Description

Adequate resource allocation in Intensive Care Medicine is especially challenging due to limited resources and increasing demands for ICU capacities due to an aging population and medical advances. Several studies in the past were trying to predict ICULOS using different models. The Objective and aim of our retrospective multicenter study are to forecast ICU length of stay (ICULOS) and length of mechanical ventilation (LOMV) in ICU patients of different groups (regarding gender, age group, medical vs surgical admission) worldwide for the next years up to the year of 2040 using statistical forecasting models.

To achieve this objective, historical ICU data spanning from 2005 to 2023 is collected from international ICU databases worldwide as well as population data from national and international databases and employ different statistical forecasting models (ARIMA-Model (Auto-Regressive Integrated Moving Average), logistic regression, Poisson Regression and ETS (Exponential smoothing)) to make these predictions. The Validity of the 4 different models is assessed with out-of-time-cross validity by splitting the data in 2 subsets for generation and testing of the model in a ratio of approximately 75:25 of the dataset. The most valid model of the 4 different models will be chosen. The statistical analysis follows he guidelines for Accurate and Transparent Health Estimates Reporting (GATHER Statement) von Stevens et al. from the year 2016.

The ultimate goal of this project is to provide valuable insights to healthcare system decision-makers worldwide regarding future requirements of ICU beds and ventilator capacities. With this insight we want to enable healthcare- system decision makers worldwide to proactively anticipate and allocate appropriate ICU resources for the future.