Microbial Etiology Of Acute Febrile Illness In Children Presenting To Hospitals In Ghana

ABSTRACT

Background

Acute febrile illness (AFI) is responsible for a significant number of childhood mortality

and morbidity and remains a common clinical presentation at most hospitals. Lack of

appropriate screening techniques present a challenge in identifying potential pathogens

associated with AFI. This study investigated the microbial etiology of AFI among febrile

children, evaluated the potential use of inflammatory mediators as biomarkers of fever, and

developed a web-based model to predict the infection status of febrile children.

Methods

858 children aged 1-15 years with acute uncomplicated fever were clinically screened

using point-of-care and advanced diagnostic methods. A panel of laboratory tests

comprising malaria microscopy, malaria Rapid Diagnostic Test (RDT), complete blood

count, blood and urine cultures and polymerase chain reaction (PCR) were employed to

screen samples for parasitic, bacterial, and viral pathogens. Concentrations of serum

cytokines and hematological parameters were respectively measured using a Luminexbased

magnetic bead assay and a fully automated hematology analyzer. The test of

sensitivity and specificity, as well as the area under the curve (AUC) of the receiver

operating characteristic (ROC) of cytokines and hematological parameters, were used as

measures of diagnostic accuracy to predict fever and in the selection of the suitable data

mining technique to model malaria and bacterial infection status of febrile children.

Results

Etiologies of fever were identified in 43.7% (374/858) of children studied. From blood

samples analyzed, 38.6% (331/858) tested positive for the Plasmodium parasite which was

the most frequent pathogen detected. From 140 blood and 137 urine cultures performed, 59

organisms were identified. The most common organisms isolated were Staphylococcus

aureus (4.7%), Escherichia coli (3.2%), Group D Streptococcus (2.5%), Pseudomonas

aeruginosa (1.8%), Non-typhoidal salmonellae (1.4%), Coagulase negative staphylococci

(1.4%), Citrobacter freundii (1.1%), Enterobacter clocoae (1.1%), Salmonella Typhi

(0.9%), Streptococcus pneumonia (0.7%) and Klebsiella pneumonia (0.4%). Pathogens

detected using TaqMan-based PCR from 166 blood samples included: Dengue virus

(1.2%), Coxiella burnetti (0.6%), Rickettsia (3.0%), HIV (0.6%) and Plasmodium

falciparum (37.9%). Of the enrolled children, 3.2% had Plasmodium-bacteria coinfections:

Plasmodium-Staphylococcus aureus (0.9%), Plasmodium-dengue (0.3%), and

Plasmodium-Rickettsia (0.6%). From the cytokine analysis, tumor necrosis factor (TNF-α)

with sensitivity of 84.4% (95% CI: 75.5-91.0) and specificity of 72.2% (95% CI: 46.5-

90.3) was the best predictor of fever, having AUC for the ROC curve to be 0.7.

Lymphocyte (LYM-%) was the best hematological predictor of fever, with sensitivity of

65.2%, specificity of 67.3% and a ROC of 0.78. Naïve Bayes model, which incorporated

only clinical symptoms, proved useful for the development of the interactive tool to predict

infection status of children with AFI.

Conclusion

Malaria remains a major contributor of AFI in the study area, despite additional diagnoses

of bacterial and viral origin. Dengue virus, Rickettsia felis and Coxiella burnetti were

detected among the children but not clinically diagnosed. Febrile illnesses due to comorbid

infection are common and call for differential diagnosis of AFI to ensure judicious

use of drugs and to avoid evolution of drug resistance. In addition, routine hematological

parameters including lymphocyte, and circulating cytokines such as TNF-α are useful and

independent prognostic factors for fever. A web-based clinical decision tool has been

developed to predict infection status of febrile patients.