WARN-D: developing an early warning system for depression in students
My ERC Starting Grant, funded with €1.5 million for 5 years as part of the Horizon 2020 research and innovation program, is focused on building the early warning system WARN-D to reliably forecast depression in young adults before it occurs. Why depression, and why prediction?
Mood disorders like depression are common, debilitating, and often chronic, and depression is considered among the most pressing health-related problems of modern society. Unfortunately, compared to breakthroughs in treating diseases like certain cancers, we haven’t made a lot of progress in increasing efficacy of depression treatments over the last 2 decades, and young people are disproportionately affected. This is why experts agree that prevention is the most effective way to change depression’s global disease burden. And the biggest barrier to successful prevention is to identify people at risk for depression in the near future — and we cannot do that at the moment.
3 interdisciplinary pillars of the grant
We will try to solve the challenge who should receive prevention, and when, by developing the personalized early warning system WARN-D. To figure out the problem of personalized detection, my team and I will follow 2,000 young adults over 2 years, and integrate emerging theoretical, measurement, and modelling approaches from different scientific fields so far unconnected.
- Regarding theory, we conceptualize depression as a complex, dynamical, biopsychosocial system in which causal relations and vicious cycles between a host of variables can move the system from a healthy to a clinical state, consistent with the Network Approach to Psychopathology that I co-developed. This borrows heavily from the growing discipline of complexity science that has led to massive breakthroughs in other disciplines.
- Regarding measurement, in addition to traditional mental health surveys every few months, we will assess how young adults are doing in their daily lives. To do so, we will use smart-phone based ecological momentary assessment (EMA) and collect temporal dynamics of variables like feelings, experiences, thoughts, and behaviors. Further, we will collect smart-watch based digital phenotype data such as sleep patterns or activity.
- Regarding statistical modeling, we will use dynamical network models to study the relations among these within-person mood systems, and use parameters of these models, combined with baseline, EMA, and digital phenotype data—as well as theory-driven predictors from the early warning literature—to construct the prediction model WARN-D via state-of-the-art machine learning models.
Overall, this project combines and integrates numerous modern tools to develop a tailored early warning system with the goal to forecast depression reliably before it occurs. Improving detection of future onset is a necessary next step to enable tailored prevention programs to kick in.
More specific goals include, among others, to:
- Map the (healthy) biopsychosocial mood system, and clinical deviations from it
- Integrate self-report EMA data with more objective digital phenotype data in statistical models
- Explore interindividual differences in within-person mood systems
- Find out how variables in the external field (e.g. one-time stressors or personality traits) affect the mood system
- Test the long-term stability of the mood system
- And understand the nature of depression onset as well as early warning signals for depression
We hope that WARN-D can, after validation, be transferred to other populations and problems. For instance, it could be used to forecast PTSD in soldiers going into a war zone, or to forecast burnout in teachers or medical professionals during stressful periods, or to forecast upcoming manic or depressive episodes in patients with a history of bipolar disorder. So keep your eyes open for future WARN-X projects!
Here is a 4-minute video where I introduce the study (recorded 2020): https://www.youtube.com/watch?v=fsq_XaiXjxo
Like science in general, grant writing is a team sport, and I want to thank everybody who has supported the project. For any ideas related to the grant — or if you think your CV fits exceptionally well and you’d like to work with us — please reach out to me. We are always looking for thesis students and interns, and plan to hire a 3-year postdoc at the end of 2022.