Combining clinical and diagnostic surveillance to estimate the burden of measles disease: a modeling study

BackgroundThe clinical case definition for measles is highly sensitive and has low specificity. Diag-nostic confirmation can resolve this uncertainty for individual cases and is a crucial tool for confirmation of measles outbreaks. However, in under-resourced settings, it is prohibitive to confirm all suspected cases and routine measles surveillance comprises a combination of both clinically and diagnostically confirmed cases. MethodsWe developed a dynamic model of measles, rubella, and other sources of febrile rash to simulate time series of a suspected measles surveillance system. We simulated partial reporting of suspected cases and limited routine diagnostic testing using assays with sensitivity and specificity that correspond to current or proposed rapid diagnostic tests. We estimated the time series of reported measles cases as the product of suspected cases and the proportion of diagnostic positive cases. We then estimated the reporting rate and annual incidence for measles using the time-series SIR model. ResultsReconstructing the time series of reported measles cases using the fraction of diagnostic positive cases results in unbiased estimates of the reporting rate and the annual incidence at moderate vaccination levels for all reasonable levels of test sensitivity and specificity, even for low proportions tested. At high vaccination levels, diagnostic tests with low sensitivity (< 90%) lead to slight bias in annual incidence. Temporal variation in the prevalence of measles among suspected cases require that the proportion of cases attributable to measles be estimated frequently (i.e., monthly) to avoid bias in estimates. ConclusionCombining routine, systematic diagnostic confirmation of suspected measles cases with suspected cases surveillance can improve estimates of the reporting rate and annual incidence using diagnostic tests with sensitivity and specificity consistent with proposed rapid diagnostic tests.