Dr Kenji Fujita, Dr Marion Matheson, Dr Bhavya Valecha and Professor Sarah Hilmer recently published a study in Drug Safety.
Their study developed a two-stage system that combines rule-based extraction with an open-source large language model to identify deprescribing recommendations in hospital discharge summaries. Tested on 850 discharge summaries from six NSW hospitals, the system achieved an F1 score of 0.91 and can be run securely on a local computer. It could be used as a screening tool, with clinicians reviewing the flagged recommendations.
Link to publication: Development and Validation of a Two-Stage NLP System Combining Rule-Based Extraction and LLM-Based Classification for Deprescribing Recommendations in Discharge Summaries
