Enhancing Outpatient Care Using an LLM-Based Model for Clinical Ward Disposition

Enhancing outpatient care in the Emergency Department (ED) requires optimized patient flow to reduce waiting times and improve hospital resource utilization. This work presents a Clinical Ward Disposition (CWD) system designed to optimize patient flow and support admission decisions following initial nursing assessments. The proposed system uses a hybrid CNN–BiLSTM model for optical character recognition (OCR) to digitize handwritten admission notes, achieving an accuracy of 86.57%. The extracted clinical information is analyzed using ClinicalBERT and XLNet models to determine the most appropriate ward for patient admission, achieving a disposition accuracy of 90.44%. The proposed system improves patient care by automating note digitization, extracting clinical data, and supporting efficient ward disposition decisions.