Background and aims Diagnosis of serious bacterial illness (SBI) in children can be challenging despite it being endemic. Few studies have looked at clinical features and investigations in diagnosis of SBI. We compared tachycardia with C-reactive protein (CRP) and WCC to establish which is better for predicting SBI.

Methods Prospective cohort study of children <6 yrs referred to an acute assessment unit with febrile illness (July 2012–March 2014). Four variables: Tachycardia (APLS guideline) at admission, tachycardia during stay, CRP and WCC were recorded. ROC curves were plotted for WCC and CRP.

Results 219 children participated. 100 had confirmed SBI. 101 had CRP and WCC checked as decided by admitting clinician. There were significant differences for all four variables (Chi-square test). Tachycardia at presentation (p = 0.021), Tachycardia later in illness (<0.001), CRP > 5 (p = 0.001) and WCC (<5/ >15)(p = 0.031).

Conclusions In keeping with our previous study tachycardia at presentation did not reliably predict SBI with or without focus, however its negative predictive value was 65%, which increased to 77% for tachycardia later in illness. The optimal cut-off for CRP from ROC curve that maximised sensitivity was 19 (sensitivity-94.1%, specificity-45%). WCC was neither specific nor sensitive. In summary neither the tachycardia nor the WCC and CRP were reliable in predicting SBI.

• Download figure
• Open in new tab
• Download powerpoint

Abstract PS-366 Figure 1

ROC curve of CRP