The complexity of psychopathology:! A network approach Denny%Borsboom% University%of%Amsterdam www.psychosystems.org
Overview The network idea! Conceptual consequences:! Disorders and their boundaries! Kinds or categories! The particular and the general! Discussion
The standard model error error error mood sleep interest Depression
Reductionist daydreams
Great expectations, limited yields Research%into%the% biological%basis %for%disorders% has%proven%exceptionally%tricky% Limited%advances%in%pinpointing%neurobiological% correlates,%let%alone%causes% Genetic%findings%suggest%no%large%and%specific% genetic%effects%for%common%mental%disorders% Also%note%that%although%medication%prescription% has%risen%spectacularly,%prevalence%is%relatively% constant
The network alternative sleep tired conc worry
I: Disorders and their boundaries
What magnetism and depression have in common 8 X 1 = 1 X 2 = 1 X 3 = 1 X 4 = 1 X 5 = 1 X 6 = 1 X 7 = 1 X 8 = 1 X 9 = 1 X 10 = 1 X 11 = 1 X 12 = 1 X 13 = 1 X 14 = 1 X 15 = 1 X 16 = 1 (b) repressented by a network shaped as a lattice as in s alligned upwards and 1 indicates that the south (b) adheres to a PMRF in that the probability of a nly dependent on the state of its direct neighbors. hyp rmo sui qmo ple sex fut fal sad int sen slo sel irr con sle ene anx app wak par wei res bod pan mod ach dig fal: falling asleep sle: sleep during the night wak: waking up too early hyp: hypersomnia sad: feeling sad irr: feeling irritable anx: feeling anxious rmo: respons of mood to events mod: mood in relation to time of day qmo: quality of mood app: change of appetite wei: change of weight con: concentration problems sel: view of oneself fut: view of future sui: suicidal thoughts int: general interest ene: energy level ple: capacity for pleasure (not sex) sex: interest in sex slo: feeling slowed down res: feeling restless ach: aches/pains bod: other bodily symptoms pan: panic/phobic symptoms dig: digestion problems sen: interpersonal sensitivity par: leaden paralysis Figure 1: Network constructed with `1 regularized logistic regression of each variable on the other variables. Neighborhood selection is optimized with EBIC. only depends on the direct neighbors (north, south
Ising model for the entire DSM-IV 1 12 6 5 10 2 7 4 10 8 7 8 3 6 5 1 4 13 11 2 9 7 1 6 4 3 2 8 19 5 9 18 12 9 3 2 6 5 1 4 10 7 8 3 7 15 5 17 16 11 14 13 15 16 14 4 3 14 6 2 16 9 1 15 13 5 8 12 10 7 11 3 3 4 3 6 1 2 1 4 2 4 5 2 6 4 8 3 11 17 4 5 1 7 2 6 3 7 9 2 6 10 1 4 8 5 3 4 5 2 1 2 1 3 4 1 5 2 3 7 Major depressive episode Dysthymia Mania or hypomania Generalised anxiety disorder Social phobia Specific phobia Panic disorder Agoraphobia Post-traumatic stress disorder Attention-deficit/hyperactivity disorder Alcohol abuse or dependence Nicotine dependence Psychotic symptoms 17 18 13 12 10 1 9
So... From a network perspective, boundaries between disorders are inherently fuzzy! Excessive comorbidity cannot be avoided, as it is inherent in the structure of the network! This may imply that diagnostic practice should be revised (no more diagnostic silos )
II: Mental disorders: Kinds or continua?
Diathesis-stress Two ways we can manipulate a network:! by putting the network under stress: activating and deactivating symptoms! by changing network vulnerability (diathesis): increasing and decreasing connectivity strength
LOW Connectivity
MEDIUM Connectivity
HIGH Connectivity
So... From a network perspective, the kinds or continua question is misguided! Networks can behave as kinds or as continua, depending on their parameters! Many possible extensions: early warnings, dynamical systems models, other constructs
III: Personalized networks
C P E! "#$% &%! + " ' ( ')$*'% + F D " + ( +)$*' +... + #! T H Bringmann, L., Vissers, N., Wichers, M., Geschwind, N., Kuppens, P., Peeters, F., Borsboom, D., & Tuerlinckx, F. (2013). A network approach to psychopathology: New insights into clinical longitudinal data. PLoS ONE.!
