Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Curr Opin Psychiatry. Author manuscript; available in PMC 2013 July 11.
Published in final edited form as:
PMCID: PMC3708605

The role of latent internalizing and externalizing predispositions in accounting for the development of comorbidity among common mental disorders


Purpose of the review

Although numerous studies have examined the latent structure of internalizing and externalizing mental disorders, the effects of this structure in predicting the development of comorbidity have remained unexamined until recently.

Recent findings

A novel approach to study the effects of latent internalizing and externalizing predispositions to the development of comorbidity was used to analyze data from 14 countries in the WHO World Mental Health (WMH) Survey Initiative. Pervasive significant positive associations were found between temporally primary and secondary internalizing and externalizing disorders in survival analyses, with time-lagged associations consistently stronger within-domains than between-domains. The vast majority of these associations were explained by latent internalizing and externalizing variables. Specific phobia and OCD were the most important internalizing components and hyperactivity disorder and oppositional-defiant disorder the most important externalizing components. Several intriguing residual time-lagged associations remained significant, though, even after controlling latent predispositions.


The latent variable model suggests that common causal pathways account for most comorbidity among internalizing-externalizing disorders. These pathways should be the focus of future research on the development of comorbidity, although isolation of consistent residual time-lagged associations between certain pairs of primary-secondary disorders is also important in pointing the way to subsequent focused study.

Keywords: Comorbidity, Epidemiology, Risk factors


Comorbidity is the norm among common mental disorders, as more than 50% of people with a mental disorder in a given year have multiple disorders [1, 2]. The structure of this comorbidity has been the subject of considerable interest over the past decade. Beginning with an influential paper by Krueger [3], numerous researchers have used factor analysis to document that associations among hierarchy-free anxiety, mood, behavior, and substance disorders can be accounted for by correlated latent predispositions to internalizing (anxiety, mood) and externalizing (disruptive behavior, substance) disorders. The internalizing dimension is sometimes further divided into secondary dimensions of fear (e.g., panic, phobia) and distress (e.g., major depressive episode, generalized anxiety disorder) [4-9]. These results have been used to argue for a reorganization of the classification of mental disorders in the DSM and ICD diagnostic systems [10-14], although other data suggest that this theoretical structure might be insufficiently robust to serve as the basis for such a reorganization [4, 15]. For example, the distinction between fear and distress disorders does not emerge in all studies [4, 15-17] and model fit deteriorates when additional disorders are added or when the model is estimated separately among people at different life-course stages [13, 15].

Despite these inconsistencies, the general finding of strong comorbidity within the internalizing and externalizing domains has raised the question whether common risk factors exist for the disorders in either of these domains and, if so, whether risk factors for individual disorders documented in previous studies are actually risk factors for these broader predispositions. The issue of specificity vs. generality of risk factors is of considerable importance, as a number of hypotheses about causal pathways posit the existence of very specific associations between particular risk factors and particular outcomes. These interpretations would be called into question if empirical research showed that the risk factors had less specific predictive effects [18]. In addition, evidence that a risk factor had a broad effect on a wide range of disorders would increase interest in that risk factor as an intervention target [19].

Although the use of latent variable models to study risk factor specificity is only in its infancy, research already has shown that this line of analysis has considerable value in distinguishing specific versus nonspecific risk factors. For example, Kramer and colleagues [20] found that the widely-observed association of gender with depression became insignificant when controls were included for latent internalizing and externalizing dimensions, arguing that gender is more directly associated with these overall latent dimensions than with depression or any other disorder within these dimensions. In another example, Kessler and colleagues [21] found that the effects of childhood adversities on onset of individual mental disorders were largely mediated by more direct effects on predispositions for internalizing and externalizing disorders.

One special class of latent variable risk factor studies uses samples of twins to estimate the effects of genetic factors on comorbidity. These studies have shown that much of the comorbidity between particular pairs of mental disorders in epidemiological samples, such as between eating disorders and substance use disorders [22] or nicotine dependence and major depression [23], can be explained by a latent variable model that assumes the existence of genetic influences. More elaborate studies of a related sort have shown that much of the comorbidity among anxiety disorders [24] and among personality disorders [25] can be explained by models that assume the existence of genetic influences. Other studies have shown that the inter-generational continuity of childhood-onset externalizing disorders can be explained by a model that assumes the existence of genetic transmission [26] and that decomposition of factor analyses into separate additive genetic and environmental components finds stable internalizing and externalizing factors only for genetic, not environmental, influences [27**, 28].

