Robust Analytics for Malaria Policy and Adaptive Malaria Control
Evidence-based malaria management is uniquely challenging. Malaria, human disease caused by infection with malaria parasites, is part of a complex adaptive system – a system of non-linear interactions among humans, mosquitoes, parasites, and malaria managers that can change over time through the evolution of resistance, changing management strategies, or economic development. While all malaria transmission systems share some common features, each malaria system can have some locally peculiar elements because of differences in the intensity, seasonality, and stability of malaria transmission; the local mix of vector species and their behaviors; mosquito ecology and population fluctuation and regulation in response to forcing by climate and hydrology, heterogeneous resource landscapes, and biotic factors; human demography and behaviors; health systems, and many other factors. Because of the complexity and peculiarity, effective management of malaria requires information and analytical methods that are up to the tasks of learning about malaria transmission systems and giving policy advice that is tailored to the local context and robust to uncertainty.
The need for high quality analytics for malaria motivated development of a bespoke inferential framework that we have called RAMP (Robust Analytics for Malaria Policy). Iterative, robust analytics and engagement with malaria control programs is called adaptive malaria control, which borrows heavily from adaptive management developed for natural resources management.
Adaptive malaria control describes an iterative process of analysis and engagement to help malaria programs make operational decisions, optimize the allocation of resources, improve technical efficiency, and develop evidence-based policies to reduce burden and eliminate malaria. A goal for adaptive malaria control is to provide robust policy advice despite uncertainty. Through careful analysis of uncertainty and through consultation with program managers, surveillance systems are reviewed to identify key data gaps, and plans are made to fill those gaps and improve future decisions.
Malaria analytics – analysis to support malaria policy – is based on a synthtesis of information about malaria epidemiology, transmission dynamics, and control from all available data sources. Since data describing the local features of malaria are far from ideal, and since analysts must be accountable for the advice they give, the analysis should endeavor to characterize and quantify uncertainty and to propagate that uncertainty through the analysis.
We developed RAMP around malaria research and surveillance metrics and theory developed over a century of studying malaria. RAMP is a system of linked concepts, algorithms, protocols, and procedures that connect data and decision support. To be useful, malaria policy advice should recommend a course of action, but the advice should not fabricate a false sense of confidence about a course of action. Instead, the advice should present alternatives, an accurate sense of the degree of the confidence about outcomes, and any information that could help sway the decision one way or another. The motivating idea for robust analytics is that the policy advice would not change if the analysis had been done in some other reasonable way.
The purpose of this website is to be a resource for information about RAMP and adaptive malaria control. We have organized the discussion arount four major activities.
Information Systems
To implement adaptive malaria control, malaria programs must improve their data systems to support routinely updated data streams and iterative analysis. Malaria is managed on monthly and annual cycles, and on multi-year funding and strategic planning cycles. Adaptive malaria control must support these reporting and policy evaluation cycles by developing information systems – including protocols and procedures for data collection, data curation, and repeated routine analyses – to ensure that the analysis is repeated consistently, that it is of the highest quality, and that can be updated in response to changing needs.
Malaria Intelligence
A key concept in adaptive malaria control is malaria intelligence, information about malaria that is needed for decision support in a country or management region. Malaria intelligence is focused on estimating the quantities that are needed for outbreak response, monitoring and evaluation, scenario planning, stratification and sub-national tailoring, and strategic planning.
Malaria programs must rely on surveillance metrics for routine decision support, but such metrics are not replicable. Given the intrinsic biases, it is necessary to find some way of estimating the magnitude of the bias and correcting it. Malaria intelligence leverages malaria research data to make the most of malaria surveillance data for malaria analytics. Facility data, including statistics such as case counts or the test positivity rate (TPR), can be compared against malaria prevalence data, or parasite rate (PR), collected in a cross-sectional survey. At the core of malaria intelligence are algorithms that estimate the PR in relation to the local annual entomological inoculation rate (aEIR) and antimalarial drug taking.
Unlike malaria research, the scope of interest for a malaria manager is not malaria generically, but malaria in some specific place: a country, or perhaps some part of it. The goal of adaptive malaria control is to learn about malaria over time in the localities that a malaria program is responsible for managing. Malaria intelligence assets represent an up-to-date assessment of malaria in the areas under management. The assets serve as an important bridge between statistical analysis of the past, and forecasts and strategic plans for the future.
If we start from the premise that malaria analytics should be capable of guiding resource allocation decisions, then we must evaluate policy scenarios representing tradeoffs between modes of control that affect malaria in different ways: suppressing transmission through vector control or investing in targeted medical preventative measures (e.g. vaccines). Such tradeoffs are difficult to evaluate without using some sort of mathematical model, usually dynamical systems models for malaria epidemiology, transmission dynamics, and control. In adaptive malaria control, malaria intelligence is thus designed around the quantities needed for simulation models and malaria data describing transmission.
Simulation and Policy
Adaptive malaria control relies heavily on simulation-based analytics. Models of malaria transmission guide development of malaria intelligence, play a role in developing a synthetic picture of local malaria transmission. They are the tool used for scenario planning and strategic planning, and after to analyze and prioritize missing data. Since it is expensive to get new information, and since those resources could be spent in other ways, we can use simulation models to design experiments in silico that can help us to understand the value of information and to use natural experiments to learn about malaria in context.
RAMP algorithms are based around models and the supporting theory that has come from a century of malaria studies and from challenging models to malaria data. These models are usually formulated as dynamical systems describing malaria epidemiology & transmission dynamics, mosquito ecology, and malaria control. A new framework for model building to support RAMP with supporting software, called SimBA, has been developed. SimBA supports nimble model building to reduce the response times for consultation with simulation models, and scalable complexity to add realism and detail to models, making it possible to formulate models at the right level of complexity for the problem at hand.
Using SimBA software, we can develop analytics pipelines, organized workflows that can replicate a simple policy analysis many times. The software makes it possible to automate model building and simulation-based analytics to generate ensembles – each analysis is drawn from the overall uncertainty, and taken altogether, the ensemble has propagated uncertainty from malaria intelligence through policy advice.
Adapting
If all the policy analyses in an ensemble have come to a consensus, then there
etcetera
These concepts are described in greater depth in short essays that can be accessed through links in the navigation bar and sidebar.