DyNOPS

DyNOPS (Dynamic Bayesian Network Modelling of Psychopathology during Outpatient Psychotherapy) is a collaborative research project between the Institute of Psychology and the Institute of Data Science in Biomedicine at TU Braunschweig. The project aims to view psychological disorders as causal interactions between symptoms forming a continuous feedback loop, rather than as observable parts of a latent condition. DyNOPS uses dynamic Bayesian networks to understand how symptoms of depression and anxiety disorders are interconnected during outpatient psychotherapy.

Theoretical Background

For just over 10 years, it has been assumed that mental disorders are not a latent, invisible illness that produces symptoms such as a depressed mood (in the same way that a bacterial infection causes symptoms such as fever and cough), but rather that it is the reciprocal, causal interactions between the individual symptoms that form the core of the disorder (Borsboom & Cramer, 2013). The whole thing can be imagined as a network.

In ongoing psychotherapy, this could mean that an improvement in one symptom triggers – often with a delay – improvements in other symptoms. If certain symptoms act as ‘pioneers’ of change, they should be the first to be addressed in therapy. But what are pioneer symptoms? Furthermore, little is known about how personal characteristics (age, gender, type of diagnosis, comorbidities, etc.) influence the network. The DyNOPS project addresses precisely these gaps and aims to provide a sound basis for evidence-based therapy that focuses on the right interventions for each individual patient.

The focus of this study is to investigate whether these temporal phenomena occur, which symptoms might spearhead change in the network, and whether patterns of symptom interactions are associated with better therapy outcomes. In addition, this study aims to explore how patient characteristics may influence the structure of these symptom interactions and shape differences in therapy outcomes. The network will be constructed using data-driven structure learning, with optional input from clinical experts. In particular, clinicians can specify which symptom interactions are clinically possible or impossible. These can be expressed as a whitelist, which specifies the connections that must be present in the network, or a blacklist, which specifies the connections that should not be included because they are clinically impossible or meaningless.

Research Questions

  • Are there any temporal interactions between symptoms during psychotherapy?
  • Do certain temporal interactions predict improvements in overall therapy outcomes?
  • How do patient characteristics influence symptom interactions and therapy outcome?

Goals

  • Model temporal interactions between symptoms in outpatient psychotherapy with dynamic Bayesian networks. 
  • Identify central or controlling nodes in the network that precede further changes in the system, and describe and model the change of the network following the course of psychotherapy.
  • Investigate patient-level factors that are associated with differences in network structure, such as sociodemographic factors (gender, age, educational, occupational status, and relationship status), type or intensity of disorders (mood vs. anxiety disorders, mild vs. major depressive episode, global severity) and comorbidities (present personality disorder, present ADHD or autism)

Project Team

DyNOPS Logo

For further information, please contact:
Prof. Dr. Tim Kacprowski
tim.kacprowski[a.t_))tu-braunschweig.de
Telefon (0531) 391-9541

Scroll to Top