ICU scoring systems: current perspectives and future directions.

Abstract

Purpose of review: This review aims to summarize the recent publications and future perspectives on the use of ICU scoring systems mainly for the assessment of ICU performance, resource use and benchmarking. Additionally, we provide current limitations and future directions on the use of scoring systems.

Recent findings: Generalizability and precision remain major challenges to the use of ICU-score systems. Recent innovations in this field have been driven by the expansion of national and international critical care registries, alongside advancements in data science.Models developed using data from specific regions lack broader applicability. Simplified scoring systems have been proposed to address the urgent need for a global ICU predictive model. Scoring systems can facilitate research, outcome prediction, and healthcare quality comparisons across different settings. A global ICU score system would need minimal data collection requirements, but its use would be inherently limited by the trade-off between generalizability and precision. In parallel, the search for more precise models has led to recent advances. Artificial intelligence-based models have improved predictive abilities compared to traditional scores. Omics data integration and diverse variables and dimensions may interact to predict outcomes. Dynamic models can update such predictions. However, implementation challenges persist, including the need for validation across diverse settings and addressing issues such as transparency, reproducibility, and potential biases.

Summary: Traditionally, ICU scoring systems enable the assessment of patients' severity of illness and consequently the risk-adjusted evaluation of ICU performance and resource use. The expansion of national ICU registries has advanced their use internationally for quality assessment, quality improvement and benchmarking. Novel approaches and methodologies, including the use of machine learning and data science, are making progress in improving the scores performance and expanding their use beyond risk-adjusted mortality.