The relationship between reported fever and Plasmodium falciparum infection in African children

Abstract

Background: Fever has traditionally served as the entry point for presumptive treatment of malaria in African children. However, recent changes in the epidemiology of malaria across many places in Africa would suggest that the predictive accuracy of a fever history as a marker of disease has changed prompting calls for the change to diagnosis-based treatment strategies.

Methods: Using data from six national malaria indicator surveys undertaken between 2007 and 2009, the relationship between childhood (6-59 months) reported fever on the day of survey and the likelihood of coincidental Plasmodium falciparum infection recorded using a rapid diagnostic test was evaluated across a range of endemicities characteristic of Africa today.

Results: Of 16,903 children surveyed, 3% were febrile and infected, 9% were febrile without infection, 12% were infected but were not febrile and 76% were uninfected and not febrile. Children with fever on the day of the survey had a 1.98 times greater chance of being infected with P. falciparum compared to children without a history of fever on the day of the survey after adjusting for age and location (OR 1.98; 95% CI 1.74-2.34). There was a strong linear relationship between the percentage of febrile children with infection and infection prevalence (R2 = 0.9147). The prevalence of infection in reported fevers was consistently greater than would be expected solely by chance and this increased with increasing transmission intensity. The data suggest that in areas where community-based infection prevalence in childhood is above 34-37%, 50% or more of fevers are likely to be associated with infection.

Conclusion: The potential benefits of diagnosis will depend on the prevalence of infection among children who report fever. The study has demonstrated a predictable relationship between parasite prevalence in the community and risks of infection among febrile children suggesting that current maps of parasite prevalence could be used to guide diagnostic strategies in Africa.

Figure 1

Graph showing the least-square linear regression line y = 1.376x (black dashed lines) over a scatter plot showing the continuous relationship between the percentage of febrile children with infection and infection prevalence among all children and plotted with the 95% confidence interval. Two lines surrounding the best-fit line (grey solid lines) define the confidence interval. We also show the point at which febrile infections exceeds 50% and the corresponding value for infection prevalence (black dotted horizontal and vertical lines) including a line illustrating the 50%-50% random chance of symptomatic infection.

Figure 2

Graph showing the least-square linear regression line y = 1.460x (black dashed lines) over a scatter plot showing the continuous relationship between the percentage of febrile children with infection and infection prevalence among children who did not report a fever and plotted with the 95% confidence interval. Two lines surrounding the best-fit line (grey solid lines) define the confidence interval. We also show the point at which febrile infections exceeds 50% and the corresponding value for infection prevalence among the asymptomatic population (black dotted horizontal and vertical lines) including a line illustrating the 50%-50% random chance of symptomatic infection.