Prediction of elective surgical red blood cell transfusion reveals surgeon-level variation

Surgeon identity meaningfully improves preoperative prediction of allogeneic RBC transfusion in 279,905 elective procedures at Helsinki University Hospital, increasing AUPRC from 0.440 to 0.472 (+0.032).

• Why it matters: Patient and procedure factors do not fully explain transfusion variation, and quantifying surgeon-level contribution can improve patient blood management and elective blood supply planning.
• What they did: In a retrospective 2021–2025 cohort, four CatBoost models predicted any allogeneic RBC transfusion through postoperative day 7 using only variables available ≥7 days pre-surgery, with performance measured by AUPRC on a 20% test set.
• The result: Transfusion occurred in 2.2% of cases, and adding pseudonymised surgeon identity to a procedure-adjusted model increased AUPRC by +0.032 (95% CI 0.020–0.044), with surgeon identity ranking third among 72 features by permutation importance.

AI summary of the abstract · may contain errors · not medical advice · read the paper.