Time Series Forecasting: A Powerful Approach for Short-Term ICU Census Prediction in Critical Care (Motivated by the Study on ICU Capacity Planning Using MODS and NEMS by Murray et a

Accurate forecasting of intensive care unit bed occupancy is critical for managing clinical resources, optimizing staff allocation, and maintaining patient care quality. This report explores the role of time series forecasting in predicting short-term intensive care unit census, drawing on a recent study that combined historical arrival patterns with patient-level data to develop a hybrid forecasting model. Using daily data from a large medical-surgical intensive care unit, researchers implemented a time series approach for estimating incoming patient volume and integrated a survival model based on admission characteristics to project patient length of stay. The combined method outperformed traditional forecasting techniques such as moving averages and standard autoregressive models, delivering more accurate and stable projections across 1- to 7-day horizons—even during periods of high variability, such as the COVID-19 pandemic. Key evaluation metrics, including root mean squared error and prediction interval coverage, demonstrated the model’s robustness and practical utility. The report outlines the data requirements, model validation steps, and visualization tools used to assess performance. It concludes with a discussion of the strengths and limitations of time series forecasting and identifies future directions for enhancing model adaptability through advanced analytics and real-time data integration in high-acuity healthcare environments.