Can clinical signs or symptoms combined with basic hematology data be used to predict the presence of bacterial infections in febrile children under - 5 years?

Background: Infectious diseases in children living in resource-limited settings are often presumptively managed on the basis of clinical signs and symptoms. Malaria is an exception. However, the interpretation of clinical signs and symptoms in relation to bacterial infections is often challenging, which may lead to an over prescription of antibiotics when a malaria infection is excluded. The present study aims to determine the association between clinical signs and symptoms and basic hematology data, with laboratory confirmed bacterial infections.

Methods: A health survey was done by study nurses to collect clinical signs/symptoms in febrile (axillary temperature ≥ 37.5 °C) children under - 5 years of age. In addition, blood, stool and urine specimen were systematically collected from each child to perform bacterial culture and full blood cell counts. To determine the association between a bacterial infection with clinical signs/symptoms, and if possible supported by basic hematology data (hemoglobin and leucocyte rates), a univariate analysis was done. This was followed by a multivariate analysis only on those variables with a p-value p < 0.1 in the univariate analysis. Only a p-value of < 0.05 was considered as significant for multivariate analysis.

Results: In total, 1099 febrile children were included. Bacteria were isolated from clinical specimens (blood-, stool- and urine- culture) of 127 (11.6%) febrile children. Multivariate logistical regression analysis revealed that a general bacterial infection (irrespective of the site of infection) was significantly associated with the following clinical signs/symptoms: diarrhea (p = 0.003), edema (p = 0.010) and convulsion (p = 0.021). Bacterial bloodstream infection was significantly associated with fever> 39.5 °C (p = 0.002), diarrhea (p = 0.019) and edema (p = 0.017). There was no association found between bacterial infections and basic haematological findings. If diarrhea and edema were absent, a good negative predictive value (100%) of a bacterial bloodstream infection was found, but the positive predictive value was low (33.3%) and the confidence interval was very large (2.5-100; 7.5-70.1).

Conclusion: Our study demonstrates that clinical signs and symptoms, combined with basic hematology data only, cannot predict bacterial infections in febrile children under - 5 years of age. The development of practical and easy deployable diagnostic tools to diagnose bacterial infections remains a priority.