Artificial Intelligence and Large Language Models as Decision-Support Tools in Hospital Compounding Pharmacy: A...

Area of Science:

• Pharmacy Practice
• Artificial Intelligence in Healthcare
• Drug Compounding

Background:

• Hospital compounding pharmacies face challenges in efficiently retrieving critical information for pediatric extemporaneous preparations.
• Assessing drug crushability, regulatory compliance, and formulation feasibility requires accessing and consolidating data from multiple sources.

Purpose of the Study:

• To evaluate the efficiency and reliability of a large language model (LLM) as a decision-support tool in hospital compounding pharmacy.
• To assess the LLM's capability in handling pediatric extemporaneous preparations, focusing on drug crushability, regulatory compliance, and formulation feasibility.

Main Methods:

• A proof-of-concept study compared an LLM-assisted workflow (Claude Sonnet 4.6) with traditional manual information retrieval.
• The LLM was prompted to extract data from authoritative sources including drug crushability lists, Summary of Product Characteristics (SmPC), off-label use lists, and formulation stability databases.
• Two drugs, propranolol hydrochloride and imatinib mesylate, were analyzed for crushability, regulatory status, and formulation data.

Main Results:

• The LLM-assisted workflow reduced information retrieval time to under 2 minutes per drug, compared to a mean of 20 minutes for the manual process.
• The LLM achieved 100% output completeness and accuracy in classifying drug crushability and regulatory status.
• Specific formulation data, including stability for propranolol (146 days) and alternative data for imatinib (30 days refrigerated), were accurately retrieved.

Conclusions:

• A properly configured LLM can serve as an effective decision-support tool in hospital compounding pharmacy, enhancing efficiency and maintaining accuracy.
• The LLM system demonstrates potential for integration into routine pediatric galenical preparation practices, improving workflow and compliance.