnose psychiatric comorbidity, i.e. the co-occurrence of two or more mental disorders (Andrews, Henderson, & Hall, Reference Andrews, Henderson and Hall2001; Klerman, Reference Klerman, Maser and Cloninger1990; Maser & Cloninger, Reference Maser, Cloninger, Maser and Cloninger1990; Plana-Ripoll et al., Reference Plana-Ripoll, Pedersen, Holtz, Benros, Dalsgaard, de Jonge and McGrath2019; Trull, Scheiderer, & Tomko, Reference Trull, Scheiderer, Tomko and Widiger2012; Wittchen, Reference Wittchen1996). While this may seem like an intuitive extension of Feinstein's definition, it is, in fact, not unproblematic.

When it comes to psychiatric comorbidity, it is, in most cases, not clear that the founding principle of Feinstein's original definition, viz., the notion of distinct additional clinical entities, can be met. To qualify as distinct additional clinical entities, mental disorders must have either known etiology or circumscribed pathology. Most mental disorders, however, have unknown etiology. Psychiatric comorbidity is therefore typically based solely upon description of psychopathology. If mental disorders had well circumscribed psychopathology, it could potentially compensate for the absence of known etiology, and they would still qualify as distinct additional clinical entities. However, mental disorders are precisely not assumed to be completely demarcated from each other in the sense that single symptoms of one mental disorder cannot occur in another disorder – e.g. as stated in DSM-5-TR (American Psychiatric Association, 2022; cf. American Psychiatric Association, 1980, p. 6), ‘there is no assumption that each category of mental disorder is a completely discrete entity with absolute boundaries dividing it from other mental disorders or from no mental disorder’. This assumption, underlying the classification of mental disorders, is natural given that many symptoms are found across mental disorders (e.g. anxiety), and only few symptoms are considered diagnostically specific for any disorder (e.g. Mølstrøm, Henriksen, & Nordgaard, Reference Mølstrøm, Henriksen and Nordgaard2020). However, the absence of known etiology and circumscribed psychopathology jointly implies that the founding principle of Feinstein's concept of comorbidity, i.e. the notion of distinct additional clinical entities, typically is lacking in psychiatry.

Recognizing this, First (Reference First2005) proposed a distinction between three types of psychiatric comorbidity: (1) true comorbidity pertaining to cases of mental disorders that fulfil Feinstein's requirement of distinct additional clinical entities; (2) artifactual comorbidity which is a by-product of the diagnostic manuals’ decision ‘to “split” diagnostic entities into numerous specific narrowly-defined disorders’ (First, Reference First2005, p. 207); and (3) spurious comorbidity which is best avoided (e.g. comorbidity between autistic disorder and Asperger's disorder in DSM-IV). Since true comorbidity is rare in psychiatry and spurious comorbidity must be avoided, psychiatric comorbidity, First (Reference First2005) argued, primarily consists of cases of artifactual comorbidity. Consequently, First (Reference First2005, p. 206) noted that it could be misleading to speak of psychiatric comorbidity – ‘It is important to understand that comorbidity in psychiatry does not imply the presence of multiple diseases or dysfunctions’. Still, First advocated the use of psychiatric comorbidity, citing the instruction from ICD-10 (World Health Organization, 1992), recommending that ‘clinicians should follow the general rule of recording as many diagnoses as are necessary to cover the clinical picture’. However, in the context of ICD-10, this instruction does not imply that a diagnosis should be made for every symptom constellation (mental disorder) for which a patient fulfils diagnostic criteria. In ICD-10, it only pertains to symptom complexes that cannot be accommodated by the index disorder. For example, a diagnosis of schizophrenia is sufficient to cover almost all other symptoms – e.g. anxiety and changes of personality are common features of schizophrenia. Since such symptoms can be accommodated by the index disorder (schizophrenia), no additional diagnosis is typically needed ‘to cover the clinical picture’. Finally, the renouncement of the founding principle of comorbidity (i.e. distinct additional clinical entities) seems premature and leaves us with a concept of psychiatric comorbidity (artifactual comorbidity) that, strictly put, negates the very meaning of the concept of comorbidity. If the concept of psychiatric comorbidity is to be meaningful, a qualification of distinct additional clinical entities other than known etiology and circumscribed pathology is needed (see the final section for a proposal).

