What time-varying network models based on functional analysis tell us about the course of a patient's problem

Background: Relations among psychological variables are assumed to be complex and to vary over time. Personalized networks can model multivariate complex interactions. The development of time-varying networks allows to model the variation of parameters over time. Objectives: We aimed to determine the value of time-varying networks for clinical practice. Methods: We applied time-varying mixed graphical models (TV-MGM) and time-varying vector autoregressive models (TV-VAR) to intensive longitudinal data of nine participants with depressive symptoms (n = 6) or anxiety (n = 3). Results: Most of the participants showed temporal changes in network topology within the assessment period of 30 days. Time-varying networks of participants with small, medium, and large time variability in edge parameters clearly show the different temporal evolvements of dynamic interactions between variables. The case example indicates clinical utility but also limitations to the application of time-varying networks in clinical practice. Conclusion: Time-varying network models provide a data-driven and exploratory approach that could complement current diagnostic standards by reflecting interacting, often mutually reinforcing processes of mental health problems and by accounting for variation over time. They can be used to generate hypotheses for further confirmatory and clinical testing.