International Journal of Drug Delivery Technology
Volume 16, Issue 1s

Ai-Driven Pharmacovigilance Systems Using Multimodal Retrieval-Augmented Generation For Evidence-Based Drug Safety Monitoring.

Rama Krishna Kumar Lingamgunta *1, Shalmali Joshi 2, Ramkumar Raju, Djeabalaradjou3

1IT Principal Cigna Evernorth Services Inc,  Raleigh,  North Carolina,  USA Email ID : ramkumar2606@gmail.com, ramakrishnakumar.lingamgunta@evernorth.com ORCID: 0009-0001-6201-7620 BR> 2Senior Advanced Analytics Analyst, Atlanta, GA, USA ORCID: 0009-0000-4329-3841 Email: joshishalmalij@gmail.com BR> 3Independent Researcher, Healthcare IT Program Manager ORCID - https://orcid.org/0009-0002-7595-9411 Email: omniramkr@gmail.com


ABSTRACT

Pharmacovigilance plays a critical role in ensuring patient safety by monitoring and evaluating adverse drug events after medicines are introduced into real-world clinical practice. However, existing drug safety workflows remain heavily manual and fragmented, relying on labor-intensive review of structured clinical data, unstructured physician notes, patient narratives, and regulatory reference documents. This fragmentation limits timely signal detection and increases the risk of missed or delayed identification of serious adverse drug reactions. Recent advances in large language models offer new opportunities to support pharmacovigilance activities, but their use in regulated healthcare settings is constrained by concerns around hallucination, lack of transparency, and limited auditability. This paper presents an AI-driven pharmacovigilance system that leverages multimodal retrieval-augmented generation to support evidence-based drug safety monitoring. The proposed system integrates structured electronic health record data, unstructured clinical text, and external drug safety documents to retrieve relevant evidence prior to generating safety assessments. Rather than producing autonomous decisions, the system generates explainable summaries of potential adverse drug events, including temporal relationships, seriousness classification, and evidence-linked rationale, while preserving human oversight. Each generated output is grounded in retrieved source material and accompanied by traceable citations to support regulatory review and audit requirements. The architecture is designed to align with real-world pharmacovigilance workflows, enabling drug safety teams to prioritize high-risk cases, reduce manual review burden, and improve consistency in adverse event evaluation. By combining multimodal evidence retrieval with governance-aware generation, the proposed approach addresses key limitations of traditional pharmacovigilance systems and demonstrates how AI can be safely applied to support drug safety monitoring in regulated clinical environments..

Keywords: Pharmacovigilance, drug safety monitoring, adverse drug events, multimodal data integration, retrieval-augmented generation, clinical text analysis, electronic health records, explainable artificial intelligence, human-in-the-loop systems

How to cite this article: Lingamgunta RKK, Joshi S, Raju R, Djeabalaradjou D; Ai-Driven Pharmacovigilance Systems Using Multimodal Retrieval-Augmented Generation For Evidence-Based Drug Safety Monitoring...Int J Drug Deliv Technol. 2026;16(1s): 208-211; DOI: 10.25258/ijddt.16. 208-211