Adolescent Depression and Externalizing Problems: Testing Two Models of Comorbidity in an Inpatient Sample

Article excerpt

In children, psychiatric comorbidity is more the rule that the exception. However, from a treatment perspective, comorbidity presents a serious dilemma: do you separately target each type of problem or is a qualitatively different approach to treatment needed? Research has not provided a sound solution to this question because it has not adequately differentiated between additive and interactive effects of multiple disorders or child dysfunction. In an extensive review of this literature, Nottelmann and Jensen (1995) cite a number of methodological difficulties but overlook this endemic problem. The typical research strategy is to compare levels of dysfunction among four groups: morbid for disorder A, B, both A and B, and (for comparison purposes) neither A nor B. This strategy confounds "additive" (A + B) with "interactive" (A x B) models of comorbidity, obscuring distinctions between quantitative and qualitative processes. A second issue that Nottelmann and Jensen do address is the paucity of studies utilizing continuous (rather than categorical) measures of psychopathology, suggesting that both are needed to fully appreciate the implications of comorbidity in child and adolescent populations.

In this study, we use data on depression and externalizing psychopathology in an adolescent inpatient sample to compare the utility of an additive versus interactive model of comorbidity using both categorical and continuous measurement procedures. Among numerous researchers who have sought to clarify differences among children with "pure" or "mixed" patterns of internalizing and externalizing difficulties, there is a growing realization that, in young people especially, these qualitatively different expressions of psychological dysfunction frequently co-occur (Overbeek et al., 2001; Zahn, Climes-Dougan, & Slattery, 2000). In the general population, rates of comorbid depression and behavioral problems typically increase from childhood to adolescence (e.g., Esser et al., 1990). These rates are further inflated in adolescent inpatient samples by referral and admission biases (Nottelmann & Jensen, 1995). In fact, conservative estimates suggest that at least a third of all adolescents hospitalized for psychiatric conditions simultaneously present with depression and conduct disorder (Puig-Antich, 1982; Ryan et al., 1987). We took advantage of high prevalence rates in hospital samples to examine whether adolescent inpatients with comorbid depression and externalizing pathology might differ in degree or kind from adolescent inpatients with one or the other. Related studies mostly have focused on implications of comorbidity for course and outcomes of the disorders (see Nottelmann & Jensen's, 1995, review; see also Somersalo, Solantaus, & Almgvist, 1999), relying on the four-group approach described above. In contrast, in this study we used a multifactorial design to distinguish additive from interactive processes, and focused on potential differences between comorbid and single-disorder groups in personality and functional impairment. An additive model posits that personality or functional difficulties associated with either depression or externalizing problems (or both) will be found in linear combination among individuals with both disorders. Alternatively, an interactive model predicts nonadditive differences in dysfunction, with comorbid individuals reporting higher (or possibly lower) levels of dysfunction than predicted from a simple combination of problems associated with depression and externalizing difficulties alone.

To appreciate the importance of this distinction, consider a relatively early but relevant study of comorbid internalizing and externalizing problems, which involved assessing teachers' and mothers' ratings of problem behavior among children identified as withdrawn, aggressive, both, or neither (Ledingham, 1981). Post hoc comparisons of the four group means showed that, on several dimensions, adolescents with both types of difficulties had more problem behaviors that those in the single-disorder groups, with the single-disorder groups differing from controls and often from one another as well. …