Author: M. S. Heimvik
Co-Authors: M. A. Helmich, K. M. Pålerud, R. Hagen, A. Hoffart, S. U. Johnson
Institution: University of Oslo, COPE (Complexity in treatment Outcome, Psychopathology and Epidemiology) research group, Modum Bad psychiatric hospital
This project explores how machine learning can be used to personalize treatment for patients with complex anxiety disorders. By analyzing routinely collected clinical data from Modum Bad (2016–2024), we aim to predict which treatment, Cognitive Behavioral Therapy (CBT) or Metacognitive Therapy (MCT), offers the greatest benefit for each patient.
- Apply the Personalized Advantage Index (PAI) to estimate individual treatment response.
- Investigate whether routinely collected baseline measures can be used to inform optimal treatment selection.
- Examine how clinical complexity influences treatment benefit and assignment.
The dataset consists of real-world clinical data from Modum Bad Psychiatric Centre, including psychometric assessments taken pre- and post-treatment.
- Feature selection using tree-based models (e.g.,
mobForest) - Outcome prediction via cross-validated regression and PAI
- Evaluation using clinical and statistical metrics (e.g., BAI, Reliable Change Index)
- Analysis: Summer 2025
- Manuscript: Winter 2025
- Submission: Winter 2025
This project is preregistered. For detailed hypotheses, methodology, and analysis plans, see Heimvik, M., Helmich, M. A., & Johnson, S. U. (2024, September 20). Predicting Optimal Treatment Outcomes for Patients With Complex Anxiety Disorders. https://doi.org/10.17605/OSF.IO/TNKZY
For questions or collaboration inquiries, please contact:
Margrete S. Heimvik – margrsh@uio.no