Comparing accuracy of clinical prediction rules to predict pneumonia in children and adolescents with acute febrile respiratory illness

It is a common challenge for emergency physicians to differentiate pneumonia from simple upper respiratory tract infections. Several clinical prediction rules exist to assist the diagnosis process and guide the clinical decisions of ordering investigations such as chest X‐ray (CXR).This study aims to validate and compare the accuracy of various prediction rules in the setting of children and adolescents presenting with acute febrile respiratory illness (AFRI).This was a prospective multicentre study. Three hundred and fifty‐five patients, aged 6–18 years, were recruited. Patients with immunocompromised state or hypoxia were excluded. Pneumonia was defined as diagnosis by CXR or subsequent diagnosis of pneumonia upon re‐attendance within 7 days. Clinical rules including Diehr rule, Heckerling rule, Bilkis simpler rule, the AFRI rule, the paediatric acute febrile respiratory illness rule (PAFRI) were compared in terms of accuracy of predicting pneumonia in the recruited subjects and presented as receiver operating characteristic curves.Five patients were excluded. In the 350 patients included, 38 were diagnosed as pneumonia by CXR and 1 was subsequently diagnosed as pneumonia upon re‐attendance. The area under the receiver operating characteristic curve of Diehr rule, Heckerling rule, Bilkis simpler rule, AFRI rule and PAFRI rule were 0.703, 0.565, 0.59, 0.807 and 0.846 respectively. The PAFRI rule is superior to other prediction rules in terms of diagnostic accuracy. At the cut‐off of PAFRI ≥0, the rule has high sensitivity of 97.44% and negative predictive value of 99.09%.Among the rules compared, the PAFRI rule has the highest diagnostic accuracy in assisting emergency physicians to identify pneumonia among children and adolescents aged 6–18 years presenting with AFRI.