Deep Learning Models Predict Emergency Department Boarding Time
A recent study released on arXiv introduces a forecasting framework for multi-horizon time series aimed at estimating the boarding time in emergency departments (ED), which refers to the wait time for admitted patients to receive an inpatient bed. The research team assessed models at intervals of 6, 8, 10, 12, and 24 hours, utilizing actual data from a university-associated urban hospital in the U.S., alongside external factors such as weather, holidays, and local events. The DLinear and NLinear deep learning models, based on decomposition and normalization respectively, demonstrated exceptional performance across all timeframes. This framework seeks to facilitate proactive operational strategies to mitigate the ongoing global issue of ED overcrowding.
Key facts
- ED boarding time is the duration admitted patients remain in the ED while awaiting inpatient bed placement.
- The forecasting framework predicts boarding time at 6, 8, 10, 12, and 24-hour horizons.
- Real-world data from a university-affiliated urban hospital in the United States was used.
- External contextual data included weather, holidays, and major local events.
- DLinear and NLinear deep learning models showed superior performance.
- The study is published on arXiv with identifier 2605.18839.
- The framework supports proactive operational decision making.
- ED overcrowding is a persistent operational challenge worldwide.
Entities
Institutions
- arXiv
- university-affiliated urban hospital in the United States
Locations
- United States