Pathogen spectrum and triple-criteria interpretation algorithm of BALF-targeted NGS in children with refractory severe community-acquired pneumonia in mountainous Northeastern Sichuan: a single-center retrospective study

Background Severe community-acquired pneumonia (SCAP) causes substantial pediatric hospitalization and respiratory failure in mountainous western China. Conventional microbial assays show limited sensitivity for fastidious pathogens and mixed infections. Targeted next-generation sequencing (tNGS) using bronchoalveolar lavage fluid (BALF) enables broad-spectrum pathogen detection at lower cost than whole-metagenomic sequencing. However, many previous tNGS studies directly equate multiple microbial nucleic-acid signals with true coinfection, which may lead to unnecessary combined anti-infective therapy. Methods This single-center retrospective study enrolled 86 children aged < 14 years with refractory SCAP who underwent BALF-tNGS testing at Guangyuan Central Hospital between March 2023 and September 2024. We developed a triple-criteria interpretive algorithm combining an exploratory empirical pathogen-load cut-off (>1.0 × 10 6 copies/mL), inflammatory biomarkers (CRP, PCT), and chest imaging to discriminate genuine active coinfection from airway colonization. This load threshold was empirical and not internally validated by ROC analysis in our dataset. Patients were stratified into four age subgroups; diagnostic concordance was assessed among 34 patients with parallel routine microbial tests. Results In total, 229 pathogen isolates were identified. Viruses predominated (58.9%, 135/229), followed by bacteria (32.8%, 75/229). The three most frequent pathogens were Haemophilus influenzae (32.56%), Streptococcus pneumoniae (30.23%), and respiratory syncytial virus A (RSV-A, 18.60%). While 82.6% (71/86) of patients harbored ≥2 pathogen nucleic-acid signals, only 17.4% (15/86) fulfilled criteria for clinical coinfection, dominated by virus-bacterial dual infection. Viral detection rates tended to decline with increasing age, whereas bacterial positivity was higher among preschool- and school-age children. A prominent spring epidemic peak was observed during May-June 2023. Only 22.86% of paired routine microbial tests showed full concordance with tNGS. These single-center data cannot be generalized to all pediatric severe-pneumonia populations. Most abundant multi-pathogen tNGS signals represented colonization or residual dead-microbe nucleic acid instead of true active infection. Conclusion Our triple-criteria interpretive framework offers standardized rules for grassroots mountain hospitals to reduce over-prescription of combined antibiotics and antivirals when interpreting BALF-tNGS results for children with refractory SCAP.