Using georeferenced phylogenetic trees, phylogeography allows researchers to elucidate interactions between environmental heterogeneities and patterns of infectious disease spread. Concordant with the increasing availability of pathogen genetic sequence data, there is a growing need for tools to test epidemiological hypotheses in this field. In this study, we apply tools traditionally used in ecology to elucidate the epidemiology of foot-and-mouth disease virus (FMDV) in Uganda. We analyze FMDV serotype O genetic sequences and their corresponding spatiotemporal metadata from a cross-sectional study of cattle. We apply step selection function (SSF) models, typically used to study wildlife habitat selection, to viral phylogenies to show that FMDV is more likely to be found in areas of low rainfall. Next, we use a novel approach, a resource gradient function (RGF) model, to elucidate characteristics of viral source and sink areas. An RGF model applied to our data reveals that areas of high cattle density and areas near livestock markets may serve as sources of FMDV dissemination in Uganda, and areas of low rainfall serve as viral sinks that experience frequent reintroductions. Our results may help to inform risk-based FMDV control strategies in Uganda. More broadly, these tools advance the phylogenetic toolkit, as they may help to uncover patterns of spread of other organisms for which genetic sequences and corresponding spatiotemporal metadata exist.

Estimates of malaria burden and intervention uptake in Africa are primarily based on household surveys. However, their expense and infrequency limit their utility. We investigated whether data collected during antenatal care (ANC) can provide relevant information for decision-makers. Malaria test positivity rates and questionnaire data from ANC attendees at 39 health facilities were compared to questionnaire data and positivity rates among children from two cross-sectional surveys in the facilities' corresponding catchment areas. Trends in parasitemia among ANC attendees were predictive of trends in parasitemia among children at the council level (mean absolute error 6.0%). Primigravid ANC attendees had the lowest rates of net ownership (modeled odds ratio [OR] 0.28, 95% CI 0.19-0.40) and use (OR 0.58, 95% CI 0.42-0.79). ANC attendees reported higher levels of care-seeking (OR 1.78, 95% CI 1.48-2.14), malaria testing (OR 4.16, 95% CI 3.44-5.04), and treatment for children with fever (OR 7.66, 95% CI 4.89-11.98) compared to women surveyed in households, raising concerns about social desirability bias disproportionately impacting ANC surveys. ANC surveillance is an effective strategy for tracking trends in malaria burden. More work is required to elucidate the value of administering questionnaires to ANC attendees.

Estimates of malaria prevalence and intervention coverage in Africa are primarily based on nationally representative household (HH) surveys. However, the expense and infrequency limit the utility of HH surveys for operational action by malaria programmes. We assessed whether data collected during first antenatal care (ANC1) visits, consisting of data on malaria prevalence using rapid diagnostic tests, ownership of insecticide-treated nets (ITNs) and treatment-seeking for children with fever, could provide relevant data to guide decision-makers. Malaria prevalence among ANC1 attendees in select areas of six countries (Benin, Burkina Faso, Mozambique, Nigeria, Tanzania and Zambia) was compared with prevalence data among children under 5 and school-aged children from cross-sectional HH surveys in the same areas. To examine the relationship between prevalence among ANC1 attendees and children, we fitted a linear trend to the log-OR of the risk of testing positive. The predictive performance of the model was assessed by leave-one-out cross-validation (LOOCV). District-level ANC1 prevalence and prevalence among children are correlated (Spearman’s rank correlation, r=0.79, 95% CI=0.65 to 0.88, p<0.001) and ANC1 prevalence is predictive of prevalence among children (LOOCV mean absolute error=6.5%). To understand whether data on ITN ownership collected at ANC1 are representative of ownership in the underlying communities, we assessed the district-level proportion of HH ownership in five countries (Benin, Burkina Faso, Mozambique, Nigeria and Zambia) and fitted an ordinal regression model to the ranking of ownership by district. Reported rates of treatment-seeking for children under 5 with fever, testing for malaria and treatment for the HH and ANC1 settings were compared. Estimates of malaria prevalence and ITN coverage derived from ANC1 attendees correlate well with HH survey estimates and may be useful in monitoring malaria prevalence and prevention efforts. In contrast, data on treatment-seeking does not appear useful.