Another issue, complicating matters further, is the fact that the diagnostic criteria, as stated in ICD-10 (World Health Organization, 1992), ‘do not pretend to be comprehensive statements about the current state of knowledge of the disorders’ but serve only as ‘a reasonable basis for defining the limits of categories’. In other words, the purpose of the diagnostic criteria is to delimit the different mental disorders from each other; it is not to exhaustively describe their psychopathology. Although the diagnostic criteria initially were considered as ‘gate keepers’, i.e. a sort of minimum requirement for making a diagnosis, they have increasingly come to be perceived as almost exhaustive of the psychopathology of the different mental disorders (Andreasen, Reference Andreasen2007; Kendler, Reference Kendler2016a; Kendler, Reference Kendler2022; Kendler & Zachar, Reference Kendler, Zachar, Kendler and Parnas2008). This perception is, to some extent, underpinned by contemporary textbooks of psychiatry that regularly are tailored to the diagnostic criteria and therefore predominately include information about psychopathology listed as diagnostic criteria. Consequently, knowledge of psychopathology that is not included as diagnostic criteria is disappearing from contemporary psychiatric discourse and research (Parnas, Reference Parnas2011). This unintended development may also impact assessment of comorbidity. For example, clinicians or researchers may be unaware that many symptoms and signs are well-known features of a given disorder (e.g. schizophrenia) when such symptoms and signs (e.g. obsessive-like features, concentration and attentional difficulties, self-disorders, etc.; Henriksen, Raballo, & Nordgaard, Reference Henriksen, Raballo and Nordgaard2021; Rasmussen & Parnas, Reference Rasmussen and Parnas2022) are not included in the diagnostic criteria for the disorder.

The final theoretical issue that we will address here can be summarized under the heading of reification, which designates that something essentially non-thing-like acquires a thing-like status. As Kendler (Reference Kendler2016b) aptly put it, ‘It is a conceptual error to assume that our diagnostic criteria constitute rather than index our disorders’. Despite such cautionary statements, the diagnostic categories are often taken to be the mental disorders they index (Frances et al., Reference Frances, Widiger, Fyer, Maser and Cloninger1990; Hyman, Reference Hyman2010; Jablensky, Reference Jablensky2004, Reference Jablensky2005; Kendler, Reference Kendler2014). Already in the 1970s, Kendell (Reference Kendell1975) eloquently pointed to this crucial issue, ‘Because the categories and those attributed to it has a name, it acquires a shadowy ‘existence’ of its own, and it eventually comes to be assumed that its members must differ in some fundamental way from members of the other categories in the typology’. Reification is important for the discussion of psychiatric comorbidity because reification of mental disorders and psychiatric comorbidity to some extent seem to implicate or reinforce each other. Even though the diagnostic manuals repeatedly emphasize the absence of any assumption of clear-cut boundaries between the diagnostic categories, the categories nonetheless, as Kendell described, tend to reify over time (Hyman, Reference Hyman2010; Kendell & Jablensky, Reference Kendell and Jablensky2003) and come to be viewed as separate entities.

Although psychiatric comorbidity in the literature usually is defined as the co-occurrence of mental disorders, i.e. without explicitly assuming their independence (for an exception, see Caron & Rutter, Reference Caron and Rutter1991), the concept of psychiatric comorbidity nonetheless carries an implicit assumption of independence of mental disorders. This implicit assumption of independence of mental disorders permeates much research on psychiatric comorbidity. For example, a large-scale Danish register study found that all mental disorders were associated with an increased risk of all other mental disorders (Plana-Ripoll et al., Reference Plana-Ripoll, Pedersen, Holtz, Benros, Dalsgaard, de Jonge and McGrath2019). This finding must, however, be tempered by the fact that the hierarchical rules of the applied diagnostic manual (ICD-10), which exert rules for diagnosing comorbidity, were not applied in the study. Another study – the Dunedin Longitudinal Study from New Zealand – also found psychiatric comorbidity to a large extent on a lifetime basis in a complete birth cohort (Caspi & Moffitt, Reference Caspi and Moffitt2018). Here, psychiatric comorbidity was assessed ‘without regard for hierarchical exclusionary rules’ (Caspi et al., Reference Caspi, Houts, Belsky, Goldman-Mellor, Harrington, Israel and Moffitt2014). In both studies, mutual independence of all mental disorders was presumed, exemplifying to some extent a reified approach to mental disorders, and the studies’ findings of widespread psychiatric comorbidity rest in the end on an inadequate application of the guidelines of the relevant diagnostic manual. This prompts the question of whether such results are perhaps less facts than artifacts, resulting from insufficient differential diagnosis?

