Article Text

Download PDFPDF

Review: C-reactive protein has moderate diagnostic accuracy for serious bacterial infection in children with fever

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

S Sanders

Ms S Sanders, University of Queensland, Brisbane, Queensland, Australia; s.sanders{at}


What is the diagnostic accuracy of C-reactive protein (CRP) for bacterial infections in children with fever?


Included studies compared blood or serum CRP with a reference standard for microbiological diagnosis of bacterial v non-bacterial infection or serious bacterial v benign bacterial or non-bacterial infection in children initially presenting with fever. Excluded were studies involving >10% neonates, children admitted to hospital outside of the emergency department (ED) or with a specific medical condition (eg, cancer or renal failure), or in which the reference standard was diagnosis of a specific disease (eg, meningitis, gastroenteritis, or arthritis). Outcome measures were sensitivity, specificity, positive and negative likelihood ratios, and diagnostic odds ratio.


Medline and EMBASE/Excerpta Medica (both to Dec 2007), and reference lists were searched for studies. Authors were contacted for additional information. 10 studies evaluating CRP met the selection criteria: 7 (n = 1322) for detecting serious bacterial infections with reference standards including blood or urine cultures, chest radiography, or lumbar punctures; and 3 (n = 722) for differentiating between bacterial and non-bacterial or viral infections. The CRP cut-point varied for diagnosing serious bacterial infection (20–70 mg/l) and bacterial infection (>6–35 mg/l). All studies were conducted in EDs. Assessment of study quality was based on the QUality Assessment of studies of Diagnostic Accuracy in Systematic reviews (QUADAS) scale; score range 3–8 of 10 criteria.


Serious bacterial infection. Prevalence of serious bacterial infection was 11–29% across 6 studies. The table shows pooled data for diagnostic accuracy of CRP for serious infections. Bacterial infection. Results of 3 studies that differentiated bacterial from non-bacterial infection were not pooled because of study variation. Prevalence of bacterial infection varied (range 28–82%), sensitivity was low (range 22–58%), and specificity was high (range 86–96%).


In children with fever, C-reactive protein has moderate sensitivity and specificity for diagnosing serious bacterial infection and low sensitivity with high specificity for diagnosing bacterial infections.


Sanders S, Barnett A, Correa-Velez I, et al. Systematic review of the diagnostic accuracy of C-reactive protein to detect bacterial infection in nonhospitalized infants and children with fever. J Pediatr 2008;153:570–4.

Pooled data for C-reactive protein for differentiating serious bacterial infection from benign or non-bacterial infection in children with fever*

Clinical impact ratings: GP/FP/Primary care 6/7; Paediatrics 6/7; Emergency medicine 5/7; Paediatric emergency medicine 5/7


The review by Sanders et al examined the utility of CRP for predicting bacterial infection in febrile children, most of whom were <3 years of age. It is important because pooling data from various studies can provide better estimates of test diagnostic accuracy. The review included studies with different inclusion criteria, which is of concern. For example, Isaacman and Burke1 included all febrile children, whereas Pulliam et al2 included only those without a clinically identifiable source for fever. In addition, the temperature cut-off for fever varied or was undefined in different studies, and data were obtained from ED patients and may not be generalisable to those who present at primary care practices. The inclusion of children presenting within hours of onset of fever also limits the positive predictive value and positive likelihood ratio of CRP.3 Nonetheless, the findings highlight that CRP has limited utility in diagnosing serious bacterial infection. This is partly because some viral infections, such as adenovirus, are also associated with elevated CRP.4

In an era of pan-vaccination, the risk of bacterial infection (bacteraemia) in otherwise healthy, immunised, young children is extremely low (<1%).5 Investigating fever in such patients or those with an obvious viral source is not indicated. Urine analysis should be considered in Caucasian girls and uncircumcised boys. Otherwise, work-ups should be guided by the clinical picture and, if CRP is obtained, it should be interpreted accordingly.


View Abstract


  • Source of funding: University of Queensland New Staff Start-up Grant.