Clinical factors and relative risk stratification for repeat platelet transfusion in thrombocytopenia

To identify clinical factors associated with repeat platelet (PLT) transfusion among thrombocytopenic patients and develop an exploratory relative-risk stratification framework for repeat transfusion during hospitalization.

This retrospective case-control study included thrombocytopenic inpatients who received at least one PLT transfusion. Patients with repeat transfusion (≥2 transfusions during the same hospitalization, n=145) were compared with those receiving a single transfusion (n = 145). Multivariable logistic regression using clinically prespecified predictors was performed, with ten-fold cross-validated LASSO used to assess variable stability. Model performance was evaluated by AUC, calibration measures, Brier score, and 1,000-resample bootstrap optimism correction. A simplified integer-based score was developed.

Multivariable analysis identified hematologic malignancy, active bleeding, splenomegaly, lower PLT and hemoglobin levels, absence of thrombopoietic-agent use before the first transfusion, and transfusion timing as factors associated with repeat PLT transfusion. The primary model demonstrated an apparent AUC of 0.805 (95% CI, 0.756-0.855), with a bootstrap-corrected C-index of 0.777. The corrected Brier score was 0.196, and the bootstrap-corrected calibration intercept and slope were 0.001 and 0.860, respectively. The simplified score (0-12 points) stratified patients into low-, medium-, and high-score groups, with observed repeat transfusion proportions of 20.0%, 50.7%, and 83.8%, respectively (trend χ² = 59.904, P < 0.001).

Several clinical factors were associated with repeat PLT transfusion. The proposed score may support exploratory relative-risk stratification within the development sample; external validation is required before estimating absolute risk or clinical implementation.