Clinical issues

The theoretical issues discussed above also seep into daily clinical reality, which we illustrate with two clinical vignettes.

Toward a new assessment of psychiatric comorbidity

Despite the reservations outlined above, we believe the concept of psychiatric comorbidity can be both appropriate and helpful in psychiatry. There are clinical situations, where it is relevant to make a comorbid diagnosis – e.g. a patient, who was diagnosed with a personality disorder 20 years ago, and who develops a major depression. In this case, the patient's mental condition, which has been stable for years, changes markedly with new and clearly depressive symptoms emerging. However, if the concept of psychiatric comorbidity is to be applied in a more meaningful way in psychiatry, its theoretical foundation must first be established. Inspired by Feinstein's concept of comorbidity and his differential diagnostic principles, we suggest some potential principles for an improved assessment of psychiatric comorbidity.

Since known etiology and circumscribed pathology cannot adequately qualify mental disorders as distinct additional clinical entities, we must look elsewhere for a principle to secure, as far as possible, their mutual independence. Here, it may be helpful to distinguish between two different levels of analysis: a diagnostic level and a nosological level.

The diagnostic level targets assessment of comorbidity in clinical practice and research using the existing diagnostic manuals. Here, we suggest that the distinction between trait and state conditions (see Fig. 1) can play a more crucial role in determining whether a patient's symptomatology amounts to a distinct clinical entity in addition to the index disorder. Trait conditions are mental disorders, which are defined as such in the diagnostic manuals. The symptomatology of trait conditions can fluctuate in severity, but the disorder is usually present for a longer time, often years. Examples of trait conditions can be autism spectrum disorder, ADHD, schizophrenia, or personality disorders. By contrast, state conditions are mental disorders, which are defined as such in diagnostic manuals, and which are present in one or more circumscribed episodes. Examples of state conditions can be affective disorders, e.g. major depression. In the example above, the patient has a stable trait condition (a personality disorder), when a later state condition (major depression) emerges. Crucially, the patient's more enduring symptomology is accommodated by the personality disorder, and the emerging depressive symptoms, which cannot be explained by the trait condition, are accommodated by the diagnosis of major depression. If, by contrast, we met the patient for the very first time, and the patient fulfilled criteria for both major depression and a personality disorder, we should not immediately make both diagnoses, since apparent changes of personality could be a product of the current depressive episode. In this case, we should diagnose and treat the major depression, and wait until the depression recedes to see if the changes of personality recede with it. If they do not recede, we could then perhaps diagnose a comorbid personality disorder.

Figure. 1. Illustration of state and trait conditions. (a) A state condition with a single episode of a disorder, e.g. major depression. (b) A state condition with recurring episodes, e.g. bipolar disorder. (c) A trait condition, e.g. schizophrenia. (d) A trait condition with a comorbid state condition, e.g. personality disorder with a single episode of major depression.

Applying the distinction between trait and state conditions to inform assessment of psychiatric comorbidity cannot, however, stand alone. Such an assessment ultimately rests on a comprehensive understanding of the psychopathology of mental disorders. Such an understanding, which necessarily goes beyond the limited range of symptoms and signs included in the major diagnostic manuals, forms an indispensable part of the ‘clinical judgment’ upon which diagnoses must be made (American Psychiatric Association, 2022). Such an understanding can only be achieved through varied clinical experience, reading psychopathological literature, and ongoing discussions about psychopathological phenomena with experts.

Assessment of comorbidity can be substantially improved at the diagnostic level by making suitable changes at the nosological level. Here, a classical principle for differential diagnosis can further help us restrict the use of psychiatric comorbidity. Traditionally, psychiatry has operated with a hierarchical classification of mental disorders (American Psychiatric Association, 1980), based on the clinical method (Owen, Reference Owen2023), which, in a descending order, roughly can be summarized as follows: organic disorders, schizophrenia, affective disorders, anxiety disorders, and personality disorders (Jaspers, Reference Jaspers1997, pp. 605–606, 611–612; Klerman, Reference Klerman, Maser and Cloninger1990; Owen, Reference Owen2023, p. 1704). For disorders at the top of the diagnostic hierarchy (e.g. organic disorders or schizophrenia), most, if not all, symptoms may occur. By contrast, the range of symptoms is much more limited in disorders placed lower in the diagnostic hierarchy (e.g. anxiety or personality disorders) (American Psychiatric Association, 1980). Thus, if a patient fulfils diagnostic criteria for more than one disorder, the highest placed disorder generally outranks lower placed disorders, implying that comorbid diagnoses should not be made if the symptomatology can be accommodated by the index disorder – a point vividly illustrated in the decision trees for differential diagnoses in previous editions of DSM (American Psychiatric Association, 1980, pp. 339–349, 2000, pp. 744–757) and still operative in ICD-10.

Today, emphasis on the diagnostic hierarchy of mental disorders has gradually eroded, as it has been argued that the hierarchy is empirically unfounded (e.g. Boyd et al., Reference Boyd, Burke, Gruenberg, Holzer, Rae, George and Nestadt1984; Leckman et al., Reference Leckman, Weissman, Merikangas, Pauls and Prusoff1983; Pincus, Tew, & First, Reference Pincus, Tew and First2004). This crumbling of a pillar of classical medical thinking, coupled with a somewhat blind faith in empiricism (Faust & Miner, Reference Faust and Miner1986; Kendler, Reference Kendler1990), has, however, had various ‘unintended consequences’ (Andreasen, Reference Andreasen2007; Wakefield, Reference Wakefield2016). In the case of psychiatric comorbidity, the weakening of the diagnostic hierarchy may have contributed to the inflation of psychiatric comorbidity, because symptom constellations that previously would have been regarded as mere psychopathological aspects of hierarchically higher placed disorders – e.g. anxiety or changes of personality as features of schizophrenia or major depression – now often are perceived as comorbidities. Despite going against current trends, opting for a more hierarchical diagnostic system, which is both simple to understand and easy to administer, could improve clinical practice and psychiatric research by reducing informational complexity and combating ‘compartmentalization’ of the human psyche and a superficial and partial understanding of patients’ difficulties (Aragona, Reference Aragona2009; Jablensky, Reference Jablensky2004; Widiger & Ford-Black, Reference Widiger and Ford-Black1994).

Two often invoked arguments against a diagnostic hierarchy are that it prevents a full representation of patients’ symptomatology and that it restricts research into lower placed disorders (Clark, Reference Clark2005). The first objection hinges on the erroneous assumption that the diagnostic criteria are exhaustive of the psychopathology of the different disorders. Regarding the second objection, we cannot see why research on lower placed disorders necessarily would be restricted. Moreover, it could be beneficial to study such disorders when they occur without comorbidity – is anxiety (occurring without a comorbid disorder) really the same condition as anxiety occurring in, e.g. schizophrenia?

Even without a strict hierarchical diagnostic system, psychiatric comorbidity may still be restricted using exclusion criteria (Robins & Guze, Reference Robins and Guze1970). While exclusion criteria are frequently used, they have become increasingly vague and open to interpretation, exemplified by the recurring exclusion criterion in DSM-5 stating that the disturbance is ‘not better explained’ by another disorder. How can we decide if symptoms of a given disorder is ‘better explained’ by another disorder and what is the basis of that judgment? More explicit exclusion criteria (e.g. diagnosis A must not be made if diagnosis B or C is made) may diminish the role of interpretation and restrict psychiatric comorbidity.

Finally, other criteria for diagnostic validity have been proposed (e.g. biological markers, treatment response, or risk factors) but none of them seem able to provide strong differential diagnostic tools as long as the etiology and pathogenesis of most mental disorders remain largely unknown (e.g. Robins & Guze, Reference Robins and Guze1970; and the validators for change of diagnostic criteria at the DSM proposal submission portal).