It is important to note that the findings of strong genetic influences on comorbidity are constrained by the additivity (i.e., no interactions between genetic and environmental effects) and equal environment (i.e., comparability of environmental similarity between identical and non-identical twins) assumptions that are needed to identify the coefficients in standard behavior genetic models. These assumptions have long been the subject of controversy, especially the additivity assumption [29]. Great care is consequently needed in interpreting these results because of their sensitivity to these assumptions [30**]. An additional important implication, even if we are prepared to accept the results of variance-covariance decomposition analyses based on twin studies, is that the term “genetic” has a much broader meaning than typically appreciated. For example, as noted famously by Lewontin many years ago [29], a genetic effect on tryptophane metabolism that has mediating effects through “melatin deposition to skin color to hiring discrimination to lower income” would emerge in a standard twin analysis as documenting strong “heritability for ‘economic success’” even if the true driving force behind the association was hiring discrimination based on skin color.

The risk factor studies described above treated latent measures of internalizing and externalizing predispositions as independent variables in causal models that predict individual disorders. Most of these studies use cross-sectional data and assess comorbidity at a point in time. Although several other studies have used longitudinal data to determine whether the structure of internalizing and externalizing disorders is stable over time [9, 15, 16], none tried to predict onset or persistence of disorders prospectively. Other longitudinal studies examined temporal progression [31-34] or sequencing [35-39] between earlier and later disorders, documenting strong persistence of disorders over time and predictive associations between some but not other temporally primary and later disorders. Again, though, none of these studies investigated the extent to which associations of earlier disorders with later disorders were explained by latent internalizing or externalizing variables. For example, Fergusson and colleagues [31] found that childhood conduct disorder but not ADHD predicted subsequent onset of substance disorders, while Beesdo et al. found that temporally primary social anxiety disorder predicted subsequent onset and persistence of major depression [40], but did not study whether these associations were due to effects of latent internalizing or externalizing predispositions.

Previous reviews of the developmental psychopathology literature suggest that analysis of the effects of latent predispositions to mental disorders on onset and progression of individual disorders could be very useful in identifying modifiable risk pathways [41, 42]. The confirmatory factor analysis approach that has dominated the literature on latent variables in comorbidity does not allow this kind of investigation. However, a new approach to the analysis of comorbidity makes this possible. This new approach is exposited here and illustrated with an analysis of predictive effects between latent predispositions to internalizing and externalizing disorders and subsequent first onset of Axis I disorders in the WHO World Mental Health (WMH) Surveys [43**].


Factor analytic studies of comorbidity decompose correlation matrices among point-in-time disorders to study the structure of cross-sectional associations. However, point-in-time prevalence is a joint function of lifetime risk and persistence. Factor analysis cannot break prevalence estimates into these two components. When data are available on age-of-onset (AOO) and persistence of multiple disorders, though, this decomposition can be made by using survival analysis [44] to carry out separate studies of (i) the associations of prior lifetime disorders with subsequent first onset of some other disorder and (ii) the associations of lifetime comorbidity with persistence of that other disorder. Backwards recurrence models can also sometimes be used to study predictors of persistence [45].

Consider a situation where we are studying comorbidities among D disorders with a focus on predictors of first lifetime onset. We would have D survival equations (i.e., one to predict onset of each disorder). There would be D-1 predictors in each equation (i.e., one predictor for prior lifetime occurrence of each other disorder at time t to predict onset of a focal disorder between times t and t+1) and D × (D-1) coefficients across all the equations. (Figure 1) The latent variable formulation we propose, in comparison, assumes that these coefficients are mediated by latent predispositions to internalizing and externalizing disorders that can change between times t and t+1. (Figure 2) As these two models are nested (i.e., the model in Figure 2 is a special case of the model in Figure 1), it is possible to compare model fit using standard fit indices. It is also possible to modify the model to allow for direct effects of some temporally primary disorders on subsequent onset of some secondary disorders. The model can also be elaborated to consider more than two latent variables.

