Predictors of Airspace Disease on Chest X‐ray in Emergency Department Patients With Clinical Bronchiolitis: A Systematic Review and Meta‐analysis

Background An abnormal chest X-ray (CXR) inconsistent with simple bronchiolitis is found in 7%–23% of cases. Despite national guidelines stating “current evidence does not support routine radiography in children with bronchiolitis”; the use of CXR in these patients remains high. Inappropriate use of CXR not only exposes children to excess radiation, but also increases medical costs. The majority of the time, CXRs are obtained to diagnose or rule out pneumonia. We aim to provide an evidence-based approach defining the utility of CXR in bronchiolitis for the diagnosis and treatment of bacterial pneumonia. Objectives We performed a systematic review and meta-analysis to describe potential predictors of a CXR with airspace disease in patients with bronchiolitis. Methods We searched the medical literature from 1965 to June 2015 in PubMed/EMBASE using the following PICO formulation of our clinical question, “What characteristic(s) of history/physical examination (HP Intervention—HP Comparator—a CXR positive for airspace disease (+CXR), defined as atelectasis versus infiltrate or infiltrate/consolidation; and Outcome—operating characteristics of H&P and VS predicting an +CXR were calculated: sensitivity, specificity, and likelihood ratios (LR+ or LR−). The methodologic quality of the studies was assessed using the quality assessment of studies of diagnostic accuracy tool (QUADAS-2). We created a test–treatment threshold model based on the operating characteristics of the CXR to accurately identify a child with bronchiolitis and a superimposed bacterial pneumonia while accounting for the risks of a CXR and risks of treating patients with and without a bacterial infection. Results We found five studies including 1,139 patients meeting our inclusion/exclusion criteria. Prevalence of a +CXR ranged from 7% to 23%. An oxygen saturation < 95% was the predictor with highest LR+ of 2.3 (95% confidence interval = 1.3 to 3.07) to predict a +CXR. None of the H&P and VS variables were found to have sufficiently low LR– to significantly decrease the pretest probability of finding a +CXR. Our test–treatment threshold model showed that hypoxia (O2 Sat < 95%) alone complicating bronchiolitis did not show a benefit to obtaining a CXR. Our model only suggested that a CXR maybe indicated for a child with hypoxia (O2 Sat < 95%) and respiratory failure requiring ventilatory support. Conclusion No single predictor of a +CXR was of sufficient accuracy to either support or refute ordering a CXR in a child with clinical bronchiolitis. We provide a decision threshold model to estimate a test threshold for obtaining a CXR and a treatment threshold for administering antibiotics. Application of this model requires the clinician to approximate the empiric benefit of antibiotics based on the clinical situation, highlighting the importance of clinical assessment.