2022/09/06 by Jingmeng Cui, Merlijn Olthof, Fred Hasselman +1 · 1 voice
Psychology · Environmental Science · Decision Sciences · #Mental Health Research Topics #Ecosystem dynamics and resilience #Complex Systems and Decision Making
paper · pdf · doi:10.31234/osf.io/59fu4
openalex publication_date 2022/09/06 · openalex created_date 2022/10/01 · openalex updated_date 2026/07/14
Background: The past few years have seen a rapid growth in research on early warning signals (EWSs) in the psychopathology domain. Whereas early studies found EWSs to be associated with sudden changes in clinical change trajectories, later findings showed that EWSs may not be general across variables and cases and have low predictive power. These mixed results may be explained by the diverse methods employed in clinical EWS studies, with some of these approaches and practices potentially misaligned with the underlying theory of EWSs. Methods: We aim to examine the research methodology in the field of clinical EWSs based on the underlying theory and mechanisms. We show a derivation of EWSs for general multivariate dynamic systems, identify the key assumptions of EWSs, and compare them with the common research practices in this field.Results: We identify the following key assumptions for EWS validation studies: the system should depart from a point attractor that is no longer stable during the critical transition; EWSs should be calculated strictly before the transition; EWSs and the start of the transition should involve the same set of variables. The common research practices in the field were often not in line with those assumptions.Conclusions: We suggest future research in clinical EWSs should be designed as such the transition is visually checked with distance plots, the EWS assessment period is strictly before the transition, and a consistent set of variables is used. EWS as a theory-based prediction tool bears its merits compared with data-driven approaches, making it important to align the methodology closely with the underlying theory.