Abstract
Platelet transfusion is commonly utilized in the surgical intensive care unit (ICU) to provide hemostatic support for patients with severe thrombocytopenia or bleeding. Ineffective platelet transfusion is described as either the expected platelet count increase or insufficient hemostasis following transfusion. This study sought to identify risk factors and assess their predictive value in improving patient outcomes. This was an observational, retrospective study. Clinical information was gathered from 302 platelet transfusion episodes that occurred in the surgical ICU at the First Affiliated Hospital of Sun Yat-sen University, between June and December 2022. Based on the clinical bleeding improvement indices (drainage volume, hemoglobin level) and the 24 hour-corrected count increment, the patients were divided into effective and ineffective groups. Binomial logistic regression was used to identify independent risk factors for ineffective transfusion after univariate analysis was screened for significant variables (P < .05). The predictive performance of these factors was assessed using receiver operating characteristic (ROC) curves. Patients were separated into 2 groups: effective (n = 192) and ineffective (n = 110). Baseline comparisons demonstrated significant differences in surgery frequency (≤1 vs >1), liver disease, intraoperative bleeding (≤1600 vs >1600), procalcitonin (PCT) level, C-reactive protein level, splenomegaly, and age (≥65 vs <65 years) (P < .05). The total red blood cell transfusion volume and other indicators showed no intergroup differences (P > .05). Univariate logistic regression revealed a significant association (P < .05) between platelet transfusion refractoriness and these variables, including the total red blood cell transfusion volume. Multivariate logistic regression analysis identified surgeries (>1) (odds ratio (OR) = 1.997, 95% confidence interval (CI): 1.196-3.335; P = .008), splenomegaly (OR = 3.912, 95% CI: 2.277-6.720; P < .001), and PCT (OR = 1.030, 95% CI: 1.013-1.047; P < .001) as independent risk variables (P < .05). ROC curve analysis revealed that the area under the curve of the model was 0.706 (95% CI: 0.645-0.766), with a Youden index of 0.346, corresponding to an optimal cutoff value of 0.338, sensitivity of 66.40%, and specificity of 68.20%. The number of surgeries (>1), splenomegaly, and PCT level were independent risk factors for ineffective platelet transfusion, and the constructed model had moderate predictive efficacy, providing a reference for early clinical identification and intervention.