Feasibility of forecasting future critical care bed availability using bed management data.

Area of Science:

• Health Informatics
• Computational Modelling
• Data Science

Background:

• Accurate forecasting of critical care bed availability is crucial for effective hospital resource management.
• Existing methods may require complex patient-level data, posing privacy and logistical challenges.

Purpose of the Study:

• To assess the feasibility of predicting critical care bed capacity using data-driven computational forecast modelling.
• To determine if routinely collected hospital bed management data is sufficient for forecasting.

Main Methods:

• A proof-of-concept, single-centre feasibility study utilizing retrospective analysis of prospectively collected data.
• Applied regression-based and classification data science techniques to hospital-wide bed management data.
• Forecasted critical care bed availability at 1, 7, and 14-day horizons.

Main Results:

• Demonstrated the feasibility of forecasting critical care bed capacity using only hospital bed management data and interpretable models.
• Achieved better predictive performance for 1-day forecasts (AUC 0.78) compared to 14-day forecasts (AUC 0.73).
• Feature importance analysis indicated reliance on critical care and temporal data, not data from other wards.

Conclusions:

• A data-driven forecasting tool utilizing solely hospital bed management data can predict critical care bed availability.
• This novel approach eliminates the need for patient-sensitive data in predictive modelling.
• Further research is warranted to refine this method for broader application in other hospital settings.