Machine learning based development of an early diagnosis signature for distinguishing hospitalized pediatric human respiratory syncytial virus infection from mycoplasma pneumonia.

BACKGROUND

The differentiation between human respiratory syncytial virus (HRSV) and mycoplasma pneumoniae (MP) infections in pediatric community-acquired pneumonia (CAP) remains a clinical challenge due to overlapping respiratory symptoms. A rapid, non-invasive diagnostic tool is urgently needed to guide appropriate therapeutic decisions and antimicrobial stewardship.

OBJECTIVE

This study aimed to develop and validate a blood-based biomarker signature for distinguishing HRSV from MP infections.

METHODS

We conducted a retrospective cohort study analyzing clinical data and blood samples from patients with CAP infected by HRSV and MP, diagnosed via PCR and serology. Patients were randomly split into a discovery cohort and a validation cohort. Least Absolute Shrinkage and Selection Operator (LASSO) regression was used for feature selection. ML was employed to train, optimize, and evaluate multiple classifiers, including logistic regression, random forest, and support vector machines. The diagnostic performance of the final model was evaluated by the area under the receiver operating characteristic curve (AUC), sensitivity, and specificity.

RESULTS

The analysis identified a parsimonious five-biomarker signature comprising eosinophilic granulocyte, immunoglobulin A, lactic dehydrogenase (LDH), β2-microglobulin, and the albumin to globulinratio (AGR). The optimized random forest model demonstrated superior performance, achieving an AUC-ROC of 0.89 (95% CI: 0.85-0.90) for distinguishing HRSV from MP.

CONCLUSION

We developed and validated a novel, minimally invasive blood biomarker signature that accurately distinguishes HRSV from MP infections. This model has the potential to serve as a valuable adjunctive tool for early etiological diagnosis, facilitating timely and targeted clinical management. Further prospective, multi-center studies are warranted to confirm its generalizability and clinical utility.