Abstract
Lower respiratory tract infections are common in children. Bronchoalveolar lavage fluid has long been established as the best biological sample for detecting respiratory tract infections; however, it is not easily collected in children. Sputum may be used as an alternative yet its diagnostic accuracy remains controversial. Therefore, this study sought to evaluate the diagnostic accuracy of sputum for detecting lower respiratory tract infections using metagenomic next-generation sequencing. Paired sputum and bronchoalveolar lavage fluid samples were obtained from 68 patients; pathogens were detected in 67 sputum samples and 64 bronchoalveolar lavage fluid samples by metagenomic next-generation sequencing, respectively. The combined pathogen-detection rates in the sputum and bronchoalveolar lavage fluid samples were 80.90% and 66.2%, respectively. For sputum, the positive predictive values (PPVs) and negative predictive values (NPVs) for detecting bacteria were 0.72 and 0.73, respectively, with poor Kappa agreement (0.30; 95% confidence interval: 0.218-0.578, P < 0.001). However, viral detection in sputum had good sensitivity (0.87), fair specificity (0.57), and moderate Kappa agreement (0.46; 95% confidence interval: 0.231-0.693, P < 0.001). The PPVs and NPVs for viral detection in sputum were 0.82 and 0.67, respectively. The consistency between the sputum and bronchoalveolar lavage fluid was poor for bacterial detection yet moderate for viral detection. Thus, clinicians should be cautious when interpreting the results of sputum in suspected cases of lower respiratory tract infections, particularly with regards to bacterial detection in sputum. Viral detection in sputum appears to be more reliable; however, clinicians must still use comprehensive clinical judgment.