Figure 1
Schematic of the multivariate observed variable model1
Figure 2
Schematic of the multivariate latent variable model1

This latent variable model cannot be estimated with the covariance structure analysis approach used in previous studies of the structure of comorbidity, as the number of person-years in the survival analysis varies across outcomes. It is also important to note that model does not assume a factor analytic structure in which the latent variables cause the observed disorders and the prediction errors for the observed disorders are conditionally independent. Instead, we assume that the observed disorders are the predictors of the latent variables. These predictors can be inter-correlated at time t because of joint influences. In this way, the model assumes that the latent variables represent common pathways by which the time t predictors influence multiple outcomes at time t+1.

Estimation of model coefficients is complicated by the fact that we are dealing with survival models across a range of outcomes in which first onsets vary from person to person and from year to year. Iterative methods are consequently needed to estimate model coefficients. We did this using a discrete-time (i.e., person-year) framework where each value of t represents one year of the respondent's life. We consider each of three main parts of the model separately (time t observed variables predicting time t latent variables, time t latent variables predicting time t+1 latent variables, time t+1 latent variables predicting time t+1 observed variables), estimate coefficients only in one of these three parts at a time while fixing coefficients in the other two parts to their values in the most recent iteration, and then repeat this process sequentially until estimates converge. This yields maximum-likelihood estimates of model parameters. A likelihood-ratio χ2 test can be used to compare model fit with the observed variable model.

This model can also be estimated at separate points in the life course. And it can determine if particular pair-wise associations between observed time t lifetime disorders and onset of outcome disorders at time t+1 are significant after controlling the latent variables. The latter can be done by using empirical estimates of time t latent variables (generated from model coefficients) as controls in separate bivariate survival equations that predict first onset of each disorder from prior history of each of the other disorders. And, of course, this entire system of associations can be included in more complex models that examine other predictor associations (e.g., the association of gender with first onset of major depression) to determine if the latent variables account for those associations.


As reported in more detail elsewhere [43**], this approach was used to study the structure of lifetime comorbidity among the 18 DSM-IV disorders assessed in the WHO World Mental Health (WMH) surveys, a series of community epidemiological surveys administered to 21,229 respondents across 14 different countries [46]. Preliminary analysis found that point-intime comorbidity among these disorders fit a two-factor internalizing-externalizing disorders factor model. Retrospective AOO reports were then used to estimate a series of 18 survival equations in which first onset of each core disorder was predicted by prior lifetime onset of the other 17 disorders along with basic socio-demographic controls. 98.0% of the 306 (18x17) survival coefficients were positive and 95.1% significant in bivariate analysis. 80.0% of survival coefficients were positive and 43.0% positive and statistically significant in multivariate analyses. Within-domain time-lagged associations were generally stronger than between-domain associations.

The latent variable model was then estimated and was found to fit the data much better than the observed variable model. The most important predictors of the latent variables were specific phobia and obsessive-compulsive disorder for the internalizing dimension and hyperactivity disorder and oppositional-defiant disorder for the externalizing dimension. Controls for the latent variables explained the vast majority of the originally significant time-lagged associations among observed disorders. Most of the 13 residual pair-wise time-lagged associations between observed disorders that remained significant involved either (i) within-disorder reciprocal associations (e.g., attention-deficit with hyperactivity disorders; overt and covert subtypes of conduct disorder), (ii) asymmetrical associations between well-known disorder pairs (panic predicting agoraphobia and major depression predicting GAD), and (iii) likely diagnostic confusions (agoraphobia predicting specific phobia and hyperactivity disorder, but not attention-deficit disorder, predicting bipolar disorder).


The good fit of the latent variable model suggests that common causal pathways account for most comorbidity among the disorders considered. A similar pattern was found in preliminary analysis of several other datasets [47]. Ongoing analyses of these data are exploring more refined specifications that examine variation in the relative importance of different temporally primary disorders in predicting subsequent onset of secondary disorders at different points in the life course. These analyses are also investigating the possibility of synergistic effects of comorbid primary disorders. The latent variable modeling approach used here is very flexible in allowing these types of elaborations to be considered.

