2026 年 52 巻 5 号 p. 273-281
We evaluated the utility of a next-generation Artificial Intelligence (AI) image-analysis tablet identification system (“the system”) for unit-dose (also known as “one-dose package”) identification of patient-brought medications and examined the feasibility of task shifting/sharing from pharmacists to pharmacy assistants. Sixteen participants—12 pharmacists (four each with 1 – 3, 4 – 6, and ≥7 years of experience) and four pharmacy assistants—completed unaided visual identification and AI system–assisted identification in a within-subject comparative design study. Unit-dose pouches containing 2, 7, or 15 tablets were provided for identification to simulate patient-brought medications lacking accompanying drug information. The outcomes included identification time, electronic medical record (EMR) entry time, and accuracy. For all participant groups, the AI system significantly reduced identification time compared with visual inspection alone. No significant impact of AI assistance on EMR entry time was observed. No significant differences were observed in identification time, EMR entry time, or accuracy between pharmacists at different experience strata or between pharmacists and assistants. The overall accuracy was high, with only a few isolated errors across the two methods. These findings indicate that the AI system can improve workflow efficiency without compromising safety, supporting the practical feasibility of task shifting/sharing of routine unit-dose identification from pharmacists to trained assistants. The adoption of AI-assisted identification can enable pharmacists to allocate more time to advanced clinical activities, including therapeutic optimization, pharmacotherapeutic monitoring, and interprofessional collaboration.
