A Clinical Prediction Model for Inadequate Hemoglobin Response After Red Blood Cell Transfusion.

BACKGROUND

Transfusion of one unit of packed red blood cells is traditionally expected to increase hemoglobin concentration by approximately 1 g/dL in adults. However, substantial variability in hemoglobin response is frequently observed in clinical practice. Smaller-than-expected increments may raise concerns regarding ongoing bleeding, hemodilution, or altered erythrocyte distribution. Understanding predictors of inadequate hemoglobin response may help clinicians interpret post-transfusion laboratory values and avoid unnecessary repeat transfusion.

METHODS

We conducted a prospective observational cohort study of 518 hemodynamically stable hospitalized adults receiving a single unit of red blood cell transfusion. Hemoglobin concentrations were measured at baseline and at 12, 24, and 48 h after transfusion. Inadequate hemoglobin response was defined a priori as a 24-h hemoglobin increment (ΔHb24) < 1 g/dL. Multivariable logistic regression was used to identify predictors of inadequate response. Model discrimination and calibration were evaluated using receiver operating characteristic analysis and calibration plots. A simplified clinical prediction score and nomogram were derived from the regression model.

RESULTS

The mean 24-h hemoglobin increment was 1.41 ± 0.37 g/dL. Inadequate hemoglobin response occurred in 106 patients (20.5%). Patients with inadequate response had higher baseline hemoglobin levels and a greater prevalence of splenomegaly. In multivariable analysis, higher baseline hemoglobin (OR 1.35 per 1 g/dL increase, 95% CI 1.18-1.54) and splenomegaly (OR 1.70, 95% CI 1.10-2.65) were independently associated with inadequate hemoglobin response. The prediction model demonstrated moderate discrimination (AUC 0.64; 95% CI 0.58-0.70) and acceptable calibration. A simplified clinical score stratified patients into low-, intermediate-, and high-risk categories for inadequate response. Decision curve analysis demonstrated modest clinical utility across a range of threshold probabilities.

CONCLUSIONS

Inadequate hemoglobin response after single-unit red blood cell transfusion occurs in approximately one fifth of patients and appears to be influenced primarily by recipient physiological factors rather than donor unit characteristics. A simple clinical prediction model may assist clinicians in interpreting post-transfusion hemoglobin changes and reduce unnecessary repeat transfusions.