Supporting elimination of lymphatic filariasis in Samoa by predicting locations of residual infection using machine learning and geostatistics.

The global elimination of lymphatic filariasis (LF) is a major focus of the World Health Organization. One key challenge is locating residual infections that can perpetuate the transmission cycle. We show how a targeted sampling strategy using predictions from a geospatial model, combining random forests and geostatistics, can improve the sampling efficiency for identifying locations with high infection prevalence. Predictions were made based on the household locations of infected persons identified from previous surveys, and environmental variables relevant to mosquito density. Results show that targeting sampling using model predictions would have allowed 52% of infections to be identified by sampling just 17.7% of households. The odds ratio for identifying an infected individual in a household at a predicted high risk compared to a predicted low risk location was 10.2 (95% CI 4.2-22.8). This study provides evidence that a 'one size fits all' approach is unlikely to yield optimal results when making programmatic decisions based on model predictions. Instead, model assumptions and definitions should be tailored to each situation based on the objective of the surveillance program. When predictions are used in the context of the program objectives, they can result in a dramatic improvement in the efficiency of locating infected individuals.

Finding hotspots: development of an adaptive spatial sampling approach.

The identification of disease hotspots is an increasingly important public health problem. While geospatial modeling offers an opportunity to predict the locations of hotspots using suitable environmental and climatological data, little attention has been paid to optimizing the design of surveys used to inform such models. Here we introduce an adaptive sampling scheme optimized to identify hotspot locations where prevalence exceeds a relevant threshold. Our approach incorporates ideas from Bayesian optimization theory to adaptively select sample batches. We present an experimental simulation study based on survey data of schistosomiasis and lymphatic filariasis across four countries. Results across all scenarios explored show that adaptive sampling produces superior results and suggest that similar performance to random sampling can be achieved with a fraction of the sample size.

Designing malaria surveillance strategies for mobile and migrant populations in Nepal: a mixed-methods study.

BACKGROUND: As malaria cases have declined throughout Nepal, imported cases comprise an increasing share of the remaining malaria caseload, yet how to effectively target mobile and migrant populations (MMPs) at greatest risk is not well understood. This formative research aimed to confirm the link between imported and indigenous cases, characterize high-risk MMPs, and identify opportunities to adapt surveillance and intervention strategies to them. METHODS: The study used a mixed-methods approach in three districts in far and mid-western Nepal, including (i) a retrospective analysis of passive surveillance data, (ii) a quantitative health facility-based survey of imported cases and their MMP social contacts recruited by peer-referral, and (iii) focus group (FG) discussions and key informant interviews (KIIs) with a subset of survey participants. Retrospective case data were summarised and the association between monthly indigenous case counts and importation rates in the previous month was investigated using Bayesian spatio-temporal regression models. Quantitative data from structured interviews were summarised to develop profiles of imported cases and MMP contacts, including travel characteristics and malaria knowledge, attitudes and practice. Descriptive statistics of the size of cases' MMP social networks are presented as a measure of potential programme reach. To explore opportunities and barriers for targeted malaria surveillance, data from FGs and KIIs were formally analysed using a thematic content analysis approach. RESULTS: More than half (54.1%) of malaria cases between 2013 and 2016 were classified as imported and there was a positive association between monthly indigenous cases (incidence rate ratio (IRR) 1.02 95% CI 1.01-1.03) and the previous month's case importation rate. High-risk MMPs were identified as predominantly adult male labourers, who travel to malaria endemic areas of India, often lack a basic understanding of malaria transmission and prevention, rarely use ITNs while travelling and tend not to seek treatment when ill or prefer informal private providers. Important obstacles were identified to accessing Nepali MMPs at border crossings and at workplaces within India. However, strong social connectivity during travel and while in India, as well as return to Nepal for large seasonal festivals, provide opportunities for peer-referral-based and venue-based surveillance and intervention approaches, respectively. CONCLUSIONS: Population mobility and imported malaria cases from India may help to drive local transmission in border areas of far and mid-western Nepal. Enhanced surveillance targeting high-risk MMP subgroups would improve early malaria diagnosis and treatment, as well as provide a platform for education and intervention campaigns. A combination of community-based approaches is likely necessary to achieve malaria elimination in Nepal.

Household electricity access in Africa (2000-2013): Closing information gaps with model-based geostatistics.

