Large Language Models Support Antimicrobial Stewardship Pharmacy Services

Large language models (LLMs) using artificial intelligence (AI) exhibit significant potential in supporting antimicrobial stewardship efforts for clinical pharmacists, according to a study in Exploratory Research in Clinical and Social Pharmacy.¹ With the knowledge and reasoning to rationalize antibiotic use, LLMs showed the ability to identify contradictions and facilitate collaboration between pharmacists and AI.

“AI has the potential to improve resistance surveillance and prediction, optimize antimicrobial stewardship, and accelerate the discovery of new medications and diagnostics,” according to a study in Infection Prevention in Practice.² “The incorporation of AI-driven methodologies into antimicrobial resistance (AMR) control efforts marks a move from reactive to predictive, data-driven decision-making.”

The researchers utilized a dual-framework evaluation to test 3 mainstream models, DeepSeek, DouBao, and Qwen, on their ability to handle both theoretical pharmacy knowledge and practical clinical case analysis.

In the theoretical portion, which involved a standardized 100-point examination based on national antimicrobial guidelines, the models achieved a respectable average score of 86. However, their performance was even more impressive when tasked with reviewing 20 real-world irrational prescriptions, where the average score climbed to 92.3.¹

READ MORE: Amazon, Fleming Initiative Partner to Combat Antimicrobial Resistance

This finding is particularly notable because it runs counter with the “knowledge-practice gap” often observed in international AI research, where models typically score lower on practical tasks than on theoretical tests. The models proved exceptionally adept at identifying clinical contradictions, such as an antibiotic being prescribed without a matching diagnosis.¹

This high level of accuracy in practical reasoning suggests that AI can serve as a powerful ally in the fight against AMR, a crisis that was directly responsible for 1.27 million deaths globally in 2019. Beyond mere prescription review, this emerging pharmacointelligence is optimizing the entire medication-related workflow, from inventory management to personalized therapy adjustments.^2-4

In specialty pharmacy, according to US Pharmacist, AI is already being used to accelerate drug discovery, significantly shortening the traditional 12-year development timeline for new antibiotics. For instance, the AI platform Pharma.AI was instrumental in developing INS018_055, an antifibrotic molecule that moved from initial discovery to phase I clinical trials in under 30 months.⁴

This shift toward predictive, data-driven discovery is essential for rejuvenating a pipeline that has long been stagnant due to high costs and low financial returns.²

Despite these technological leaps, the human element remains the cornerstone of antimicrobial stewardship programs. Pharmacists are vital members of multidisciplinary teams, performing prospective audits, de-escalating therapies, and providing essential education to both patients and health care providers, according to a study in Cureus.⁵

Systematic reviews have shown that pharmacist-led interventions significantly improve guideline compliance and reduce the time to optimal therapy without negatively impacting patient readmission rates.

AI is viewed not as a replacement for these clinicians but as a tool to mitigate pharmacist burnout by automating routine administrative tasks and handling the heavy workload of initial prescription screening.^1-3

However, the integration of LLMs into clinical practice is not without risks.^3,4

The main study noted that despite LLMs excelling at logic, they often struggle with institution-specific compliance rules and localized administrative guidelines. Furthermore, concerns regarding algorithmic bias, data privacy, and the lack of empathy in AI interactions necessitate continued pharmacist oversight.^1,3,4

Moving forward, the goal is a collaborative “human-machine” model where AI handles the preliminary screening of prescriptions, allowing clinical pharmacists to focus their expertise on complex cases and direct patient care.¹

By extending these technology-supported stewardship services into ambulatory and urgent care settings, pharmacists can better address the global threat of AMR. This balanced approach ensures that while AI transforms the speed of health care, the professional judgment and ethical responsibility of the pharmacist remain at the center of patient safety.^2-5

“[LLMs] may support antimicrobial stewardship efforts by identifying clinical logic contradictions and facilitating pharmacist-AI collaboration,” concluded the authors of the current study.¹ “Further implementation research is needed to integrate such tools into pharmacy workflows and evaluate their impact on prescribing practices.”

READ MORE: Infectious Disease Resource Center

REFERENCES

1. Zhang W, Xu J, Dong T, et al. Exploring large language models as a prescription decision support tool for rational antibiotic use: a dual-framework analysis using standardized examinations and real-world clinical cases. Explor Res Clin Soc Pharm. July 4, 2026:100821. doi:10.1016/j.rcsop.2026.100821

2. Salama RA, Abdel Kader RG, Wadid NA. Artificial intelligence in combating challenges in antimicrobial resistance: a narrative review. Infect Prev Pract. 2026 Feb 14;8(2):100522. doi: 10.1016/j.infpip.2026.100522.

3. AI in pharmacy: how artificial intelligence is transforming patient care. National Association of Boards of Pharmacy. April 8, 2026. Accessed July 8, 2026. https://nabp.pharmacy/news/blog/ai-in-pharmacy-how-artificial-intelligence-is-transforming-patient-care/

4. Tignor K, Schalliol L, Glover A. Artificial intelligence and the future of specialty pharmacy. US Pharmacist. 2025;50(1):29-32. https://www.uspharmacist.com/article/artificial-intelligence-and-the-future-of-specialty-pharmacy

5. Dighriri IM, Alnomci BA, Aljahdali MM, et al. The role of clinical pharmacists in antimicrobial stewardship programs (ASPs): a systematic review. Cureus. 2023 Dec 8;15(12):e50151. doi: 10.7759/cureus.50151.