Duplicate Therapy Risk Identification in Order Sets Using Network Analysis

This project investigated patterns of duplicate medication therapy within standardized inpatient order sets using social network analysis. Initiated by the Patient Safety and Quality team of Advocate Health’s Division of Pharmacy, the study aimed to identify clinically relevant duplication risks not evident through traditional one-by-one order set reviews. Electronic Medical Record (EMR) data from October 2024 were used to construct a medication co-occurrence network derived from over 815,000 active orders across multiple healthcare markets. In the network, nodes represented medications and edges denoted co-occurrence within the same order set. NetworkX in Python was employed to calculate centrality metrics, detect community structures, and visualize network topology. Key findings included the identification of highly central and frequently duplicated medications, particularly among perioperative pain management and diagnostic support regimens. Opioid analgesics, specific antibiotic classes, and insulin analogs formed dense clusters, highlighting areas of therapeutic redundancy risk. During exploratory analysis, pharmaceutical subclass was selected as the primary categorization, improving clinical specificity compared to broader therapeutic classes. The insights generated provide practical value for refining order set design, strengthening clinical decision support tools, and mitigating duplication-related safety risks. This project demonstrated the utility of network analysis in uncovering systemic patterns in medication use and established a foundation for future work involving modularity scoring, community refinement, and expanded duplicate therapy mapping.