Using Large Language Models (LLMs) to understand medication shortage-related trends in social media

Start Date

15 September 2023

End Date

1 July 2025

Project Participants

Australian Government Department of Health Disability and Ageing, University of Technology Sydney

Status: Completed

Tags: artificial intelligence (AI), data analytics, health system, large language models (LLMs), medicines, public health, public health surveillance, regulatory affairs, responsible AI

Opportunity 

Medicine shortages create challenges for patients, healthcare services and regulators. Existing reporting systems often only highlight shortages after patient or market impacts are evident.  

This was a feasibility project that explored leveraging large language models (LLMs) to create a dashboard prototype for identifying and analysing  emerging medication shortage-related trends in social media. 

Project Objectives  

The project aimed to: 

  • assess technical, ethical and legal feasibility of using LLM technologies to reliably analyse unstructured, publicly-available social media data to identify signals relevant to medicine shortages 
  • develop a platform to explore whether this approach could provide earlier situational awareness of potential medicine shortages and related emerging trends. 

The project developed MedShort, an LLM-powered prototype platform that could collect, de-identify, analyse and provide a visual report on medicine shortage discussions from social media data. By establishing a reusable privacy‑preserving governance model aligned with Australian regulatory and ethical expectations, the project confirmed LLM-based social media monitoring could enhance early situational awareness of medicine-related events, but challenges remain on using publicly-available social media data as a supplementary signal source for medicine shortages. 

The project demonstrated publicly-available social media data may be able to supplement regulatory intelligence for medicine shortage monitoring. This provides a foundation for future applications in pharmacovigilance, adverse event monitoring and broader public health intelligence. The findings contributed directly to policy discussions on AI-enabled public health surveillance capabilities, and responsible use of alternative data sources in regulatory monitoring.

Integrity, Excellence,
Teamwork and Authenticity

Hand holding a smartphone against a colourful, defocused background