Critical care outcome prediction equation model, version 7.

Area of Science:

• Health Services Research
• Biostatistics
• Critical Care Medicine

Background:

• Mortality prediction models are crucial for benchmarking, auditing, research, and epidemiological studies in healthcare.
• Accurate prediction of intensive care unit (ICU) mortality risk is essential for resource allocation and quality improvement.

Purpose of the Study:

• To describe the development and validation methodology for version 7 of the critical care outcome prediction equation.
• To monitor adult ICU mortality risk using a parsimonious model derived from administrative data.

Main Methods:

• A multiphase study involving data extraction, curation, aggregation, modeling, and validation.
• Utilized jurisdictional administrative datasets from 28 public and 19 private hospitals in Victoria, Australia.
• Included 215,148 adult hospital separations receiving intensive care over five years (July 2019-June 2024).

Main Results:

• The overall case fatality rate (CFR) was 12.0 per 100 separations.
• Model performance metrics included a Brier score of 0.08 and an area under the receiver-operator characteristic curve of 0.85.
• 54.1% of CFR outlier ICU-years were reclassified as SMR inliers, indicating improved classification accuracy.

Conclusions:

• Version 7 of the critical care outcome prediction equation is a parsimonious model for predicting hospital mortality in adult ICU admissions.
• The model is derived from administrative data common across jurisdictions, facilitating broader application.
• The validated model provides a reliable tool for monitoring ICU mortality risk.