Based on the results of the WMH analyses as well as the confirmatory results of the additional analyses described in the previous paragraph, it appears that the common pathways defined by latent internalizing and externalizing variables (and possible expansion of these latent variables to include more refined distinctions among disorders) should be the focus of future research on the development of comorbidity. The analyses carried out so far show that more differentiation across disorders can be found in the predictive associations of temporally primary disorders with subsequent onset of secondary comorbid disorders than in the point-in-time associations between prevalent disorders. This makes it possible for us to pinpoint critical seed disorders that are associated with high risk of subsequent onset of lifetime comorbidity. Parallel analyses are needed to determine which primary disorders (or constellations of comorbid disorders) predict disorder persistence so as to increase our understanding of the dynamic influences on episode comorbidity.

It is also important, though, to recognize the existence of several important residual associations that cannot be explained by the mediating role of latent predispositions. We noted in the introduction that latent variable models can be useful in helping to determine when associations thought to be specific (e.g., a positive association between female gender and depression) are really part of a more general pattern (e.g., a positive association between female gender and internalizing disorders, with no special elevation of the association with depression compared to other internalizing disorders). The flip side of that issue is that latent variable models provide a unique way to search through a large number of associations to distinguish the few that are specific from the larger number that are general. Consistent evidence across studies of the existence of particular specific associations can be valuable in calling attention to the importance of these associations as a preliminary to carrying out more focused studies of these associations.


  • Although numerous factor analysis studies of point-in-time comorbidity among mental disorders document the existence of latent predispositions to internalizing and externalizing disorders, it has only been recently that research has begun to study the effects of latent predispositions in predicting the development of comorbidity.
  • A novel survival analysis with mediating latent variables applied to 14 epidemiological surveys collected in the WHO World Mental Health Survey Initiative showed that the consistently significant positive associations found in observed-variable models between most temporally primary lifetime disorders and subsequent first onset of most secondary disorders could largely be explained by the mediating effects of latent internalizing and externalizing variables, although, a number of intriguing residual time-lagged associations remained even after controlling these latent predispositions.
  • The good fit of the latent variable model suggests that the common causal pathways that account for most comorbidity among internalizing and externalizing disorders should be the focus of future research on the development of comorbidity, although the analysis approach described here provides a unique way to pinpoint significant residual associations that should also be the subject of further focused study.


We thank Adrian Angold, Jane Costello, Brian Cox, Danny Pine for helpful comments on an earlier draft of the paper. This report was prepared as part of the work of the World Health Organization World Mental Health (WMH) Survey Initiative. The work of Zaslavsky in preparing the report was additionally supported by NIMH grant R01-MH66627. We thank the staff of the WMH Data Analysis Coordination Centre for consultation on data analysis. Their work is supported by NIMH (R01-MH070884, R13-MH066849, R01-MH069864, R01-MH077883), NIDA (R01-DA016558), the Fogarty International Center of the National Institutes of Health (FIRCA R03-TW006481), the John D. and Catherine T. MacArthur Foundation, the Pfizer Foundation, and the Pan American Health Organization. The WMH Data Analysis Coordination Centre has also received unrestricted educational grants from Astra Zeneca, BristolMyersSquibb, Eli Lilly and Company, GlaxoSmithKline, Ortho-McNeil, Pfizer, Sanofi-Aventis, Shire, and Wyeth. A complete list of WMH publications can be found at The views and opinions expressed in this report are those of the authors and should not be construed to represent the views of any of the sponsoring organizations, agencies, or governments.

Financial Disclosure: Kessler has been a consultant for GlaxoSmithKline Inc., Kaiser Permanente, Pfizer Inc., Sanofi-Aventis, Shire Pharmaceuticals, and Wyeth-Ayerst; has served on advisory boards for Eli Lilly & Company, Johnson & Johnson Pharmaceuticals, and Wyeth-Ayerst; and has had research support for his epidemiological studies from Bristol-Myers Squibb, Eli Lilly & Company, GlaxoSmithKline, Johnson & Johnson, Ortho-McNeil, Pfizer, and Sanofi-Aventis. Dr. Russo is an employee of Shire Pharmaceuticals.


This paper was prepared with support from grants funded by the National Institute of Health.

