Mapping psychiatric comorbidity network: A pilot multi-method weighted network analysis with a focus on key disorders

Psychiatric comorbidity is a prevalent phenomenon that imposes a significant burden on patients, healthcare systems, and society. However, current research on psychiatric comorbidity is often limited to single disorders or partial associations. This study aims to utilize network analysis methods to construct a psychiatric comorbidity network and explore the network structural characteristics under different network weight definitions.

Based on the psychiatric outpatient data from Changhua Christian Hospital in Taiwan from January 1, 2016, to June 30, 2024, the ICD-10 diagnostic codes (F00-F99) of all patients that appeared at least three times were extracted. Three different comorbidity networks were constructed using co-occurrence counts, Jaccard index, and partial correlation coefficient estimated by the mixed graphical model (MGM) as the weights of the network edges. Network structure was analyzed using indicators such as degree centrality, modularity, and community detection.

The dataset included 16,954 patients. The comorbidity frequency network showed that mood disorders (F34) and anxiety disorders (F41) had the highest weighted degree centrality. In the Jaccard coefficient network, the weighted degree centrality of developmental disorders (F8x) increased. The MGM network highlighted the central role of substance use disorders (F1x).

Our findings suggested the roles and interrelationships of different disease categories in the comorbidity network. The results provide new perspectives and data support for clinical practice and future research.