Up until ten years ago, most of what we knew about comorbidity, its prevalence, patterns and significance, was derived from clinical samples. Early research using clinical data focused on the co-occurrence of symptoms and the implications of this for diagnostic hierarchies in psychiatric classification (e.g., Foulds & Bedford, 1975; Sturt, 1981). However, a major limitation of using clinical data to determine patterns of symptom co-occurrence is that these analyses are confounded by a treatment-seeking bias. Focussing on clinical samples restricts the range of symptom presentation to the more severe cases. The development of structured diagnostic interviews and their use in large-scale epidemiological surveys have made it possible to study the co-occurrence of symptoms and syndromes across the whole spectrum of severity.

The Epidemiologic Catchment Area (ECA) study, which was the first large-scale community survey of the prevalence of mental disorders, determined that 18% of the total population or 60% of those with at least one DSM-III disorder, also had at least one other psychiatric disorder in their lifetime (Robins & Regier, 1991).The National Comorbidity Survey (NCS), which studied a probability sample of the US population aged 18-54, reported strikingly similar rates of comorbidity. Fifty six per cent of respondents with a lifetime history of at least one DSM-III-R disorder also had at least one other lifetime disorder (Kessler, 1995). Stated another way, nine out of ten severe 12 month disorders occurred in the 14% of the sample with a lifetime history of three or more disorders (Kessler et al., 1994).

With the exception of these two surveys, most of the large-scale epidemiological surveys have presented data on co-occurrence of disorders within a one, six or 12 month period, rather than over a person's lifetime. Similar findings emerge. The Netherlands Mental Health Survey and Incidence Study (NEMESIS, Bijl, Ravelli, & van Zessen, 1998) found over a 12 month period that 45% of people who met criteria for one disorder also met criteria for one or more additional disorders. In the Mental Health Supplement to The Ontario Health Survey, the figure was around 20% (Offord et al., 1996).When data from the ECA and the NCS are restricted to examine co-occurrence of disorders within a six month period the prevalence of comorbidity is similar (Kessler, 1995). In summary, multiple diagnoses, both current and past, are more common than expected from the prevalences of individual disorders and single diagnoses are less common, as though the burden of mental disorders tends to be concentrated in certain individuals. This finding is independent of country or instrument and is unlikely to be artefact.

There is much discussion in the clinical and phenomenological literature about the possible causal mechanisms underlying the clustering of disorders in certain individuals. Andrews et al. (1996; Andrews et al. 1990) studied the common neurotic disorders in volunteer twin and clinic samples and related comorbidity to the presence of a general vulnerability factor to these disorders. That is, while the clinical phenomenology may be distinct, the underlying disorders may not be. A full discussion of this issue is beyond the scope of this chapter and readers are referred to Chapter 2 in this monograph for a more detailed discussion.

When patterns of lifetime comorbidity were examined in the ECA and the NCS, three important findings emerged. Firstly, disorders within diagnostic categories were generally more commonly comorbid than disorders from different diagnostic groups. For example, major depression was most strongly associated with dysthymia (OR = 12.8 in the NCS, OR = 14.3 in the ECA) and mania (OR = 16.9 in the NCS, OR = 31.8 in the ECA) and least strongly associated with the substance use disorders (OR = 1.9 – 2.4 in the NCS, OR = 1.9 – 3.5 in the ECA) and antisocial personality disorder (OR = 2.0 in the NCS, OR = 2.6 in the ECA). Secondly, and a notable exception to this finding, odds ratios within the anxiety disorder group were generally lower than between the anxiety and affective disorders. Thirdly, despite the strong focus in the clinical literature on comorbidity between substance use disorders and affective or anxiety disorders, these were found to be among the weakest lifetime comorbidities in both the ECA and the NCS.Top of page

When data from the ECA and the NCS was restricted to six month diagnoses, patterns were similar, with six month associations between disorder pairs generally stronger than lifetime associations. Again, anxiety and affective disorders are reported as an exception to this rule. Both the ECA and the NCS report that although six month associations of affective or anxiety disorders with substance use disorder are generally higher than lifetime associations, they are not strikingly so. When anxiety and affective disorders are comorbid with substance use disorders they are likely to be concurrent, that is to be present at the same time. It should be noted, however, that associations between anxiety and affective disorders were much stronger than between either of these disorder groups and substance use disorders, a pattern also observed in the UK National Survey of Psychiatric Morbidity (Jenkins et al., 1997; Meltzer, Gill, Petticrew, & Hinds, 1995). In summary, patterns of comorbidity observed across community samples indicate that despite the current focus of treatment and policy initiatives, comorbidities between the common mental disorders and substance use disorders are not the most prevalent comorbidities.