Preoperative Hemoglobin Discriminates Perioperative Transfusion in Reverse Shoulder Arthroplasty: A Single-Center Preliminary Study with Internal Validation and a Candidate Blood-Preparation Threshold

Background and Objectives: Perioperative transfusion remains a burden after reverse shoulder arthroplasty (RSA). Yet no RSA-specific tool with internally validated discrimination guides preoperative blood preparation. This single-center preliminary study pursued three aims: to derive and internally validate the discrimination of preoperative hemoglobin for transfusion in a protocol-uniform RSA cohort, to characterize its net benefit by decision-curve analysis, and to translate that discrimination into a candidate blood-preparation threshold. Materials and Methods: This retrospective study analyzed 214 consecutive elective degenerative primary RSA procedures by a single surgeon. Each was managed with maximal intravenous iron and without routine tranexamic acid. The primary outcome was perioperative transfusion. The analysis fitted Firth penalized models to 213 procedures that included 17 transfusion events. The area under the curve (AUC) quantified preoperative-hemoglobin discrimination, corrected for optimism by bootstrapping, and decision-curve analysis quantified net benefit. Results: Transfusion occurred in 7.9% of procedures. Lower preoperative hemoglobin was the prespecified primary predictor, with an adjusted odds ratio of 0.37 per +1 g/dL. As a univariable continuous predictor, it discriminated transfusion with an AUC of 0.867. Internal bootstrap validation showed negligible optimism. Decision-curve analysis showed in-sample net benefit, though only a narrow margin over crossmatch-for-all at the lowest thresholds. At the in-sample Youden cutoff, a hemoglobin above 11.4 g/dL identified about two-thirds of patients. Among them, 1.4% were transfused, for a negative predictive value of 0.99. Conclusions: Preoperative hemoglobin showed internally validated discrimination for transfusion. The 11.4 g/dL threshold and its net benefit are apparent (in-sample) and remain fragile given 17 events. Together they define a candidate RSA blood-preparation triage. Because the protocol withheld tranexamic acid, the threshold does not generalize to tranexamic-acid-based practice. These are preliminary, single-center findings. External multicenter validation in larger cohorts is required before adoption.