Joshua E. Cogan

Global Malaria Epidemiology

Many countries are implementing measures to control and eliminate malaria. To maximise the effectiveness, it is vital that planners have access to the best metrics about malaria risk. In a major new Bill and Melinda Gates Foundation funded project, the Malaria Atlas Project (MAP) is working to synthesise complex mathematical models to compute available epidemiological data with geostatistical models for large, spatially-structured data sets, in order to more accurately predict malaria risk.

#
 
Improved modelling framework

There exist a wide range of diverse malaria metrics, which are measured to varying degrees in different locations. MAP is developing a geospatial modelling framework to relate these metrics. A recent project was to characterise the prevalence-to-incidence relationship through mathematical modelling and ensemble prediction. MAP is currently working to incorporate serological markers of historical parasite exposure into the spatio-temporal framework of existing point prevalence analyses, for example mapping transmission intensity across Cambodia in collaboration with Prof Chris Drakely and Dr Jonathan Cox.

Related work is to refine earlier mathematical modelled representations of the relationships between prevalence and classical epidemiological parameters (Force of Infection, Entomological Inoculation Rate, Reproductive Number Under Control and Basic Reproductive Number). Building on the generalised modelling package for multi-scale simulation led by Dave Smith, the second phase is to embed this tool within the geographic framework, to allow mapping from environmental covariates to the observed metrics of exposure in human hosts.

#
 
Seasonality

Most malaria affected countries experience seasonal variation in transmission. Georeferenced time-series data from peer-reviewed articles and routine case reporting are collated and an inferential model constructed to relate transmission data to a suite of temporally dynamic environmental covariates (temperature, vegetation, humidity etc.) from the MODIS remote sensing platform. The ultimate aim is to be able to accurately predict seasonal malaria transmission patterns (onset, duration, magnitude) in locations where malaria survey data is sparse, using only the environmental covariate data which is available for all locations.

#
 
Vectors

In earlier work, MAP has systematically assembled data on the 41 locally dominant anopheles vectors of human malaria, and produced geospatial models to predict their global distribution. The next stage is to integrate vector distribution models with estimated relative vector abundance and capacity indices to feed into malaria risk modelling.

#
 
Intervention coverage and receptive risk

Although many countries have extensive malariometric data from recent years, it is often preferable to compare historical (or receptive) risk levels, prior to the scale up of control interventions. MAP is exploring alternative pathways to define receptive risk.

#
 
Human mobility

Working in collaboration with Prof Andrew Tatem at the University of Southampton, MAP is using mobile phone records, micro-censuses and remotely sensed data to quantify patterns of human movement, mobility and migration. Such movement matrices can infer by probability the possible locations where malaria exposure occur, to allow conversion of incidence rates reported at health facilities to a geographical catchment area, which will greatly enhance malaria risk modelling. MAP is collaborating with Google Earth Engine to develop a geographic processing toolkit as a pre-cursor to this catchment estimation.

#
 
Optimised risk maps

There a huge challenges in generating risk stratifications within diverse settings. Large heterogeneity exists in data metrics and survey type, and a wealth of new high-level covariate products emerge including human movement patterns and vector species abundance as described above. MAP is developing a single overarching methodological architecture which will allow risk stratifications to be responsive to contextual factors, but produce standardized outputs.

#
 
Improved intervention coverage models

Three major interventions used globally are insecticide treated nets (ITN), artemisinin-based combination therapy (ACT) and indoor residual spraying (IRS). Exploratory analysis suggest that ITN coverage is highest where malaria risk is lower, and lowest where nets are most needed. Earlier work by MAP has generated detailed models of ITN, ACT and IRS coverage in parallel to spatiotemporal reconstructions of changing endemicity through time in Africa. MAP is working to extend the African ITN model to triangulate data from net manufacturers, national malaria control programmes and household surveys into coverage models. This will allow definitive evaluation of ITN coverage and enable future ITN distribution can be more targeted and risk driven. Similar models will be produced for ACTs, IRS, (and where regionally appropriate) intermittent preventive treatment in pregancy (IPTp) and seasonal malaria chemoprevention (SMC).