Conference is now cancelled

Disentangling detection in a disease elimination setting

Emily Nightingale (London School of Hygiene and Tropical Medicine)

In a successful disease elimination setting, cases become increasingly sparse in space and sporadic in time. This is what we want – for overall incidence and transmission to reduce – however at the same time this makes it increasingly difficult to detect each individual case and assess the progress of the elimination programme. As awareness, motivation, resources and immunity decline it is increasingly important to understand the difference between true and observed incidence.

The SPEAK India consortium aims to support the elimination of visceral leishmaniasis (VL) from the remaining endemic area of North-Eastern India. Data are routinely collected for each diagnosed case, detailing the individual’s pathway from initial reporting to treatment and follow up, along with personal characteristics. GPS locations are also obtained for all resident villages of cases and health facilities with VL capacity. Longer delays and greater distance travelled for diagnosis and treatment can serve as warning signs for gaps in surveillance, and detection effort may additionally vary due to the perception of some areas as “non-endemic”. Insight into how these processes are manifesting for VL in India may allow us to draw a more realistic picture of progress towards true elimination.

We hope to use the results of a blanket screening study conducted across a sample of villages to define a “best case scenario” detection probability, against which to evaluate routinely-reported incidence in the same area. We will explore the possibility of modelling disease incidence at the village level as a log-Gaussian Cox process with spatially varying detection probability, fit using the INLA/SPDE approach implemented in the R package INLAbru. This example could also provide an interesting opportunity to experiment with some of the more recent additions to the INLAbru package, such as exceedance distributions and barrier models.