Dimensions or categories?

2011-08-24

Some further reading notes on the dimensional vs. categorical approaches to mental disorders.

I just uploaded a BibTeX file as gist 1106828 on github. An htmlized version is available here: dsm5_minimal.html. It is all about the revised version of the Diagnostic and Statistical Manual of Mental Disorders (DSM). Most of these papers come from the references list available on dsm5.org.

As a follow-up to a previous post on Psychometrics, measurement, and diagnostic medicine, here is a good article describing why a dimensional approach to the assessment and diagnosis of personality disorders is necessary in place of the well-established but controversial categorical approach:

Assessment and diagnosis of personality disorder: Perennial issues and an emerging reconceptualization, by Lee A. Clark, Annual Review of Psychology, 2007, 58: 227-257.
This chapter reviews recent (2000-2005) personality disorder (PD) research, focusing on three major domains: assessment, comorbidity, and stability. (a) Substantial evidence has accrued favoring dimensional over categorical conceptualization of PD, and the five-factor model of personality is prominent as an integrating framework. Future directions include assessing dysfunction separately from traits and learning to utilize collateral information. (b) To address the pervasiveness and extent of comorbidity, researchers have begun to move beyond studying overlapping pairs or small sets of disorders and are developing broader, more integrated common-factor models that cross the Axis I-Axis II boundary. (c) Studies of PD stability have converged on the finding that PD features include both more acute, dysfunctional behaviors that resolve in relatively short periods, and maladaptive temperamental traits that are relatively more stable-similar to normal-range personality traits-with increasing stability until after 50 years of age. A new model for assessing PD–and perhaps all psychopathology–emerges from integrating these interrelated reconceptualizations.

The next DSM should see the incorporation of the five factor model of personality, as well as the addition of a dimensional adjunct to each of the traditional categorical diagnoses of the DSM [1]. There are many other issues that the DSM working groups wish to address, including the revision of the hierarchical structure of the "pure" diagnostic categories because of the high rate of co-occurrence (e.g., there exists a high comorbidity between generalized anxiety disorder and major depressive disorder), the frequent use of the "not otherwise specified" (NOS) label, or the heterogeneous mix of conditions within current diagnostic boundaries [2]. Also, nowadays biological markers are increasingly used as predictive features or risk factors. There clinical utility (as diagnostic criteria), as well as that of age-, gender-, and culture-related specifiers should be field-tested shortly.

As a sidenote, the revision of the DSM won't solve the problem of causality, as discussed by P. Fayers in his papers on causal vs. indicator variables [3,4], where the direction of causality is assumed to run from the construct to the indicators, whence changes in the construct are expected to be manifested in changes in all indicators a multi-item scale is composed of. Another good paper is [5]. This reminds us of the causal networks consisting of symptoms and direct causal relations between discussed by Borsboom [6,7]. However, it is hoped that incorporating dimensional measures will facilitate the definition of more useful diagnostic categories by highlighting some dimensional discontinuities between disorders and clear thresholds between pathology and normality [2]. Also, the use of simple dimensional scales would obviously help to overcome the "shopping list" actually offered by the DSM-IV, which yield to an increased

The idea of making a distinction between categorical attributes and unidimensional scales was also debated by Paul De Boeck during the IV EAM conference, but see [8]. A must-read for every psychometricians, whether they are interested in diagnostic medicine or not. According to De Boeck and coworkers, the dimensional approach requires "within-category heterogeneity and between-category quantitative differences", while the categorical approach suggests "within-category homogeneity and between-category qualitative differences". Also, an important point justifying their work on a Dimension/Category Framework ("DimCat") is highlighted in the introduction (emphasize is mine):

The issue under consideration here is whether the latent nature of manifest variables is category-like or dimension-like. One assumption might be that the nature of the latent and manifest variables match. As discussed below, however, manifest dimensions can be turned into manifest categories (e.g., in segmentation into groups) and manifest categories into manifest dimensions (e.g., in sum scores on a test). Thus, the relations between different kinds of manifest variables and between different kinds of manifest and latent variables are not so simple as they might at first appear. Consequently, a conceptual and methodological framework that encompasses all of these possibilities is needed.

Well, all that to say that diagnostic medicine will lead to many opportunities to develop innovative psychometrical models; we should see an increasing interest in the identification of networks of comorbidities, measurement of health outcomes, and characterisation and delineation of mental disorders.

References

  1. Kraemer, H.C. (2007). DSM categories and dimensions in clinical and research contexts. International Journal of Methods in Psychiatric Research, 16(Supp. 1): S8-S15.
  2. American Psychopathological Association, et al. (2010). The conceptual evolution of DSM-5. American Psychiatric Publishing.
  3. Fayers, P.M. and Hand, D.J. (2002). Causal Variables, Indicator Variables and Measurement Scales: An Example from Quality of Life. Journal of the Royal Statistical Society A, 165(2), 233-261.
  4. Fayers, P.M., Hand, D.J., Bjordal, K., and Groenvold, M. (1997). Causal indicators in quality of life research. Quality of Life Research, 6(5), 393-406.
  5. Edwards, J.R. and Bagozzi, R.P. (2000). On the nature and direction of relationships between constructs and their measures. Psychological Methods, 5, 155–74.
  6. Borsboom, D. (2008). Psychometric perspectives on Diagnostic systems. Journal of Clinical Psychology, 64(9), 1089-1108.
  7. Cramer, A.O.J., Waldorp, L.J., van der Maas, H.L.J., Borsboom, D. (2010). Comorbidity: A network perspective. Behavioral and Brain Sciences, 33, 137-193.
  8. De Boeck, P., Wilson, M., and Acton, G.S. (2005). A Conceptual and Psychometric Framework for Distinguishing Categories and Dimensions. Psychological Review, 112(1): 129-158. HTML version
---

Articles with the same tag(s):

Multi-Group comparison in Partial Least Squares Path Models
Data Science from Scratch
Stata for health researchers
R Graphs Cookbook
Bad Data
Data science at the command-line
Reproducible research with R
Twenty canonical questions in machine learning
Do a large amount of consulting
Stata for structural equation modeling

---