Investigation of Drug Interactions through Analysis of Prescribed Medications Association Rules Using the FP-growth Algorithm

The discovery of hidden patterns in pharmaceutical data can contribute to improving the performance of hospital pharmacies. One of the applications of advanced data analysis techniques is the identification of drug interactions.

This study was conducted using data mining techniques with the FP-growth algorithm in the RapidMiner Studio® 10.1 environment to extract association rules and frequent pharmaceutical patterns. Data preprocessing and modeling were performed based on the CRISP-DM model. The type and level of drug interactions were determined based on the algorithm's results and by referencing the database at www.drugs.com.

The results included 17 association rules and 126 prescribing patterns, ranging from single-drug to four-drug combinations. Of the 64 two-drug prescribing patterns, 56 had no interaction, 6 had moderate interactions, 1 had a minor interaction, and 1 had a major interaction. Additionally, of the 19 three-drug patterns, 18 had no interaction, and only 1 had a moderate interaction. No interactions were observed in the four-drug prescribing pattern.

The findings of this study can assist stakeholders in improving the pharmaceutical supply chain, optimizing prescriptions, reducing drug interactions, and lowering costs. The discovered patterns may also be used as part of a clinical decision support system. Although no significant drug interactions were observed in this study, the discovery of even one major interaction highlights its importance and further underscores the practical role of computer applications in medicine.