Session: 311. Disorders of Platelet Number or Function: Clinical and Epidemiological: Poster III
Hematology Disease Topics & Pathways:
Research, Bleeding and Clotting, Clinical Research, Platelet disorders, Supportive Care, Diseases, Real-world evidence, Treatment Considerations, Technology and Procedures, Study Population, Human, Machine learning
Methods: This study used a nationally representative database to develop a model to predict the risk of post-procedure bleeding in patients with ITP. Machine learning analyses, including random forest feature importance and Shapley additive explanations (SHAP) values, were employed. This contained a total of 34 post-procedural bleeding risk factors, including the amount of platelet transfusion.
Results: The random forest model had an area under the receiver-operating characteristic curve of 93.6%. The analysis identified the following variables as most important: amount of platelet transfusion, high-risk procedure, use of anticoagulant drugs, no use of antiplatelet or anticoagulant drugs, anemia, age, low-risk procedure, moderate-risk procedure, ITP treatment, and newly diagnosed ITP. Amount of platelet transfusion, high-risk procedures, use of anticoagulant drugs, anemia, ITP treatment, and newly diagnosed ITP positively correlated with post-procedure bleeding risk. In contrast, no use of antiplatelet or anticoagulant drugs and moderate- or low-risk procedures were negatively associated with post-procedure bleeding risk. In the SHAP dependence plot, the amount of platelet transfusion was associated with high-risk procedures. Additionally, among patients undergoing high-risk procedures, the likelihood of post-procedure bleeding increased with age.
Conclusions: Platelet transfusion does not significantly reduce the risk of post-procedural bleeding in patients with ITP. The risk of post-procedural bleeding is more closely related to the bleeding risk of the procedure and the patient's medical condition. Minimizing inappropriate platelet transfusions and addressing factors that can increase bleeding risk before procedures are crucial.
Keywords: Immune thrombocytopenia, Procedure, Platelet transfusion, Bleeding risk, Machine learning analysis
Disclosures: No relevant conflicts of interest to declare.