Figure 3
The error in prevalence measurements becomes more important at low MOI. a) Prevalence estimates are biased due to imperfect detection. Assuming that infecting clones within a particular host are independent from each other, the probability of missing all of them and therefore falsely classify an individual as negative, is highest for low multiplicity of infection. This graph shows - for different values of q - how the number of clonal infections in a host affects the estimates of prevalence. The probability of correctly recognizing a positive individual with n infections is calculated as 1 - (1 - q)n. It follows that the effect of detectability on prevalence estimates is highest at low multiplicity of infection and therefore low transmission, for example when being close to local elimination. However, low transmission intensity might prevent acquisition of immunity and therefore raise the value of detectability. It is therefore desirable to report estimates of q and multiplicity of infection together with prevalence estimates. b) The distribution of MOI. Contrary to common practice, observations from all four surveys are pooled for the calculation of MOI. This corresponds to the assumption that clones are present throughout all surveys if detected once. With the help of figure 3.a, the bias on prevalence estimates in this population, as introduced by imperfect detection, can be roughly estimated.