Estimates of the changing age-burden of Plasmodium falciparum malaria disease in sub-Saharan Africa

Abstract

Estimating the changing burden of malaria disease remains difficult owing to limitations in health reporting systems. Here, we use a transmission model incorporating acquisition and loss of immunity to capture age-specific patterns of disease at different transmission intensities. The model is fitted to age-stratified data from 23 sites in Africa, and we then produce maps and estimates of disease burden. We estimate that in 2010 there were 252 (95% credible interval: 171-353) million cases of malaria in sub-Saharan Africa that active case finding would detect. However, only 34% (12-86%) of these cases would be observed through passive case detection. We estimate that the proportion of all cases of clinical malaria that are in under-fives varies from above 60% at high transmission to below 20% at low transmission. The focus of some interventions towards young children may need to be reconsidered, and should be informed by the current local transmission intensity.

Figure 1. Incidence of clinical malaria in 0–5 year olds plotted against prevalence.

Solid symbols indicate studies in which both clinical incidence and parasite prevalence were reported; hollow symbols indicate studies in which only clinical incidence is reported and parasite prevalence was estimated from alternative data (based on the prior EIR). The legend states the country, and also the place and year(s) where necessary to identify the study. Units for incidence are episodes per person per year.

Figure 2. Observed and fitted incidence of clinical malaria by age.

(a–w) Model fits are the median and 95% interval of the predicted incidence over the joint posterior distribution of the parameters and study-specific random effects. Sites are labelled by country and place, with the study year(s) to distinguish studies in the same place. Plots are ordered by posterior EIR. The blue solid symbols are the observed data; the green lines and shaded areas are the fitted model with 95% credible intervals; and the hollow orange symbols are the fitted model averaged over the age groups in the observed data.

Figure 3. Fitted incidence and age distribution of cases plotted against prevalence.

(a) The estimated relationship between parasite prevalence in 2–10 year olds and clinical incidence of disease in 0–5 years olds. The shaded areas represent the 95% credible intervals. (b) The estimated relationship between parasite prevalence in 2–10 year olds and overall clinical incidence (red solid line). The green dashed line shows the relationship estimated in Patil et al. (c) The shifting age-burden of disease at different levels of endemicity. The figure shows the estimated proportion of cases in each age group plotted against prevalence. (d) The estimated relationship between prevalence and clinical disease incidence in 0–5 year olds by detection method: blue—daily ACD; dark green—weekly ACD; orange—PCD. Lines are model predictions; solid symbols are observed incidence and prevalence; hollow symbols are observed incidence and model-fitted prevalence (based on the prior EIR).

Figure 4. Age distribution of cases across Africa.

(a) 0–5 years old; (b) 5–15 years old; (c) over 15 years old.

Figure 5. The relationship between incidence and prevalence, if transmission has recently declined.

Panels show (a) incidence in under-fives against prevalence in 2–10 year olds, and (b) the proportion of cases in under-fives, following a change in transmission. (c) The prevalence before the decline in EIR plotted against current prevalence. Each coloured line is for a different reduction in EIR over the previous 10 years from 0 to 90%, as marked in panel c.

Figure 6. Model flow diagram.

S, susceptible, D, clinical disease, A, asymptomatic patent infection, U, subpatent infection, T, treated and P, prophylaxis. The parameters are defined in the text.