ISSN 1671-3710
CN 11-4766/R

Advances in Psychological Science ›› 2023, Vol. 31 ›› Issue (11): 2129-2141.doi: 10.3724/SP.J.1042.2023.02129

• Regular Articles • Previous Articles     Next Articles

Prediction of depression onset and development based on network analysis

ZHANG Weixia1, XI Min2, YIN Tiantian1, WANG Cheng1, SI Shubin3,4   

  1. 1Sports Department, Northwestern Polytechnical University,2Hospital of Northwestern Polytechnical University,3School of Mechanical Engineering, Northwestern Polytechnical University,4Key Laboratory of Industrial Engineering and Intelligent Manufacturing Ministry of Industry and Information Technology, Xi'an 710072, China
  • Received:2022-10-26 Online:2023-11-15 Published:2023-08-28

Abstract: Depression is a public health issue that needs to be addressed urgently in modern society, and prevention is one of the most effective ways to tackle this problem. The key to effective prevention is to accurately identify potential depression patients, capture warning signals of changes in depressive states, and take preventive measures timely. Traditional models of common cause consider depression as potential factors resulting from multiple symptom manifestations, neglecting the dynamic relationships among symptoms. From the perspective of a complex system, depression is a network system composed of multiple symptoms interacting with each other, and the structural and dynamic characteristics of this network can provide new theoretical perspectives and measurable indicators for predicting the occurrence and evolution of depression. Structural characteristics refer to the topological properties of symptom networks, while dynamic characteristics refer to the patterns that the network system exhibits during evolution. Starting from the key issue of predicting the occurrence and evolution of depression, this paper discusses the relationship between symptom networks and depression from a theoretical perspective, and further examines the performance of topological structure features and critical phenomena-related indicators of depression symptom networks in predicting depression onset and mutation. In terms of the structural features of symptom networks, connectivity, density, centrality, and hubs are able to predict the onset of depression. In terms of dynamic features, the presence of critical slowing down and critical fluctuations provides the basis for predicting phase transitions of depression system. However, there are some urgent problems to be solved in current research: (1) In the construction of symptom networks, the determination of node content is relatively single, often only including emotion, and other manifestations of depression are often ignored; (2) The biggest challenge of critical phenomena in depression research is that the appearance of relevant indicators is not synchronous with the change of symptoms. That is to say, there is no clear mapping relationship between clinical manifestations and warning indicators. In empirical research, the relationship between the occurrence of critical phenomena and phase transitions in depression systems is complex, and the occurrence of critical phenomena does not necessarily mean that phase transitions occur, and phase transitions may not necessarily be accompanied by the occurrence of critical phenomena. (3) The dynamics of the system include self-dynamics and interaction dynamics, and dynamic analysis is based on network structure analysis. However, in current empirical research, predictions based on network structure and those based on critical phenomena are artificially “divorced”, and there is no research on dynamic analysis based on network structure analysis. To increase the accuracy of early warning signals in predicting depression, future research should construct more systematic and comprehensive networks, and optimize the method of determining depression states by using integrated or machine learning-based warning indicators.

Key words: network, transition of depression, prediction, critical phenomenon, early warning signals

CLC Number: