Using LLMs to analyse medicine shortages in social media
AAII will lead a cross-disciplinary team of experts to leverage the AI capabilities of LLMs for proactive healthcare solutions to medicine shortages.
By using LLMs in the novel context of social media discussions of medication shortages, this project leverages digital health advancements and demonstrates the transformative potential of AI in health informatics.
A cross-disciplinary team of experts from UTS's AAII and Faculty of Health will partner with DHCRC and the Department of Health and Aged Care to apply the AI capabilities of Large Language Models (LLMs) to analyse medicine shortages through social media data.
The two year project, led by AAII's Associate Professor Guodong Long, will employ LLMs to develop a framework which can analyse and identify medication-related trends in publicly available social media content.
The primary focus will be upon causes, impacts and potential mitigation strategies when it comes to medicine shortages in social media discussions on major social media platforms, such as Twitter, Reddit, Facebook, LinkedIn and TikTok. Currently, there is a significant gap in timely responses to medicine shortages both locally and in international supply chains.
Using AI-powered LLM technology, the team will create a tool that provides timely insights into medicine shortages, facilitating early warning systems, informed decision-making and targeted interventions. The tool promises to harness the power of both social media and AI, transforming unstructured data into both accessible and actionable knowledge.
The initial phases of the project will involve investigating the feasibility of setting up the LLM technology infrastructure and processing the received text relating to medication shortages from major social media platforms. The team will address the technical challenge of enabling the LLM to process unstructured social media data posed by the informal language, slang, abbreviations, and emojis used in public posts. Notably, the expected outputs are purely statistical, with a focus upon keywords extracted from enormous volumes of received text, as opposed to investigating individual social media accounts.
The later phases will involve fine-tuning the LLM model to accurately recognise and categorise discussions related to medication shortages, including understanding the context, sentiment, and the underlying themes. Ultimately, the team will identify trends and analyse public sentiment related to medication shortages and refine the model based on these findings, iteratively enhancing its performance.
The project's impact will manifest in improved healthcare decision-making and patient outcomes, informed policy directives, and enhanced understanding of patient behaviours and public perceptions. By generating actionable insights, the project has the potential to revolutionise how healthcare systems respond to medicine shortages and contribute to a more resilient and efficient healthcare infrastructure.
