Robinaugh2020 - The network approach to psychopathology, a review of the literature 2008-2018 and an agenda for future research

  • Type:#article
  • Year read:#read2021
  • Subject: Network theory
  • Bibtex: @robinaugh2020
  • Bibliography: Robinaugh, D. J., Hoekstra, R. H. A., Toner, E. R., & Borsboom, D. (2020). The network approach to psychopathology: A review of the literature 2008–2018 and an agenda for future research. Psychological Medicine, 50(3), 353–366.

Example citation

Key takeaways

  • They review 363(!) articles in the field 2008-2018.
  • “From a network perspective, mental disorder is characterized not only by the state of the network (i.e. elevated symptom activation), but also by the structure of the network: in particular, a strongly connected network in which inter-symptom relationships are sufficient to maintain elevated symptom activation over time”
  • three

Five areas of work in network theory

Philosophy of psychiatry

The network theory of psychopathology should be contrasted to essentialist and “underlying cause” explanations, like the Germ theory of disease

The theoretical works deal with psychiatric comorbidity, the p-factor theory of psychopathology, Equifinality and Equipotentiality, among others.

Network science

Early theoretical work drew from reserach on network science.

When causal relations among symptoms are strong, the onset of one symptom will lead to the onset of others : causality hypothesis

Symptom networks can have contagion effects where activation spreads through the network : connectivity hypothesis

Central symptoms have greater potential to spread activation throughout the network : centrality hypothesis

Symptoms occuring in multiple disorders can bridge activation between the networks : comorbidity hypothesis

Affect dynamics and momentary experience

Some authors have argued that we should shift our focus to the experiences that occur in the moment, i.e. seconds or minutes, rather than days/weeks which we are typically doing. Panic attacks vs panic disorder.

See Bak et al., 2016 of a n = 1 example in psychosis

Cognitive behavioral theory

The CBT tradition is to create models with plausible causal relations. Vicious cycle theories of OCD and PD.

Salkovskis1985 - Obsessional Compulsive Problems A Cognitive-Behavioural Analysis

Systems science

Beyond network science: dynamical systems theory, catastrophe theory, cybernetics. Feedback loops from CBT theories have been investigated in cybernetics and dynamical systems theory.

Some of this work predated the network theory explosion in psychopathology, and is interesting work.

Next steps

Formal theories that specify how a specific disorder operates as a causal system are scarce. The authors think this is a critical next step for network theory, perhaps using computational models. Some challenges:

  • Symptoms are dimensional and not binary
  • Effects vary in time-scale (seconds/minutes vs days/months)
  • Higher order interactions plausible (sleep moderating the effect of trauma memories on emotional reactivity)

Their recommendations:

  • Use CBT models
  • Affective dynamics research can inform the time-scale
  • Dynamical systems literature provide tools to model and evaluate complex systems

Network methodology

  • Cross-sectional: a single network estimated between-persons at one point in time
  • n = 1 time-series: covariation of symptoms over time within one individual
  • Larger sample time-series: Both within- and between-person information

What is the “true” structure of the causal system? It remains unclear.

Network structure

The most popular method: pairwise Markov random field

Network characteristics

  • Node centrality
  • Node predictability
  • Node clustering
  • Community structure
  • Similarity of intra-individual network structure

Next steps

  • Most mental disorders do not meet the assumptions for PMRF networks.
  • How to aggregate findings?
  • How to do adequate data collection?

Empirical studies

Depression (n = 69) and PTSD (n = 31) studies odminate. The two most common network characteristics studied are network connectivity and node centrality.

Network connectivity

Mental disorders have symptoms that are highly interconnected, even when controlling for shared variance among symptoms. So there is meaningful clustering of symptoms.

Connectivity hypothesis: (cross-sectional studies)

  • Higher connectivity in more severe disease
  • Higher baseline connectivity in individuals who did not remit after treatment
  • Genetic risk does not seem to be associated with higher connectivity
  • Treatment does not seem to affect connectivity, and some have found it leads to more connectivity

Mixed results in time-series studies.

Together, these studies utilizing time-series data provide qualified support for the notion that connectivity of negative mood state networks is associated with psychopathology, but minimal evidence that broader networks of momentary experiences exhibit such associations.

Node centrality

Centrality hypothesis

  • No evidence that symptoms in the DSM play a “privileged” role in networks. Some have found no difference between DSM and non-DSM symptoms of depression.
  • Central symptoms are more predictive of subsequent diagnosis in depression and PTSD.

Consistent with centrality hypothesis, but also a common cause framework.

Future directions

Data collected for other purposes

The data collection procedures in different studies varies greatly. Studies need to be design with the explicit goal of investigating mental disorders as complex systems.

No experimental manipulation of symptoms

How do symptoms affect and interact with each other?

Be more cautious when deriving hypotheses

Complex systems are difficult to predict so we should be careful here.

Agenda for future research

(a) Identifying robust empirical phenomena

Most studies to date are exploratory, we need to understand which findings are replicable and generalizable. Can be achieved by methodological development, meta-analytic techniques, and developing techniques for confirmatory network analyses.

(b) Developing formal theories that can explain (a)

Basically developing computational models, and letting empirical and theoretical work inform each other.