Construction and evaluation of a nomogram prediction model for the efficacy of apheresis platelet transfusion / 中国输血杂志

ABSTRACT

【Objective】 To guide the scientific and rational platelet transfusion in clinical and reduce refractory platelet transfusion by constructing a nomogram prediction model for the efficacy of platelet transfusion. 【Methods】 The basic information of 298 patients underwent prophylactic transfusion of apheresis platelet in our hospital from 2019 to 2020 was collected, including gender, age, blood group, number of blood transfusions, hours before expiration, platelet count before and after transfusion, body temperature after transfusion, and splenomegaly. Potential risk factors causing refractory platelet transfusion were screened out by univariate logistic regression analysis, and the independent risk factors were further evaluated by multivariate binary logistic regression. The independent risk factors were then imported into Rstudio software to fit the effect prediction nomogram mode, which was verified by bootstrap repeated sampling,, evaluated by ROC, and calibrated by Calibration. 【Results】 212 out of 298 apheresis platelet transfusions were effective, with the effective rate at 71.14%. The binary logistic regression analysis showed that the non-immune factors causing refractory platelet transfusion were gender, blood transfusions≤ 4 occasions, blood transfusion more than 4 occasions, body temperature, splenomegaly, and platelet count before transfusion; the OR value and 95% CI were 3.773 (1.796~7.924), 29.081 (8.668~97.568), 10.416 (1.525~71.134), 7.376 (3.145~17.301), 17.385 (6.428~47.018), and 0.947 (0.904~0.992), respectively. Regression model Hosmer-Lemeshaw test P value was 0.805. The AUC of fitted nomogram prediction model was 0.895, 95% CI (0.855~0.936). 【Conclusion】 The independent risk factors causing refractory platelet transfusion are gender, body temperature, number of blood transfusions, splenomegaly and platelet count before transfusion. The area under the curve of the fitted nomogram prediction model is 0.899, which has a good predictive ability and presents the probability of ineffective transfusion visually and quantitatively, so as to guide scientific and rational use of platelets, and reduce refractory platelet transfusion.