The drug prescription network: a system-level view of drug co-prescription in community-dwelling elderly people.

Networks are well suited to display and analyze complex systems that consist of numerous and interlinked elements. This study aimed at: (1) generating a series of drug prescription networks (DPNs) displaying co-prescription in community-dwelling elderly people; (2) analyzing DPN structure and organization; and (3) comparing various DPNs to unveil possible differences in drug co-prescription patterns across time and space. Data were extracted from the administrative prescription database of the Lombardy Region in northern Italy in 2000 and 2010. DPNs were generated, in which each node represents a drug chemical subclass, whereas each edge linking two nodes represents the co-prescription of the corresponding drugs to the same patient. At a global level, the DPN was a very dense and highly clustered network, whereas at the local level it was organized into anatomically homogeneous modules. In addition, the DPN was assortative by class, because similar nodes (representing drugs with the same anatomic, therapeutic, and pharmacologic annotation) connected to each other more frequently than expected, indicating that similar drugs are often co-prescribed. Finally, temporal changes in the co-prescription of specific drug sub-groups (for instance, proton pump inhibitors) translated into topological changes of the DPN and its modules. In conclusion, complementing more traditional pharmaco-epidemiology methods, the DPN-based method allows appreciatiation (and representation) of general trends in the co-prescription of a specific drug (e.g., its emergence as a heavily co-prescribed hub) in comparison with other drugs.