6ER-015 Harnessing OpenAI as a strategic tool for horizon scanning in hospital pharmacy practice

<h3>Background and Importance</h3> Artificial intelligence (AI) can be a valuable tool in the field of hospital pharmacy, especially in Horizon Scanning tasks. OpenAI offers an innovative approach in the field of Horizon Scanning, leveraging natural language processing and machine learning algorithms to analyse large amounts of data. <h3>Aim and Objectives</h3> Evaluation of the OpenAI feasibility, efficacy, accuracy, and reliability as a support tool for horizon scanning in hospital pharmacy practice. <h3>Material and Methods</h3> Up to five searches were conducted to compare three OpenAI models (Perplexity AI, Microsoft Copilot, and ChatGPT) in determining the viability of authorising the use of drugs in clinical trials in the near future. 1st Search: Search and analysis parameters were delimited, indicating the variables (drugs under study) and the type of estimation (percentage of success of authorisation). OpenAI’s natural language processing capabilities were used to analyse and extract relevant information from the data sources (www.clinicaltrials.gov). 2nd Search: Search parameters were narrowed (disease group, drugs). An indirect comparison between each of the drugs aggregated by pathology was requested. 3rd Search: In addition to the above, the search was restricted to a specific area (rare diseases). 4th Search: In addition to the foregoing, the latest Horizon Scanning report on orphan drugs prepared by a reference group was provided. 5th Search: a statistical estimation of the viability, efficacy, accuracy, and reliability was requested. <h3>Results</h3> All OpenAI models reached a common conclusion about the areas where new drugs are most likely to emerge in the near future: chronic diseases, weight control, and rare and genetic-based diseases. They identified up to 143 drugs with potential likelihood of approval. However, significant limitations were observed in the models’ ability to provide accurate and reliable information on statistical comparisons and probabilistic estimates. The models had difficulties providing data on the relative efficacy and safety. Furthermore, when narrowing or specifying search parameters, the models became blocked and began to provide incorrect/fabricated data. <h3>Conclusion and Relevance</h3> OpenAI could be a valuable Horizon Scanning tool in hospital pharmacy practice, offering the potential to improve the identification of emerging trends and technologies relevant to patient care. However, today they present significant limitations that require further development to ensure their reliability and accuracy in complex tasks. <h3>References and/or Acknowledgements</h3> <h3>Conflict of Interest</h3> No conflict of interest