Household electricity access data in Africa are scarce, particularly at the subnational level. We followed a model-based Geostatistics approach to produce maps of electricity access between 2000 and 2013 at a 5 km resolution. We collated data from 69 nationally representative household surveys conducted in Africa and incorporated nighttime lights imagery as well as land use and land cover data to produce maps of electricity access between 2000 and 2013. The information produced here can be an aid for understanding of how electricity access has changed in the region during this 14 year period. The resolution and the continental scale makes it possible to combine these data with other sources in applications in the socio-economic field, both at a local or regional level.

Predicting residential structures from open source remotely enumerated data using machine learning.

Having accurate maps depicting the locations of residential buildings across a region benefits a range of sectors. This is particularly true for public health programs focused on delivering services at the household level, such as indoor residual spraying with insecticide to help prevent malaria. While open source data from OpenStreetMap (OSM) depicting the locations and shapes of buildings is rapidly improving in terms of quality and completeness globally, even in settings where all buildings have been mapped, information on whether these buildings are residential, commercial or another type is often only available for a small subset. Using OSM building data from Botswana and Swaziland, we identified buildings for which 'type' was indicated, generated via on the ground observations, and classified these into two classes, "sprayable" and "not-sprayable". Ensemble machine learning, using building characteristics such as size, shape and proximity to neighbouring features, was then used to form a model to predict which of these 2 classes every building in these two countries fell into. Results show that an ensemble machine learning approach performed marginally, but statistically, better than the best individual model and that using this ensemble model we were able to correctly classify >86% (using independent test data) of structures correctly as sprayable and not-sprayable across both countries.

Mapping land cover change over continental Africa using Landsat and Google Earth Engine cloud computing.

Quantifying and monitoring the spatial and temporal dynamics of the global land cover is critical for better understanding many of the Earth's land surface processes. However, the lack of regularly updated, continental-scale, and high spatial resolution (30 m) land cover data limit our ability to better understand the spatial extent and the temporal dynamics of land surface changes. Despite the free availability of high spatial resolution Landsat satellite data, continental-scale land cover mapping using high resolution Landsat satellite data was not feasible until now due to the need for high-performance computing to store, process, and analyze this large volume of high resolution satellite data. In this study, we present an approach to quantify continental land cover and impervious surface changes over a long period of time (15 years) using high resolution Landsat satellite observations and Google Earth Engine cloud computing platform. The approach applied here to overcome the computational challenges of handling big earth observation data by using cloud computing can help scientists and practitioners who lack high-performance computational resources.

The relative contribution of climate variability and vector control coverage to changes in malaria parasite prevalence in Zambia 2006-2012.

BACKGROUND: Four malaria indicator surveys (MIS) were conducted in Zambia between 2006 and 2012 to evaluate malaria control scale-up. Nationally, coverage of insecticide-treated nets (ITNs) and indoor residual spraying (IRS) increased over this period, while parasite prevalence in children 1-59 months decreased dramatically between 2006 and 2008, but then increased from 2008 to 2010. We assessed the relative effects of vector control coverage and climate variability on malaria parasite prevalence over this period. METHODS: Nationally-representative MISs were conducted in April-June of 2006, 2008, 2010 and 2012 to collect household-level information on malaria control interventions such as IRS, ITN ownership and use, and child parasite prevalence by microscopic examination of blood smears. We fitted Bayesian geostatistical models to assess the association between IRS and ITN coverage and climate variability and malaria parasite prevalence. We created predictions of the spatial distribution of malaria prevalence at each time point and compared results of varying IRS, ITN, and climate inputs to assess their relative contributions to changes in prevalence. RESULTS: Nationally, the proportion of households owning an ITN increased from 37.8 % in 2006 to 64.3 % in 2010 and 68.1 % in 2012, with substantial heterogeneity sub-nationally. The population-adjusted predicted child malaria parasite prevalence decreased from 19.6 % in 2006 to 10.4 % in 2008, but rose to 15.3 % in 2010 and 13.5 % in 2012. We estimated that the majority of this prevalence increase at the national level between 2008 and 2010 was due to climate effects on transmission, although there was substantial heterogeneity at the provincial level in the relative contribution of changing climate and ITN availability. We predict that if climate factors preceding the 2010 survey were the same as in 2008, the population-adjusted prevalence would have fallen to 9.9 % nationally. CONCLUSIONS: These results suggest that a combination of climate factors and reduced intervention coverage in parts of the country contributed to both the reduction and rebound in malaria parasite prevalence. Unusual rainfall patterns, perhaps related to moderate El Niño conditions, may have contributed to this variation. Zambia has demonstrated considerable success in scaling up vector control. This analysis highlights the importance of accounting for climate variability when using cross-sectional data for evaluation of malaria control efforts.