2SPD-016 Smart counting: the role of artificial intelligence in controlled medication management

<h3>Background and Importance</h3> The management of narcotics and other controlled medications is a critical responsibility of hospital pharmacists (HP), who must ensure their traceability. The traditional manual counting process for these medications is prone to human error and time-consuming. The application of artificial intelligence (AI) in the counting of narcotics can improve inventory control and reduce the need for human supervision, optimising the management of these high-risk products. <h3>Aim and Objectives</h3> The objective of this study was to assess the ability of an AI model in counting narcotics and controlled medications, comparing its performance to the manual counting conducted by an HP. <h3>Material and Methods</h3> The study included 39 narcotic medications and seven other controlled medications. The inclusion criteria required more than five packages per specialty, and the pharmaceutical forms considered included tablets, vials, ampoules, patches, and containers of active ingredients in powder form. Procedure: A manual count of each medication was performed. An AI application was trained using images of complete packages and loose units for recognition. Images of both complete and loose packages were captured for each medication, and the AI performed the count. The AI’s results were compared with the manual count. If they did not match, the process was repeated up to three times before considering the AI count as incorrect. <h3>Results</h3> Out of the 46 available medications, 36 were included in the study. Seven were excluded due to having fewer than five packages, and three were excluded due to their pharmaceutical form. The AI correctly counted 20 packages (55.6%) on the first attempt, six (16.7%) on the second attempt, and three (8.3%) on the third attempt. In seven cases (19.4%), the AI failed to count accurately. <h3>Conclusion and Relevance</h3> AI has the potential to optimise the counting of controlled medications, saving time and human resources. However, its accuracy still depends on factors such as image quality, lighting, and packaging arrangement, indicating the need for further development of this technology to ensure more reliable and safer results before widespread implementation. <h3>References and/or Acknowledgements</h3> <h3>Conflict of Interest</h3> No conflict of interest