Determining the relative importance of risk and protective factors for adjustment disorder symptoms during the COVID-19 pandemic by mixed-effects random forests
- Autor(en)
- Annett Lotzin, Emily Finne, Georg Schildbach, Elena Acquarini, Dean Ajdukovic, Marina Ajdukovic, Xenia Anastassiou-Hadjicharalambous, Vittoria Ardino, Ida Hensler, Filip K. Arnberg, Maria Böttche, Małgorzata Dragan, Margarida Figueiredo-Braga, Odeta Gelezelyte, Piotr Grajewski, Simon Groen, Marie José Van Hoof, Jana Darejan Javakhishvili, Evaldas Kazlauskas, Chrysanthi Lioupi, Brigitte Lueger-Schuster, Luisa Sales, Lela Tsiskarishvili, Irina Zrnic Novakovic, Ingo Schäfer
- Abstrakt
BACKGROUND: The COVID-19 pandemic exposed individuals to numerous psychosocial and health-related stressors associated with adjustment disorder (AjD) symptoms, yet it remains unclear which factors are most predictive.
METHODS: Using mixed-effects regression random forests (MERF), a machine learning approach that combines random forests with mixed-effects regressions, we analyzed longitudinal data from 15,155 adults across 11 European countries collected at three time points between June 2020 and January 2022. We evaluated 245 candidate predictors, including sociodemographic, pandemic-related, and health-related factors, for their relative importance in predicting AjD symptoms (ADNM-8).
RESULTS: The seven most influential predictors, ranked in descending order of importance, were uncertainty about the pandemic's duration and risks, poor health, social isolation, conflicts at home, loss of daily structure, fear of infection, and restricted personal contact with close others.
CONCLUSIONS: AjD symptoms were most strongly linked to factors related to lack of control (e.g., uncertainty, loss of daily structure, fear of infection), as well as current poor health and reduced social connectedness. Interventions that enhance a sense of control through clear communication, help individuals re-establish daily routines, and strengthen social connectedness may mitigate AjD symptoms during future public health crises. Our findings also highlight the potential of machine learning approaches for identifying complex patterns across high-dimensional predictors of clinical symptoms, which may improve prediction accuracy in mental health research.
- Organisation(en)
- Institut für Klinische und Gesundheitspsychologie
- Externe Organisation(en)
- Universitätsklinikum Hamburg-Eppendorf, Medical School Hamburg, Universität zu Lübeck, Università degli Studi di Urbino "Carlo Bo", University of Zagreb, University of Nicosia, Karolinska Institute, Uppsala University, Freie Universität Berlin (FU), University of Warsaw, Universidade do Porto, Universidade de Coimbra, Vilnius University (VU), De Evenaar, iMindU GGZ, Amsterdam UMC, Ilia State University, Centro de Saúde Militar de Coimbra
- Journal
- PSYCHOLOGICAL MEDICINE
- Band
- 56
- ISSN
- 0033-2917
- DOI
- https://doi.org/10.1017/S0033291726104048
- Publikationsdatum
- 06-2026
- Peer-reviewed
- Ja
- ÖFOS 2012
- 501010 Klinische Psychologie
- Schlagwörter
- ASJC Scopus Sachgebiete
- Applied Psychology, Psychiatry and Mental health
- Sustainable Development Goals
- SDG 3 – Gesundheit und Wohlergehen
- Link zum Portal
- https://ucrisportal.univie.ac.at/de/publications/5c07b8b6-efd9-4798-9db9-1972c13aadaf