Infection is a common complication of leukemia patients undergoing chemotherapy. Blood culture results are often needed to guide clinical use, but repeated sampling is often necessary to improve the positive rate and eliminate contamination. The purpose of this paper is to find predictive factors of blood culture results among clinical and laboratory indicators and try to establish a prediction model, so as to better choose the time of blood culture examination, predict the results, and better guide clinical treatment. We retrospectively collected clinical and laboratory data of febrile acute leukemia patients undergoing chemotherapy. The samples were randomly assigned to the training set and the validation set, and the prediction model was constructed from the training set. The calibration curve was made in the validation set and the Hosmer-Lemeshow test was performed to evaluate the prediction performance of the prediction model. A total of 229 patients were included. Univariate and multivariate analyses suggested that temperature at fever and procalcitonin were variables of significant difference between positive and negative blood culture patients. The sensitivity of the 2 variables for predicting blood culture results was high, but the specificity was low. In the process of external validation, the predictive ability of the constructed prediction model to the blood culture results was low. This study identified clinical and laboratory parameters associated with blood culture outcomes, but the predictive model established has low predictive accuracy in external validation.
Copyright © 2023 the Author(s). Published by Wolters Kluwer Health, Inc.