Age-stratified diagnostic performance of myxovirus resistance protein A for differentiating viral and bacterial respiratory tract infections in children: a diagnostic accuracy study

Differentiating viral from bacterial respiratory tract infections (RTIs) in children remains a major clinical challenge and contributes to inappropriate antibiotic use. Myxovirus resistance protein A (MxA), an interferon-inducible host-response biomarker, has shown promise for identifying virus-driven infections. However, age-related variation in MxA expression may affect its clinical applicability in pediatric populations. In this retrospective diagnostic accuracy study, a total of 669 participants were included, comprising 414 children with RTIs and 255 healthy controls. Leftover EDTA whole-blood samples from routine clinical care were retrieved for MxA measurement. Levels of MxA and four conventional inflammatory biomarkers—C-reactive protein (CRP), procalcitonin (PCT), interleukin-6 (IL-6), and serum amyloid A (SAA)—were measured. Diagnostic performance was evaluated using receiver operating characteristic (ROC) analysis. Age-stratified analyses were conducted to account for physiological variation in MxA expression. Multivariable logistic regression models were constructed to evaluate combined diagnostic strategies. The MxA, CRP and age group model underwent internal validation using 2,000 bootstrap resamples, with assessments of discrimination, Brier score, and calibration. In healthy controls, MxA concentrations declined with age. MxA concentrations were significantly higher in viral infections compared with bacterial infections ( P < 0.0001) but did not differ between isolated viral infections and viral-bacterial coinfections. MxA demonstrated the highest diagnostic accuracy among individual biomarkers (AUC, 0.860; sensitivity, 73.12%; specificity, 90.38%), outperforming CRP (AUC, 0.702), SAA (AUC, 0.633), PCT (AUC, 0.610), and IL-6 (AUC, 0.546). Age-stratified cutoffs varied substantially among the age groups (107.48, 82.34. and 42.33 ng/mL), and specificity was not uniformly improved relative to the overall threshold. The model incorporating MxA, CRP, and age group had an apparent AUC of 0.931 and an optimism-corrected AUC of 0.924. The optimism-corrected Brier score, calibration intercept, and calibration slope were 0.0835, -0.0099, and 0.9041, respectively. MxA showed age-related variation and discriminated isolated viral from bacterial infections in this cohort. The model combining MxA, CRP, and age group retained discrimination after bootstrap correction, although some overfitting was observed. Prospective external validation and assessment of clinical utility are needed before clinical implementation.