Pneumonia is the greatest killer of children worldwide,1 with 920 136 deaths in children in 2015.2 Although less lethal in developing countries like the United States, the burden is still considerable, with annual ambulatory visit rates for community-acquired pneumonia (CAP) in US children of 16.9 to 22.4 per 1000 in the population.3
Although CAP is a clinical diagnosis, individual signs and symptoms have poor prognostic value,4 and children may undergo radiography to confirm or disprove the diagnosis. Although both viruses and bacteria can cause CAP, physicians are generally trying to rule out a bacterial infection with a chest radiograph (CXR) because the main clinical decision is whether to prescribe antibiotics.
Some physicians may be concerned that CXRs can have false-negative results in the early stages of CAP or in the setting of dehydration.5 To address these concerns, Lipsett et al6 report the results of a prospective observational cohort study in children 3 months to 18 years of age undergoing CXR for suspected CAP. The primary outcome was a clinical diagnosis of pneumonia during the 2-week follow-up period for children discharged from the emergency department (ED) with a negative CXR result who had not received antibiotics. Of the 411 children meeting these criteria, 5 (1.2%) were diagnosed with pneumonia during the follow-up period (although even those diagnoses are questionable), yielding a negative predictive value (NPV) of 98.8% (95% confidence interval 97.0%–99.6%).
Physicians should take comfort in the fact that CXR for CAP has a high NPV. However, another possibility is that some or most of the children in this study who were diagnosed with pneumonia in the ED did not have bacterial pneumonia. In short, how accurate is the CXR when ruling in bacterial pneumonia, or what is the positive predictive value of a CXR when diagnosing bacterial CAP? We would argue that the answer to this question is at least equally as important as the one addressed by the authors.
There are several reasons to suspect that the positive predictive value of a CXR may not be as high as the NPV demonstrated in this study. Three-quarters of the children in the current study were <6 years of age, which matches the age range of the patients in a recent large etiological study of pediatric CAP in the United States, in which a pathogen was found in 81% of children.7 Researchers in that study found that of 2222 children hospitalized with radiographic evidence of pneumonia, bacteria were only found in 15%, and almost half of those had codetection of a virus. The breakdown of bacterial etiologies was Mycoplasma pneumoniae (8%), Streptococcus pneumonia (4%), Staphylococcus aureus (1%), and Streptococcus pyogenes (1%). The utility of antibiotic treatment of CAP caused by M pneumoniae in children has been questioned.8,9 These findings reveal the possibility that only a small fraction of pediatric patients with CAP benefit from antibiotic therapy and, by extension, radiography. In fact, a single-center study in the inpatient setting revealed a high rate of discontinuation of antibiotics for children who are diagnosed with CAP by the ED and admitted to the hospital (62%),without any noticeable effect seen in outcomes, such as readmissions or transfer to higher level of care.10 Presumably the burden of bacterial disease is even less in ambulatory children with CAP.
So, does performing a CXR improve outcomes for children with suspected bacterial CAP? The only randomized controlled study of CXR in the management of suspected CAP in children was performed in 522 ambulatory children 2 months to 5 years of age in Cape Town, South Africa.11,12 Antibiotic use was higher in the radiograph group (60.8% vs 52.2%; P = .05), with no difference seen in the primary outcome of time to recovery or any of the subsidiary outcomes.
In their study, Lipsett et al6 focus on the possibility (regardless of how slight) of missing the diagnosis and making an error of omission (not treating bacterial pneumonia when it is present). Less attention is paid to errors of commission (eg, treating a child with antibiotics for pneumonia when the actual diagnosis is viral bronchiolitis). This latter error occurs not by failing to obtain a CXR but rather by obtaining one when it is not indicated. Both errors may have large or small consequences. These consequences will differ in each situation, but in medicine, there is an inherent bias to put more weight on errors of omission rather than commission regardless of the consequences.13
The authors of the current study conclude that, “A negative CXR result excludes pneumonia in the majority of children. Children with negative CXR results and low clinical suspicion for pneumonia can be safely observed without antibiotic therapy.”6 Perhaps it is also time to say, “Children with low clinical suspicion for pneumonia can safely be observed without CXRs.” This path would decrease radiation, cost, inconvenience, and unnecessary antibiotic prescriptions.
negative predictive value
Opinions expressed in these commentaries are those of the authors and not necessarily those of the American Academy of Pediatrics or its Committees.
FUNDING: No external funding.
COMPANION PAPER: A companion to this article can be found online at www.pediatrics.org/cgi/doi/10.1542/peds.2018-0236.
POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.