Optimal geospatial design of malaria surveillance programs

Targeted surveillance programs are critical to achieving malaria elimination goals. Geospatial maps of disease risk support the targeting of resources for a range of public health activities, including surveillance. Such maps include those developed by Applicants Shearer, Flegg and Gething, which employ a range of machine learning-based geostatistical tools to estimate spatial variation in disease risk. The geospatial design of disease surveillance programs – e.g., the number and location of surveillance sites – should be optimised to achieve public health and research objectives, within logistical, financial, and ethical constraints on resources.

Existing approaches to prioritise surveillance locations tend to only (formally) consider predictive mapping outputs and corresponding uncertainties. New approaches are required for determining the optimal geospatial design of malaria surveillance program. Optimal design is a research area combining statistical information-theory (i.e., information obtained by measuring at a location), with optimisation tools to determine the “best” allocation of limited resources for a given aim. To design optimal disease surveillance programs, we require new optimisation methods that capture these objectives and incorporate geospatial output which this project will work to achieve.

The work conducted in this Seed Grant project will involve two parts:

1. Refine and develop the optimisation method. A key component of an optimisation problem is the utility function which, in our context, measures the utility of selected surveillance locations. For example, identifying the optimal location of surveillance sites that maximises the probability of detecting Pk while minimising resources. Defining a suitable utility function will involve testing different approaches to combining various geospatial data-sources (e.g. site accessibility, health facility locations) and outputs from predictive risk maps to robustly identify “optimal” resource allocation. Note: a previously developed Pk risk map (informed by case data) will be used to identify locations at high risk of Pk transmission, no other epidemiological data is required.
2. Apply to case study of P. knowlesi malaria surveillance in Indonesia as a proof-of-concept. Using the findings from Part 1, we will construct and evaluate a utility function that incorporates factors relevant to surveillance program design for the Pk case study and determine the “best” allocation of surveillance resources. Stakeholder engagement with these results will be done under the auspices of the ACIAR DFAT project.

Surveillance

Aug 2020 — Oct 2023

$15,000

Australia

Australian Centre for Research Excellence in Malaria Elimination (ACREME)

Plasmodium knowlesi