Reshaping the use of digital tools to fight malaria

by   Sekou L. Remy, et al.
University of Oxford

In this extended abstract we present our approach to marshal digital tools in the fight against malaria. We describe our scalable infrastructure which leverages abstractions to support effective deployment of existing computational models and their associated data.



There are no comments yet.


page 2


Interactive Cognitive Assessment Tools: A Case Study on Digital Pens for the Clinical Assessment of Dementia

Interactive cognitive assessment tools may be valuable for doctors and t...

Decentralised Trust for the Digital Economy

We propose a research initiative to explore and evaluate end-user techno...

Serverless on FHIR: Deploying machine learning models for healthcare on the cloud

Machine Learning (ML) plays a vital role in implementing digital health....

Attentive Neural Processes and Batch Bayesian Optimization for Scalable Calibration of Physics-Informed Digital Twins

Physics-informed dynamical system models form critical components of dig...

On Supporting Digital Journalism: Case Studies in Co-Designing Journalistic Tools

Since 2013 researchers at University College Dublin in the Insight Centr...

Modelling and Analysing the Landing Gear System: a Solution with Event-B/Rodin

This paper presents a solution to the landing gear system case study usi...

A learning perspective on the emergence of abstractions: the curious case of phonemes

In the present paper we use a range of modeling techniques to investigat...
This week in AI

Get the week's most popular data science and artificial intelligence research sent straight to your inbox every Saturday.

Ii Problem

Applying computational models and relevant data to interesting malaria settings is often difficult to configure (including acquisition of the requisite data), difficult to execute (especially at scientifically relevant scales), and difficult to interpret.

Iii Approach

In [4] we demonstrated an approach which addresses this problem, with the emphasis on exploration and interpretation. We developed an infrastructure by applying common computing abstractions in software development and deployment, and then applied three classes of algorithms to generate insight from the developed infrastructure.

Iii-a Infrastructure for Modeling and Computing

Fig. 1: Conceptual architecture of the infrastructure.

Iii-A1 Worker

Using containers111, we have packaged malaria models in a manner which is easy to deploy at scale and in multiple types of computing environments. In this container we also couple the model with software to communicate the desired input file, and to process the output files as needed. Together, these tools are referred to as a worker, and multiple workers can be deployed on a single machine or even distributed across the Internet.

Iii-A2 Task Clerk

We permit users to define a specific instantiation of a model that they’d like to evaluate. These parameters are use to “germinate” a seed input file for the model. The resulting file is then sent as a task to the worker so that it can be processed.

At present the parameters define two properties of an intervention policy: the portion of the households where insecticide treated bednets (ITN) are deployed, and the portion of households where indoor residual spraying (IRS)is applied. Other parameters provided by the model will be added in future work.

Iii-A3 Data store

The seed input file, as well as the results for all evaluated policies are stored in a central repository. The task clerk and all the distributed workers are connected via a common messaging fabric to this data store. Results from the model’s execution are converted to the cost per Disability Adjusted Life Year averted. This measure is implemented to provide direct comparison with earlier published work [5] This data is also stored for subsequent reporting in the case that intervention is requested in the future.

Iii-A4 Task queue

To tie each of the components together, we use a messaging fabric. The current implementation harnesses AMQP as implemented in the RabbitMQ message queue. The frontend posts jobs to a message queue which are subsequently picked up by idle workers of the appropriate type. When the worker is complete, the results are posted to a different channel on the same queue. This instantiation permits workers to be deployed in a wide range of environments, with little requirements on coordination.

Iii-B Applying Artificial Intelligence

The tools of Machine Learning or AI can now use the aforementioned infrastructure to assist with finding optimal or more optimal strategies as recommendations to malaria policy makers across the globe. AI can achieve this task in a way which has not been possible using all of the research capacity and compute we already have access to. In

[4] we framed the process of finding an optimal malaria policy as a stochastic multi-armed bandit problem, and implement three agent-based strategies to find optimal policies for a single environment. The selected algorithms were already in existence, and the novelty of our early work resides in the application domain. However we believe the domain also provides an opportunity to justify the development of new algorithms which can better manage the complexity of the malaria problem.

Fig. 2: Combined results from [4] showing AI (optimal), Current and Expert Human Solutions.

Figure 2 depicts the results of using AI to question what is the most cost-efficient combination of two interventions (ITN and IRS). The results are captured as surfaces where the optimal policies are dark red. We do not assert that these results are better than the insight that is common in current policy making (although they do challenge the prevalent view). Instead we consider these as the beginning of a dialog about both current policy and modeling approaches. Our motivation is to use these techniques to explore more complex policy recommendations.

Iv Summary

In this manuscript we presented our approach to marshal digital tools in the fight against malaria. We describe our scalable infrastructure which leverages abstractions to support effective deployment of existing computational models and their associated data. We also point to our earlier work leveraging such an infrastructure to effectively make more optimal policy recommendations with AI.

For the future of our work, we envision contributions in four complementary yet distinct areas:

  1. Evidence-based policy improvement/implementation,

  2. Augmented intelligence for complex decision making,

  3. Better simulations, effectively using computation,

  4. Transparent policy-based sharing of compute, data, models, and results towards global eradication.

Moving forward, we are assembling a diverse, geographically distributed consortium of institutions to help shape what we hope to be a catalyst for the eradication of malaria.


We are indebted to our managers Aisha Walcott-Bryant, Professor Stephen Roberts, and Komminist Weldemariam for their guidance and support.


  • [1] T. Smith, G. F. Killeen, N. Maire, A. Ross, L. Molineaux, F. Tediosi, G. Hutton, J. Utzinger, K. Dietz, and M. Tanner, “Mathematical modeling of the impact of malaria vaccines on the clinical epidemiology and natural history of plasmodium falciparum malaria: Overview,” The American journal of tropical medicine and hygiene, vol. 75, no. 2_suppl, pp. 1–10, 2006.
  • [2] S. N. Arifin, G. R. Madey, and F. H. Collins, Spatial agent-based simulation modeling in public health: design, implementation, and applications for malaria epidemiology.   John Wiley & Sons, 2016.
  • [3] M. Gambhir and C. Hettiarachchige, “Making sense of consensus: comparative modelling of malaria interventions,” The Lancet Global Health, vol. 5, no. 7, pp. e638 – e639, 2017. [Online]. Available:
  • [4] O. Bent, S. L. Remy, S. Roberts, and A. Walcott-Bryant, “Novel exploration techniques (nets) for malaria policy interventions,” in Proceedings of the Thirtieth Conference on Innovative Applications of Artificial Intelligence, 2018.
  • [5] E. M. Stuckey, J. Stevenson, K. Galactionova, A. Y. Baidjoe, T. Bousema, W. Odongo, S. Kariuki, C. Drakeley, T. A. Smith, J. Cox, and N. Chitnis, “Modeling the cost effectiveness of malaria control interventions in the highlands of western Kenya,” PLoS ONE, vol. 9, no. 10, 2014.