Project
Stress in Action: Modeling and Predicting the Effects of Stress

The NWO gravitation project Stress in Action capitalizes on the fast technological advances and big data analytics to move stress research from the lab to daily life. This project is part of the data analytic support core (DASC) team. The DASC will develop a variety of big data analytics approaches. Specific analytical questions for DASC include:
(1) How can the dynamic interaction between the contextual stress exposures and the multicomponent stress responses be best modeled? (2) How can these models account for individual differences in the effects of stress exposure? (3) How can we assess the predictive accuracy of these models.

As a major innovation, we aim to combine Machine learning (ML) and Dynamic intensive longitudinal data (DILD) techniques to interpret temporal data better. Whereas ML focuses more on data-reduction and prediction and DILD techniques on explanation and interpretability, it is anticipated that a combination of these methods will provide better insights to answer our research questions. More specifically, this project will apply joint modeling and multilevel modeling (vector-autoregressive multilevel models) to develop dynamic predictive accuracy measures for intensive longitudinal data (e.g., the prediction of mood). The random effects coming from such a model can be used for individualized predictions. In this way, we can see if and when complex multilevel models are overfitting and whether the predictive accuracy changes over time. The joint model component on the other hand can be used to predict probability and time of stress related long-term outcomes such as burnout, depression or cardiovascular diseases. Additionally, the temporal dynamics derived from VAR models could provide additional information about future health outcomes. By formally taking this relationship into account by extending the existing joint model formulation, we aim at improving our prediction and understanding on the onset of health outcomes. Ultimately, this could provide clinicians with tools that allow prognostic predictions for individuals to assess the current risk level and future trajectory – which would allow for timely interventions that prevent negative health outcomes.

Supervisors
Dr. Laura F. Bringmann
Dr. Dimitris Rizopoulos

Financed by
NWO Stress in Action Gravitation grant

Period
December 2023 – December 2028