Machine learning identifies EBV-associated HLH from routine labs

Two independent studies published June 27, 2026, developed machine learning classifiers to distinguish life-threatening EBV-associated hemophagocytic lymphohistiocytosis (EBV-HLH) from self-limited Epstein-Barr virus infectious mononucleosis in children. The XGBoost model achieved an AUC of 0.9775, sensitivity of 0.9461, and specificity of 0.9784, with top predictors including D-dimer, cervical lymphadenopathy, GGT, LDH, and CD3+CD4+ T cells. The findings could improve early triage and feature selection in clinical ML workflows.

Editorial analysis: Rapid differentiation between self-limited Epstein-Barr virus infectious mononucleosis and life-threatening EBV-associated hemophagocytic lymphohistiocytosis EBV-HLH reshapes priorities for early triage models and feature selection in clinical ML workflows. Reported facts: Two independent retrospective studies published 27 June 2026 developed and validated ML classifiers for pediatric EBV-HLH. BMC Medical Informatics and Decision Making Yingying Ye et al. reports a XGBoost model trained on 1,026 hospitalized children that achieved AUC 0.9775 , sensitivity 0.9461 , and specificity 0.9784 , with SHAP identifying D-dimer , cervical lymphadenopathy, GGT , LDH , and CD3+CD4+ T cells as top predictors. BMC Infectious Diseases Li Xiao et al. reports an external-validation cohort of 4,871 patients, EBV-HLH prevalence 12.46% , evaluation of 13 algorithms, and SHAP-based interpretation using routine CBC within 24 hours of admission.