Comparison of clinical tools for dengue diagnosis in a pediatric population-based cohort

Abstract

Background: We aimed to estimate and compare the ability of clinical tools for dengue diagnosis in a pediatric population.

Methods: We prospectively evaluated episodes of acute febrile syndrome identified during the follow-up of a population-based cohort of children and adolescents residing in a dengue endemic city. We estimated the area under the receiver operating characteristic curve (AU-ROC) for dengue diagnosis of three clinical tools: the summation of manifestations of the WHO case definition, a predefined clinical scale and a logistic regression model obtained in this study.

Results: We compared 219 dengue cases (confirmed by laboratory) and 286 patients with other febrile illnesses. In a multiple model, variables independently associated with dengue included the duration of fever, sleepiness and exanthema. Rhinorrhea, cough and minimal leukocyte count were inversely associated with dengue. This model reached an accuracy of 84.2% (for a cut-off of >0.5, sensitivity: 79.5%, specificity: 87.9%, positive predictive value: 83.7%, negative predictive value: 84.6%). The AU-ROC of this model (89.8%) was significantly higher than that obtained with either the predefined scale (82.1%) or the WHO definition manifestations (77%).

Conclusion: We validated a predefined scale and identified a multiple model suitable for the clinical diagnosis of dengue in the pediatric population.

Keywords: accuracy; case definition; children; clinical diagnosis; dengue.