Overview

Robust Analytics for Adaptive Malaria Control

Human malaria can be understood as a set of loosely coupled, managed, locally peculiar, complex adaptive systems, with non-linear interactions among mosquitoes, parasites, humans, and malaria program managers. Flexibility in malaria control options (vector control, vaccines, and drugs), and the combination of complexity and heterogeneity in malaria transmission and responses to vector control create a need for information to manage malaria and methods for synthesis.

Malaria Analytics

The challenges we address herein will inevitably involve new research in malaria, but we are more concerned with malaria analytics – the systematic analysis of data to for malaria policy and decision support.

The endpoint of analytics is to give advice about a decision or policy. Decisions are made and policies get implemented on schedules, and there is rarely enough evidence to make those decisions with any degree of certainty. Decisions will get made and actions taken regardless, so the advice should be based on the best available evidence, even if it is weak. It is better to convey advice based on weak evidence than give no advice at all, but since lives could depend on the answers, the analysis should be done with a high degree of rigor. If the available evidence is weak,
some thought should be given to identification of key data sources that are missing and that would be most likely to improve the evidence for the next round of policies. It follows that inference for decision support and policy should be directed by a slightly different set of principles than the kind of inference that is taught at most universities to support basic research.

Analysts and scientists use the same methods – any analysis of data relies on the same statistical concepts and similar algorithms regardless of the purpose of analysis – but analysis must be designed to handle uncertainty for policy. In this way, malaria analytics needs a very different set of inferential rules than the ones used for scientific research. Inference for malaria analysts involves a great deal of malaria domain specific knowledge and methods. There is a need for a bespoke inferential system to support malaria analytics developed on stable systems to manage data and information. Malaria analysts should get some professional training. The advice should be robust to uncertainty, in the sense that it has gone to great lengths to characterize and quantify uncertainty, and to propagate the uncertainty through analysis and development of policy advice. This is the essence of robust analytics.

We address these issues in greater depth in Malaria Analytics and its vignettes (sidebar).

Information Systems

• Lear about malaria

Malaria Intelligence

Developing Maps

Malaria Theory & Simulation

Complexity

Synthesis

Robust Analytics

Uncertainty

Relevant Detail

Adaptive Malaria Control

The Value of Information

Information is valuable when it is used to improve policies, but gathering information costs money.