Vol. 7, Issue 2, Part E (2025)

Background: Polypharmacy in geriatric patients is a critical health issue of global concern, which is linked to higher adverse drug events, hospitalization, and healthcare expenditures. Although deprescribing has become an effective strategy, the existing methods do not entail predictive analytics and tailored decision support features that are required to implement the strategy effectively in clinical practice. In this study, the authors introduce and analyze an AI-based system that can be used to augment pharmacist-driven deprescribing programs by applying machine learning to forecast risk and maximize drugs.
Methods: A systematic literature review was carried out according to PRISMA criteria, and the studies included were qualitative in nature and dealt with pharmacist-led deprescribing. On the basis of these insights, we created a new AI architecture, which combines natural language processing with electronic health record extraction, ensemble learning algorithms with adverse event prediction, and reinforcement learning with personalized deprescribing suggestions. A simulated patient cohort that represented clinical complexity in the real world was used to test the system.
Results: Compared to traditional clinical decision rules (AUC-ROC: 0.74-0.79), our AI framework performed better in terms of predictive accuracy (AUC-ROC: 0.92). The outcomes of the simulation showed the expected reduction of potentially inappropriate medication use by 34 percent and a reduction in adverse drug events by 28 percent in case of implementation in pharmacist-led deprescribing initiatives. The difficulties faced during implementation were defined as interoperability issues, workflow integration challenges, and the difference in the degree of the digital literacy of healthcare providers.
Conclusion: Artificial intelligence-based and pharmacist-led deprescribing is a ground-breaking concept in the management of polypharmacy. Our proposed framework will close major gaps in the contemporary deprescribing approaches since it offers a decision support based on data to improve the capacity of the pharmacist without losing the human aspect of patient care. In future studies, they are encouraged to validate and investigate the strategies of implementation in the real world so as to get maximum benefit out of AI-pharmacist collaboration.