Machine Learning in Assessing Intraoperative Blood Loss: A Systematic Review and Meta-Analysis

Abstract

Aim

To evaluate the value of machine learning in assessing intraoperative blood loss by comparing associated outcomes with those of the gold standard.

Background

Intraoperative bleeding is a leading cause of death in surgical patients and may be preventable through early and accurate assessment of blood loss. Machine learning models are used for measuring intraoperative hemorrhage with conventional assessment methods. However, outcome metrics vary across studies.

Methods

A systematic review and meta-analysis. Data were retrieved from Web of Science, PubMed, Embase, Cochrane Library, and CINAHL, with searches conducted through August 18, 2025.

Results

Twelve studies were included. The pooled correlation coefficient between machine learning models and the gold standard for assessing intraoperative blood loss was high.

Discussion

Machine learning models demonstrate high accuracy and reliability in assessing intraoperative blood loss. Heterogeneity was high, likely attributable to differences in publication year, country, study subjects, sample type, and modeling method.

Conclusion

Models should be promoted for clinical use to improve blood loss assessment accuracy and to potentially reduce perioperative risk.

Implications for Nursing

Novel machine learning models could enhance the accuracy and applicability of existing models, providing nursing staff with a more efficient tool for assessing blood loss. This will optimize the nursing decision-making process, reduce adverse events caused by underestimating or overestimating blood loss, and improve patient safety.

Implications for Nursing Policy

We provide a reference for exploring the application of artificial intelligence in other nursing fields, promoting interdisciplinary research and driving continuous innovation and progress in nursing.