Drug Clustering in a Private Hospital of Chiang Mai Province with K-Means Clustering Technique

Abstract

OBJECTIVES: The study was to establish a methodology for clustering medicines into distinct groups, hence facilitating effective management within each category.

MATERIALS AND METHODS: This study employed the k-means clustering technique to examine the drug inventory data of a hospital. This research gathered data from a private hospital located in Chiang Mai province, Thailand. The dataset consisted of 1,501 unique items that were monitored over a period of 24 months. Each item was described by parameters such as importance, annual drug expenditure (ADE), variance, and mobility. The clinical importance of the medicine was determined by Vital, Essential, Desire Analysis (VED analysis), which involved input from healthcare professionals such as physicians, pharmacists, and inventory personnel. Subsequently, the chosen medication category was subjected to k-means clustering using the RapidMiner software, which also contributed to data preparation, transformations, and the identification of appropriate parameters for the clustering procedure.

RESULTS: The VED analysis shows the result of clinical importance grouping which are vital (213 items), essential (588 items), and desirable (700 items). Only the essential group is focused on and processed by k-means clustering analysis, due to higher complexity and contribution to various patient outcomes. The clustering analysis yielded six discrete clusters of drugs, each exhibiting distinct characteristics, and management strategies have been recommended for each cluster. Some clusters can be handled using simple methods or existing inventory management policy, while others need the use of statistical forecasting tools and/or machine learning techniques to precisely predict demand.

CONCLUSION: This study provides a baseline and framework for future studies and the management of drug inventory by using mathematical methods and models. The use of VED analysis also contributes significant data clarification in terms of clinical importance to the hospital. Integrating k-means clustering is one of the current machine learning technologies for inventory management, which is needed to improve the efficiency of patient care and to ensure more effective management of drugs. As a result, by introducing a new clustering technique, hospital staff members in charge of managing the drug inventory gain new insight and guidelines into inventory policy, which they can then apply as circumstances change and implement further.