Integration of phlebotomine ecological niche modelling, and mapping of cutaneous leishmaniasis surveillance data, to identify areas at risk of under-estimation

•Published and unpublished data on the geographical distribution of sandflies in Colombia were collated.•Ecological niche models for Pintomyia (Pifanomyia) longiflocosa and Psychodopygus panamensis were developed.•In two municipalities of different departments of Colombia, the niche models were used in a spatial regression analysis of reported case incidence.•The results of the spatial regression were used to guide a larger project on under-reporting of cutaneous leishmaniasis. Passive surveillance systems are thought to under-estimate the true incidence of American cutaneous leishmaniasis (ACL) by two- to five-fold. Ecological niche models based on remotely sensed data can identify environmental factors which favor phlebotomine vectors. Here we report an integrated approach to identifying areas at risk of cutaneous leishmaniasis by applying spatial analysis methods to niche model results, and local surveillance data, in two locations in Colombia with differing vector ecology. The objective was to identify townships in which later phases of the project could implement community-based surveillance to obtain direct estimates of under-reporting. The study was carried out in one municipality in each of two departments of the Andean region of Colombia: Pueblo Rico in Risaralda, and Rovira in Tolima. Niche mapping by maximum entropy, based on published and unpublished existing locations of Pintomyia (Pifanomyia) longiflocosa and Psychodopygus panamensis, and using variables on land cover, climate and elevation. Field catches were done in each municipality to test predictions of high relative probability of presence. The niche model results were included as a predictor in a conditional autoregressive spatial model, in which the outcome variable was the number of cases by township, as detected by passive surveillance. Having rarefied 173 geolocated records, 46 of Pi. longiflocosa and 57 of Ps. panamensis were used for the niche modelling. At the national level, both species had high relative probability of presence on parts of the slopes of the three Andean cordilleras. Pi. longiflocosa also has a high relative probability of presence in the higher parts of the Magdalena valley, as does Ps. panamensis in some areas close to the Caribbean coast. At the local level, field catches confirmed that Pi. longiflocosa was the most abundant species in Rovira, and likewise Ps. panamensis in Pueblo Rico. The spatial regression showed that the incidence of ACL, according to su