There has been no severity evaluation model for pediatric patients with hemophagocytic lymphohistiocytosis (HLH) that uses readily available parameters. This study aimed to develop a novel model for predicting the early mortality risk in pediatric patients with HLH using easily obtained parameters whatever etiologic subtype. Patients from one center were divided into training and validation sets for model derivation. The developed model was validated using an independent validation cohort from the second center. The prediction model with nomogram was developed based on logistic regression. The model performance underwent internal and external evaluation and validation using the area under the receiver operating characteristic curve (AUC), calibration curve with 1000 bootstrap resampling, and decision curve analysis (DCA). Model performance was compared with the most prevalent severity evaluation scores, including the PELOD-2, P-MODS, and pSOFA scores. The prediction model included nine variables: glutamic-pyruvic transaminase, albumin, globulin, myohemoglobin, creatine kinase, serum potassium, procalcitonin, serum ferritin, and interval between onset and diagnosis. The AUC of the model for predicting the 28-day mortality was 0.933 and 0.932 in the training and validation sets, respectively. The AUC values of the HScore, PELOD-2, P-MODS and pSOFA were 0.815, 0.745, 0.659 and 0.788, respectively. The DCA of the 28-day mortality prediction exhibited a greater net benefit than the HScore, PELOD-2, P-MODS and pSOFA. Subgroup analyses demonstrated good model performance across HLH subtypes. The novel mortality prediction model in this study can contribute to the rapid assessment of early mortality risk after diagnosis with readily available parameters.
Keywords: Children; Early mortality; Hemophagocytic lymphohistiocytosis; Nomogram; Prediction model.
© 2024. The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature.