Network Temperature as a Metric of Stability in Depression Symptoms Across Adolescence

Background: Depression is characterised by diverse symptom combinations that can be represented as a network. While studies often focus on central symptoms for targeted interventions, less explored are parameters capturing the entire network, offering insights into system dynamics. We propose network temperature as a metric for psychological network stability, examining changes across adolescence, a critical period for depression development. Methods: We replicated our study across three cohorts, including 35,901 adolescents (ages 10-19) from the Adolescent Brain and Cognitive Development study (ABCD), Avon Longitudinal Study of Parents and Children (ALSPAC), and Millennium Cohort Study (MCS). Depression symptom networks were estimated using multi-group Ising models, deriving the temperature parameter (T) at multiple time points (T=1 at baseline). Network temperature captures stability by assessing node state alignment – symptoms being ‘on’ or ‘off’. Higher temperatures indicate less alignment and increased system randomness. We also investigated sex differences in network temperature to illustrate developmental variability in stratified groups.Results: Network temperature decreased across adolescence in all cohorts from baseline (ABCD T=0.60 [0.59-0.61], ALSPAC T=0.70 [0.69,0.71], MCS T=0.75, [0.74,0.76]), with the steepest decline in early adolescence. Within-person networks stabilised with age, while individual depression scores increased in variability. Male networks cooled faster and earlier than females’, which showed higher intra-network heterogeneity and decreased stability.Conclusions: Depression symptom networks stabilise, and symptom variability decreases across adolescence. Faster network cooling at earlier ages highlights a critical transition period where early intervention could be most effective. Sex differences suggest that stratifying by risk factors could inform optimal intervention timing.