Portions of this paper appeared previously in Kessler, R. C., Cox, B. J., Green, J. G., et al. (2011), The effects of latent variables in the development of comorbidity among common mental disorders. Depression and Anxiety, 28: 29–39, and Kessler, R.C., Ormel, J., Petukhova, M., et al. (2011). Development of lifetime comorbidity in the World Health Organization World Mental Health Surveys. Archives of General Psychiatry 68(1), 90-100.

The remaining authors report nothing to disclose.

1Only three observed lifetime time t internalizing disorders (e.g., i1t represents internalizing disorder 1 at time t) and externalizing disorders along with only one observed internalizing and one observed externalizing disorder at time t+1 are shown to simplify the presentation. First onset of each of these disorders between times t and t+1 was predicted by prior lifetime history of the other disorders as of time t. Estimation was made in D separate survival equations, each with D-1 predictors for prior history of the other disorders, for a total of D × (D-1) pair-wise time-lagged associations between earlier and later disorders. The D-1 predictor disorders are treated as time-varying covariates in a discrete-time (person-year) survival framework.

1Only three observed lifetime time t internalizing disorders (e.g., i1t represents internalizing disorder 1 at time t) and externalizing disorders and only three disorders of each set at time t+1 are shown to simplify the presentation. These latent variables, in turn, were predicted by lifetime history of latent internalizing and externalizing variables as of time t. These time t latent variables, finally, were predicted by lifetime history of observed internalizing or externalizing variables as of time t. Estimation was carried out using a three-part iterative procedure. See the text for more details. As in the earlier observed variable model, the predictor disorders were treated as time-varying covariates in a discrete-time (person-year) survival framework.


