Large Language Model for Predicting Altered Mental Status in Dementia Patients Using Electronic Health Records

This study demonstrates the potential of using EHRs and LLMs to predict altered mental status in dementia patients. The LLMs showed fine predictive performance, while the CoT strategy provided enhanced interpretability at the cost of reduced accuracy. These findings emphasize the need to optimize LLM in clinical applications and highlight the promise of advanced AI tools in improving dementia care through efficient risk assessment and timely interventions.