When qualitatively analyzing the network structure, it may be concluded that the data of this Patient 1 (diagnosed negative cluster assembles feeling tired, depressed and stressed, physical inconvenience, and with panic disorder) patient seem to group in two clusters, which can be labeled as a positive and a negative cluster. The positive cluster assembles the variables feeling excited and relaxed, being physically active, having pleasant daily experiences and a pleasant social environment. The unpleasant experiences. Being excited seems to have a bridge function between these two clusters. Feeling excited often co-occurs with feeling less tired, stressed, and depressed, whereas physical activity and pleasant daily experiences are associated with feeling excited. Visualizing the lag-1 correlation matrix of the same patient showed the following weighted graph in Figure 4. Figure 4. Lag-1 correlation network of Patient 1.
Patient 2 (diagnosed Version: October 2013 with between mood and context, and depression) thus less autonomous mood processes, than in the first measurement period. Physical activity is still not significantly related to e.g. mood. The lag-1 correlation network of the second measurement period is shown in Figure 10. Figure 10. Lag-1 correlation network of Patient 3 during the second measurement period.
Early warnings?
So... From a network perspective, the general structure of mental disorders may be an amalgam of individual patterns! Because structure and dynamics are coupled in networks, such individual differences can have large consequences! This may help in identifying targets for intervention
Discussion Network approaches change the conceptualization of disorders, and hence nature of the research game! Networks motivate the study of feedback and interconnection, rather than a search for latent diseases! The holistic nature of network modeling de-emphasizes searches for the genes or the neural correlates of disorders
People Psychosystems group: Angélique Cramer, Verena Schmittmann, Lourens Waldorp, Sacha Epskamp, Claudia van Borkulo, Mijke Rhemtulla; Collaborators: Laura Bringmann, Arjen Noordhof, Francis Tuerlinckx, Sophie van der Sluis, Kenneth Kendler, Steve Aggen, Hilde Geurts, Marieke Wichers, Erik Giltay, Han van der Maas, Marten Scheffer, Ingrid van de Leemput, Gunter Maris, Robert Schoevers. Students: Noémi Schuurman, Robert Hillen, Michel Nivard, Susanna Gerritse, Janneke de Kort, Charles Driver, Laura Ruzzano, Esther Lietaert-Peerbolte, Jonas Dalege, Jolanda Kossakowski, Tessa Blanken Papers Borsboom, D., & Cramer, A. O. J. (2013). Network analysis. Annual Review of Clinical Psychology.! Bringmann, L., Vissers, N., Wichers, M., Geschwind, N., Kuppens, P., Peeters, F., Borsboom, D., & Tuerlinckx, F. (2013). A network approach to psychopathology: New insights into clinical longitudinal data. PLoS ONE.! Cramer, A. O. J., Borsboom, D., Aggen, S. H., & Kendler, K. S. (2011). The Pathoplasticity of Major Depression. Psychological Medicine.! Borsboom, D., Cramer, A. O. J., Schmittmann, V. D., Epskamp, S., & Waldorp, L. J. (2011). The small world of psychopathology. PLoS ONE.! Schmittmann, V. D., Cramer, A. O. J., Waldorp, L. J., Epskamp, S., Kievit, R. A., & Borsboom, D. (2011). Deconstructing the construct: A network perspective on psychological phenomena. New Ideas in Psychology.! Cramer, A.O.J., Waldorp, L.J., Van der Maas, H.L.J., & Borsboom, D. (2010). Comorbidity: A network approach. Behavioral and Brain Sciences, 33, 137-193.! Borsboom, D. (2008). Psychometric perspectives on diagnostic systems. Journal of Clinical Psychology, 64, 1089-1108. Software Epskamp, S., Cramer, A. O. J., Waldorp, L. J., Schmittmann, V. D., & Borsboom, D. (2011). qgraph: Network representations of relationhips in data. R package version0.4.10. Available from http://cran.r-project.org/ package=qgraph. dennyborsboom@gmail.com www.psychosystems.org