1. Demyttenaere K, Bruffaerts R, Posada-Villa J, et al. Prevalence, severity, and unmet need for treatment of mental disorders in the World Health Organization World Mental Health surveys. JAMA. 2004;291:2581–2590. [PubMed]
2. Kessler RC, Chiu WT, Demler O, et al. Prevalence, severity, and comorbidity of 12-month DSM-IV disorders in the National Comorbidity Survey Replication. Arch Gen Psychiatry. 2005;62:617–627. [PMC free article] [PubMed]
3. Krueger RF. The structure of common mental disorders. Arch Gen Psychiatry. 1999;56:921–926. [PubMed]
4. Beesdo K, Hofler M, Gloster A, et al. The structure of common mental disorders: a replication study in a community sample of adolescents and young adults. Int J Methods Psychiatr Res. 2009;18:204–220. [PubMed]
5. Cox BJ, Swinson RP. Instrument to assess depersonalization-derealization in panic disorder. Depress Anxiety. 2002;15:172–175. [PubMed]
6. Krueger RF, Markon KE. Reinterpreting comorbidity: a model-based approach to understanding and classifying psychopathology. Annu Rev Clin Psychol. 2006;2:111–133. [PMC free article] [PubMed]
7. Lahey BB, Rathouz PJ, Van Hulle C, et al. Testing structural models of DSM-IV symptoms of common forms of child and adolescent psychopathology. J Abnorm Child Psychol. 2008;36:187–206. [PubMed]
8. Slade T, Watson D. The structure of common DSM-IV and ICD-10 mental disorders in the Australian general population. Psychol Med. 2006;36:1593–1600. [PubMed]
9. Vollebergh WA, Iedema J, Bijl RV, et al. The structure and stability of common mental disorders: the NEMESIS study. Arch Gen Psychiatry. 2001;58:597–603. [PubMed]
10. Andrews G, Goldberg DP, Krueger RF, et al. Exploring the feasibility of a meta-structure for DSM-IV and ICD-11: could it improve utility and validity? Psychol Med. 2009;39:1993–2000. [PubMed]
11. Goldberg DP, Krueger RF, Andrews G, Hobbs MJ. Emotional disorders: cluster 4 of the proposed meta-structure for DSM-IV and ICD-11. Psychol Med. 2009;39:2043–2059. [PubMed]
12. Krueger RF, Markon KE. Understanding Psychopathology: Melding Behavior Genetics, Personality, and Quantitative Psychology to Develop an Empirically Based Model. Curr Dir Psychol Sci. 2006;15:113–117. [PMC free article] [PubMed]
13. Watson D. Rethinking the mood and anxiety disorders: a quantitative hierarchical model for DSM-V. J Abnorm Psychol. 2005;114:522–536. [PubMed]
14. Wittchen HU, Beesdo K, Gloster AT. A new meta-structure of mental disorders: a helpful step into the future or a harmful step back to the past? Psychol Med. 2009;39:2083–2089. [PubMed]
15. Wittchen HU, Beesdo-Baum K, Gloster A, et al. The structure of mental disorders reexamined: is it developmentally stable and robust against additions? Int J Methods Psychiatr Res. 2009;18:189–203. [PubMed]
16. Krueger RF, Caspi A, Moffitt TE, Silva PA. The structure and stability of common mental disorders (DSM-III-R): a longitudinal-epidemiological study. J Abnorm Psychol. 1998;107:216–227. [PubMed]
17. Krueger RF, Finger MS. Using item response theory to understand comorbidity among anxiety and unipolar mood disorders. Psychol Assess. 2001;13:140–151. [PubMed]
18. Green JG, McLaughlin KA, Berglund PA, et al. Childhood adversities and adult psychiatric disorders in the national comorbidity survey replication I: associations with first onset of DSMIV disorders. Arch Gen Psychiatry. 2010;67:113–123. [PMC free article] [PubMed]
19. Mrazek PJ, Haggerty RJ. Reducing Risks for Mental Disorders: Frontiers for Preventive Intervention Research. National Academy Press; Washington, DC: 1994.
20. Kramer MD, Krueger RF, Hicks BM. The role of internalizing and externalizing liability factors in accounting for gender differences in the prevalence of common psychopathological syndromes. Psychol Med. 2008;38:51–61. [PubMed]
21. Kessler RC, McLaughlin KA, Green JG, et al. Childhood adversities and adult psychopathology in the WHO World Mental Health Surveys. Br J Psychiatry. 2010;197:378–385. [PMC free article] [PubMed]
22. Baker JH, Mitchell KS, Neale MC, Kendler KS. Eating disorder symptomatology and substance use disorders: prevalence and shared risk in a population based twin sample. Int J Eat Disord. 2010;43:648–658. [PMC free article] [PubMed]
23. Lyons M, Hitsman B, Xian H, et al. A twin study of smoking, nicotine dependence, and major depression in men. Nicotine Tob Res. 2008;10:97–108. [PubMed]
24. Tambs K, Czajkowsky N, Roysamb E, et al. Structure of genetic and environmental risk factors for dimensional representations of DSM-IV anxiety disorders. Br J Psychiatry. 2009;195:301–307. [PMC free article] [PubMed]
25. Kendler KS, Aggen SH, Czajkowski N, et al. The structure of genetic and environmental risk factors for DSM-IV personality disorders: a multivariate twin study. Arch Gen Psychiatry. 2008;65:1438–1446. [PMC free article] [PubMed]
26. Bornovalova MA, Hicks BM, Iacono WG, McGue M. Familial transmission and heritability of childhood disruptive disorders. Am J Psychiatry. 2010;167:1066–1074. [PMC free article] [PubMed]
**27. Kendler KS, Aggen SH, Knudsen GP, et al. The structure of genetic and environmental risk factors for syndromal and subsyndromal common DSM-IV axis I and all axis II disorders. Am J Psychiatry. 2011;168:29–39. [PMC free article] [PubMed]
[This study is the most comprehensive factor analysis study of the additive genetic and envionmental components of covariance among common internalizing and externalizing disorders. The analysis showed that the finding of strong underlying comorbidity between internalizing and externalizing disorders was much more pronounced for the genetic than the environmental components of covariance.]
28. Kendler KS, Prescott CA, Myers J, Neale MC. The structure of genetic and environmental risk factors for common psychiatric and substance use disorders in men and women. Arch Gen Psychiatry. 2003;60:929–937. [PubMed]
29. Lewontin RC. Annotation: the analysis of variance and the analysis of causes. Am J Hum Genet. 1974;26:400–411. [PubMed]
**30. Molenaar PCM. On the limits of standard quantitative genetic modeling of inter-individual variation: Extensions, ergodic conditions and a new genetic factor model of intra-individual variation. In: Hood KE, Halpern CT, Greenberg G, Lerner RM, editors. Handbook of Developmental Science, Behavior, and Genetics. Blackwell; Malden, MA: 2010. pp. 626–648.
[This methodological paper summarizes the major methodological critiques of genetics epidemiological studies that use twin data to decompose variance and covariance of observed variables into additive genetic and environmental components.]
31. Fergusson DM, Horwood LJ, Ridder EM. Conduct and attentional problems in childhood and adolescence and later substance use, abuse and dependence: results of a 25-year longitudinal study. Drug Alcohol Depend. 2007;88(Suppl 1):S14–26. [PubMed]
32. Merikangas KR, Zhang H, Avenevoli S, et al. Longitudinal trajectories of depression and anxiety in a prospective community study: the Zurich Cohort Study. Arch Gen Psychiatry. 2003;60:993–1000. [PubMed]
33. Orvaschel H, Lewinsohn PM, Seeley JR. Continuity of psychopathology in a community sample of adolescents. J Am Acad Child Adolesc Psychiatry. 1995;34:1525–1535. [PubMed]
34. Stein MB, Fuetsch M, Muller N, et al. Social anxiety disorder and the risk of depression: a prospective community study of adolescents and young adults. Arch Gen Psychiatry. 2001;58:251–256. [PubMed]
35. Burke JD, Loeber R, Lahey BB, Rathouz PJ. Developmental transitions among affective and behavioral disorders in adolescent boys. J Child Psychol Psychiatry. 2005;46:1200–1210. [PubMed]
36. Copeland WE, Shanahan L, Costello EJ, Angold A. Childhood and adolescent psychiatric disorders as predictors of young adult disorders. Arch Gen Psychiatry. 2009;66:764–772. [PMC free article] [PubMed]
37. Costello EJ, Mustillo S, Erkanli A, et al. Prevalence and development of psychiatric disorders in childhood and adolescence. Arch Gen Psychiatry. 2003;60:837–844. [PubMed]
38. Feehan M, McGee R, Williams SM. Mental health disorders from age 15 to age 18 years. J Am Acad Child Adolesc Psychiatry. 1993;32:1118–1126. [PubMed]
39. Newman DL, Moffitt TE, Caspi A, et al. Psychiatric disorder in a birth cohort of young adults: prevalence, comorbidity, clinical significance, and new case incidence from ages 11 to 21. J Consult Clin Psychol. 1996;64:552–562. [PubMed]
40. Beesdo K, Bittner A, Pine DS, et al. Incidence of social anxiety disorder and the consistent risk for secondary depression in the first three decades of life. Arch Gen Psychiatry. 2007;64:903–912. [PubMed]
41. Angold A, Costello EJ, Erkanli A. Comorbidity. J Child Psychol Psychiatry. 1999;40:57–87. [PubMed]
42. Jensen PS. Comorbidity and child psychopathology: recommendations for the next decade. J Abnorm Child Psychol. 2003;31:293–300. [PubMed]
**43. Kessler RC, Ormel J, Petukhova M, et al. Development of lifetime comorbidity in the World Health Organization world mental health surveys. Arch Gen Psychiatry. 2011;68:90–100. [PMC free article] [PubMed]
[This paper presented detailed results of the application of the latent variable model exposited in the current paper to data on the prediction of lifetime comorbidity in the WHO World Mental Health (WMH) surveys.]
44. Hosmer DW, Lemeshow S. Applied Survival Analysis: Regression Modeling of Time to Event Data. Wiley; New York, NY: 1999.
45. Zelen M. Forward and backward recurrence times and length biased sampling: age specific models. Lifetime Data Anal. 2004;10:325–334. [PubMed]
46. Kessler RC, Aguilar-Gaxiola S, Alonso J, et al. The global burden of mental disorders: an update from the WHO World Mental Health (WMH) surveys. Epidemiol Psichiatr Soc. 2009;18:23–33. [PMC free article] [PubMed]
47. Kessler RC, Cox BJ, Green JG, et al. The effects of latent variables in the development of comorbidity among common mental disorders. Depress Anxiety in press [PMC free article] [PubMed]