An intelligent digital twin framework with AI-driven optimization for patient flow and clinical scheduling in smart healthcare systems.

The operation of the emergency departments at hospitals can be faced with many operational difficulties due to their unpredictable nature, resource scarcity, and increased pressure on services. In this paper, we propose a multi-level AI-enhanced digital twin framework to analyze the patterns of patient flow and clinical scheduling through a case study conducted with the help of three real-life datasets. Our framework incorporates three levels, namely temporal forecasting, clinical robustness, and outcome grounding, to generate an indication about hospital dynamics from real-life datasets. Temporal intelligence level incorporates an LSTM-based model that has shown competitive prediction performance (R² = 0.6785, MAE = 0.0895, RMSE = 0.1103). In addition, our clinical robustness level uses a triage dataset consisting of 560k records to generate an understanding of congestion pattern based on acuity and reveals that maximum congestion is observed at times ranging from 11:00 to 14:00 caused mainly by moderate-acuity patients (ESI-3). Results Layer. The results layer demonstrates the effect of the model, which shows that the mean waiting time is about 35 min. In addition, there is a negative association between the two variables-waiting time and patient satisfaction. Trend analysis across datasets reveals similarities in temporal trends, implying that the developed framework reflects representative features of the hospital environment. Generally, this study emphasizes the importance of incorporating AI-enabled predictions and digital twin models to aid decision-making concerning the analysis of patients' flow and assessment of scheduling policies.