A Temporal Network Analysis of Drug Co-Prescription Around Antidepressants and Anxiolytics Uses

OBJECTIVES: The implementation of network analysis in medicine and epidemiology started to gain popularity. A network-based approach captures the relationships of elements in a complex system. This study evaluates the relationships of antidepressants and anxiolytics with other drug therapies in a pharmacy dispensing database.

METHODS: Exploratory temporal network analysis was performed using the University of Groningen prescription database, IADB.nl. A closed cohort with at least one prescription of anxiolytics/antidepressants dispensed anywhere from 2018 to 2022 was extracted. The drug prescription was evaluated daily using anatomical therapeutic chemical (ATC) classification, where psychopharmaca was evaluated at ATC level 4, drugs for the nervous system at ATC level 3, and other medications at ATC level 1, distinguished into a total of 24 therapeutic medication classes. An undirected drug prescription network was constructed to capture co-prescription. Eigenvector centrality (c_e) was computed to measure relative nodal importance, which reflects the degree of medication concurrence in a network. The number of prescriptions, number of patients, claim-to-patient ratio, and eigenvector centrality were extracted as weekly-aggregated time-series data. Singular spectrum analysis was performed to decompose the data into its constituting trend, harmonics, and noise components.

RESULTS: Antidepressants (c_e: 0.09) and anxiolytics (c_e: 0.04) are psychopharmaca with high eigenvector centrality. These classes of medication are likely to be co-prescribed apart from other medications. Further groups with high eigenvector centrality were medications for alimentary and metabolism (c_e: 0.21), cardiovascular (c_e: 0.19), respiratory (c_e: 0.09), blood (c_e: 0.08), and analgesics (c_e: 0.05). Among the seven medication groups with high eigenvector centrality, only antidepressants showed a positive time trend for both eigenvector centrality and the actual number of claims.

CONCLUSIONS: Seven medication therapy classes, three of which are psychopharmaca, had a high eigenvector centrality, indicating concurrence with other medication groups on a population level in the group of antidepressant or anxiolytic users.