Improving measles incidence inference using age-structured serological data

Measles is a target for elimination in all six WHO regions by 2020, and over the last decade, there has been considerable progress towards this goal. Surveillance is recognised as a cornerstone of elimination programmes, allowing early identification of outbreaks, thus enabling control and preventing re-emergence. Fever–rash surveillance is increasingly available across WHO regions, and this symptom-based reporting is broadly used for measles surveillance. However, as measles control increases, symptom-based cases are increasingly likely to reflect infection with other diseases with similar symptoms such as rubella, which affects the same populations, and can have a similar seasonality. The WHO recommends that cases from suspected measles outbreaks be laboratory-confirmed, to identify ‘true’ cases, corresponding to measles IgM titres exceeding a threshold indicative of infection. Although serological testing for IgM has been integrated into the fever–rash surveillance systems in many countries, the logistics of sending in every suspected case are often beyond the health system's capacity. We show how age data from serologically confirmed cases can be leveraged to infer the status of non-tested samples, thus strengthening the information we can extract from symptom-based surveillance. Applying an age-specific confirmation model to data from three countries with divergent epidemiology across Africa, we identify the proportion of cases that need to be serologically tested to achieve target levels of accuracy in estimated infected numbers and discuss how this varies depending on the epidemiological context. Our analysis provides an approach to refining estimates of incidence leveraging all available data, which has the potential to improve allocation of resources, and thus contribute to rapid and efficient control of outbreaks.