Subproject 6

Digital Phenotyping of emotion (dys-)regulation as transdiagnostic process and proxy for clinical and neurobiological markers of treatment (non-)response

Predictors of treatment non-response (NR) that focus on emotion regulation (ER) should be ecologically valid, and readily available at low cost. Such predictors enable treatment-informing single case-predictions. Collecting neural and psycho-physiological predictors of NR in a laboratory context is time-consuming and difficult to implement outside of the university context. Clinical practice needs proxies that transfer the predictive value of neural NR-signatures and related ER-(dys-)functions of internalizing disorders into clinical settings. It can be assessed repeatedly with minimal effort for therapists and patients during different treatment periods (e.g., before, during, and after therapy). Embedded within the Research Unit, the current project uses digital phenotyping to derive such proxies. We use data actively and passively collected by personal electronic devices such as smartphones and wearables, to derive phenotypes of individuals at risk of NR based on indicators of ER as the overarching core construct. Smartphones are particularly well-suited for this purpose. First, smartphones allow assessing mood and cognition actively and ecologically valid by means of Ecological Momentary Assessment (EMA, i.e., repeated short surveys of the current state of mind and activities). Second, the sensor-rich environment of smartphones provides multimodal objective data about behavior, physiology, and mood collected passively through built-in applications and connected intelligent devices (e.g. wearables). The present project (SP6) collects, prior to a CBT treatment, sensor- and EMA data focusing on ER in n = 468 patients with mental disorders from the internalizing spectrum (e.g., patients suffering from specific phobia, social anxiety disorder, panic disorder, agoraphobia, generalized anxiety disorder, obsessive-compulsive disorder, post-traumatic stress disorder, unipolar depressive disorders). Furthermore, we collect sensor and EMA data (two additional measurement bursts at T20, and after twelve months (post)) throughout the entire treatment in a subsample of n = 350. The sensor-based assessment on mobile devices focuses on markers that are either 1) regulatory (e.g., physical activity, smartphone usage), 2) affect the individual's ability to regulate emotions (e.g., sleep, heart rate variability), or 3) serve as proxies for regulatory efforts or environmental characteristics that influence ER (e.g., GPS tracked spatial resolution of activity patterns, physical activity, and smartphone use). ER indicators are calculated based on the dynamics of self-reported emotional experiences and/or patterns in the self-reported application of ER strategies. Combined with the data modalities of the remaining subprojects, this subproject provides a unique opportunity to establish a comprehensive cross-level understanding of ER and its value for NR prediction in naturalistic settings.


Dr. Anette Brose

Associated Researcher

Benjamin Gas

Student Assistant

Leona Hammelrath, M.Sc.

Research Associate

Prof. Dr. Christine Knaevelsrud

Principal Investigator

Johanna Suchy

Student Assistant



Hammelrath L., Heinrich M., Burggraf P., Schenk, H., Knaevelsrud C. (2023, June 05-07). Tell me how I feel! Development of an Algorithm to Predict Depression Symptoms Based on Wearable Data. [Poster presentation]. Society for Ambulatory Assessment, Amsterdam, Netherlands.


Hammelrath, L., Hilbert, K., Heinrich, M., Zagorscak, P. & Knaevelsrud, C. (2023). Select or adjust? How information from early treatment stages boosts the prediction of non-response in internet-based depression treatment. Psychological Medicine, 1–10.