Machine Learning in Assessing Intraoperative Blood Loss: A Systematic Review and Meta-Analysis.
Wenlin Zhou, Linglin Pan, Xinmei Pan and 3 others
PMID 41761466WHAT IT FOUND
Across 12 studies, machine-learning tools estimated surgical blood loss closely against reference methods.
Accuracy varied widely, was weakest in cesarean sections, and did not prove better patient outcomes.
Key findings
01The pooled analysis reported a correlation coefficient of 0.91 (95% CI 0.84 to 0.95) between machine-learning blood-loss estimates and reference standards.
02Results varied widely across studies (I2 = 99.73%), and subgroup correlations were lower for cesarean patients (r = 0.69) and for studies using the Triton system (r = 0.82).
03The review did not show absolute measurement accuracy because agreement data were generally unavailable, and patients with blood loss over 1500 mL were underrepresented.
STILL TO COME
How it was doneWhat they foundWhat it means for RNs
Read the rest of this summary
You get three full summaries a month, free, and we do not ask for a card. Search, the TL;DRs and your library stay unlimited either way.
What it does not show
The pooled result had extremely high heterogeneity (I2 = 99.73%), so the average correlation should not be read as uniform performance in every hospital or surgery. The review used correlation with reference standards, not agreement analyses, so it did not show whether machine-learning estimates were numerically accurate or systematically too high or too low. Patients with blood loss over 1500 mL were underrepresented, and most studies included patients with blood loss within the normal range. Egger's test suggested publication bias (p = 0.042), and only 12 studies were included. Subgroup differences may be confounded by country and modeling method, because many real-sample studies were from the United States and used Triton. The review included only English and Chinese literature, so studies from other regions may have been missed.
Declared interests
The supplied text lists funding from Shanghai Jiao Tong University and the Shanghai Rehabilitation Medical Association, but no competing-interest statement is reported.
The easy way to misread this
Do not conclude that machine-learning blood-loss estimates are proven to improve patient safety or replace nurse judgement. The review pooled association with reference standards, not agreement or clinical outcomes, and heterogeneity was extremely high (I2 = 99.73%).