Objectives: This study analyzes factors influencing Emergency Department (ED) overcrowding by examining the impacts of operational, environmental, and external variables, including weather conditions and football games. Materials and Methods: This retrospective observational study analyzed emergency department (ED) tracking and hospital census data from a southeastern U.S. academic medical center covering 2019–2023. These data were merged with corresponding weather, football event, and federal holiday data. The dependent variable was the hourly waiting count in the ED, our operational measure of overcrowding. Seven regression models were developed to assess different predictors across various timestamps. Results: Weather conditions were significantly correlated with increased ED waiting count in the Baseline Model, while federal holidays and weekends were consistently correlated with reduced waiting counts. Boarding count was positively correlated with ED waiting count when concurrent, but boarding counts 3 h and 6 h before showed significant negative correlations. Hospital census showed a negative correlation in the Baseline Model but shifted to a positive effect in other models, reflecting its time-dependent influence on ED operations. Football games 12 h before significantly correlated with increased waiting counts, while games 12 and 24 h after had no significant effects. Discussion: While existing research typically focuses on limited variables and narrow timeframes, the temporal relationships between operational and non-operational factors affecting ED overcrowding remain understudied, particularly the delayed impacts of external events and environmental conditions. Conclusions: This study emphasizes the importance of incorporating both operational and non-operational factors to understand ED patient flow. Identifying robust predictors such as weather conditions, federal holidays, boarding count, and hospital census can inform dynamic resource allocation strategies to mitigate ED overcrowding effectively.

Objective: Emergency department (ED) returns for mental health (MH) conditions pose a substantial burden on healthcare systems. Traditional machine learning (ML) models have shown promise in predicting ED returns but often fall short in clinical interpretability. This study evaluates whether integrating Large Language Models (LLMs) with ML methods can enhance both predictive performance and interpretability. Methods: We conducted a retrospective analysis of 42 464 ED visits involving 27 904 unique MH patients at an academic medical center in the deep south (2018-2022). Two primary outcomes were assessed: (1) 30-day ED return prediction, and (2) interpretability of model outputs. We employed LLaMA 3 (8B) for few-shot classification of chief complaints and social determinants of health (SDoH) features, including tobacco, alcohol, substance use, exercise, and home environment. These LLM-derived features were integrated into an XGBoost model. Model explanation was enhanced using a novel LLM-SHAP hybrid framework. Results: The LLM-based chief complaint classifier achieved 0.882 accuracy, 0.95 precision, 0.88 recall, and 0.86 F1-score, outperforming traditional models. SDoH classifiers reached F1-scores ranging from 0.67 to 0.96 across different categories. Incorporating LLM-extracted features into XGBoost improved AUC from 0.74 to 0.76 and AUC-PR from 0.58 to 0.61. The LLM-SHAP framework provided enhanced contextualized explanations for risk predictions, improving interpretability at both patient and population levels. Conclusion: Integrating LLMs with traditional ML models modestly improved predictive performance but substantially enhanced interpretability. This hybrid approach offers a promising path toward more actionable and explainable clinical decision support tools in emergency psychiatry.