DigitalCommunities: One Health AMR surveillance via digital twins

DigitalCommunities is a cloud-based, AI-powered solution for surveillance and early detection of food/water-borne diseases and AMR in One-Health communities. Data-driven modelling and AI-mining of community-sourced data is used to create up-to-date digital twins of the communities and issue warnings upon detected patterns.

Tania Dottorini, Primary Investigator, Professor of Bioinformatics, University of Nottingham, UK

Of 8.88M infection-related global deaths per year, 4.95M feature co-presence of AMR and 1.27M could have been entirely avoided if no AMR had been present[1,2]. Similar proportions are reflected by Brazil, Bangladesh and South Africa, which present a similar climate (Figure 1). AMR transmission can take place through different pathways, the food-chain being a major one[3]. Transmission may happen through food consumption or through contact with contaminated food/animals/environment[4] including water[5].

Globally, there is a lack of solutions for early detection of AMR. Widespread genetic and microbiological testing for surveillance is impractical due to the sheer size and costs of required sampling, and due to healthcare infrastructure limitations in LMICs. An appealing alternative is relying on predictive mathematical models to reduce the amount of required on-field testing, but the approach is hindered by the inherent complexity of tracking/reconstructing the cause-effect mechanisms underlying emergence and transmission of AMR in complex One Health communities.

Being able to digitally replicate One Health community dynamics involved with the emergence and spread of AMR and related diseases, poses significant challenges both in terms of data collection (size and diversity) and in terms of mathematical modelling (sheer complexity of interactions).

We are targeting populations affected by food/water-borne diseases in Brazil (43% of the population), Bangladesh (49%) and South Africa (46%). The ratio of deaths where AMR is involved is approximately 57% globally. If successful, we plan to evaluate deployment to other countries. The benefit is a more effective surveillance and early detection, thanks to the identification of observable variables (markers) of disease and AMR, customised around each community. For data collection, in addition to gathering from databases (Figure 6), we plan to approach the communities of Brazil by using a mobile app[12,13] they are already familiar with since 2007. In exchange, early warning feeds generated by our solution will be returned to selected actors within each community (e.g., healthcare facilities / local policy makers) with responsibility to manage interventions strategies towards the population. A similar approach will be deployed in Bangladesh and South Africa, pending availability of an equivalent app.

Finally, the accumulation of a great amount of diverse data over geographical locations and time periods will also constitute a useful reserve of information useful for further analyses available to researchers, companies, and institutions. The PI, Dottorini collaborates with global policy makers (WHO, FAO, UNICEF) and can promote further outreach.

The ultimate goal of the solution is to provide to the well-being of the public. The solution is meant to generate surveillance data and in particular, early warning of increased likelihood of AMR, for fruition by the general public. Because the primary means of access to the service for citizens is through the Internet, and -in case of Brazil – through the use of a mobile app, the benefits are globally accessible under fair and non-discriminatory terms. Further benefits derive from the body of data and knowledge generated both by the collection campaign (mixture of database interrogations, lab-generated and crowd-sourced information) and by datamining. This body of data and knowledge will be made publicly available initially across the countries that participated to the project, and later globally.

Finally, further benefits will derive from the refinement of the ML-based datamining and digital twinning methods, which can be turned into a tool of general applicability and deployability to further countries, with the purpose of obtaining further community-specific surveillance and AMR monitoring solutions. 

DigitalCommunities aims to create a tangible impact for the populations affected by food/water-borne diseases in Brazil (43% of the population), Bangladesh (49%) and South Africa (46%). The ratio of deaths where AMR is involved is approximately 57% globally. People living in disadvantaged social settings are more at risk for severe illness due to poor healthcare. In Bangladesh, vulnerable populations include also the Rohingya, segregated in refugee camps in poor health conditions, representing a major humanitarian crisis. In Brazil, the most vulnerable communities are in the north/northeast regions, favelas and indigenous reserves, and are subjected to a disproportionately higher exposure to infection. For most of these populations, in particular in Brazil and Bangladesh, the influence of climate and climate-related catastrophic events is also tangible, which makes it essential for our solution to incorporate climate/weather data in our models: floods, droughts, changes in rainfall, evapotranspiration, atmospheric water storage, precipitation, sea level, salinity and land usage, have a significant influence on emergence, re-emergence, multiplication, distribution and transmission of pathogens, as well as on the severity of the outbreaks. Tangible impact will be created by enabling the generation of early warning of diseases as well as increased likelihood of AMR emergence and spread.

Collaborative partnership and technology transfer: the consortium's strategic inclusion of a steering committee with members from Thailand, Vietnam, Turkey, India, Paraguay, Cameroon and collaboration with UN agencies will allow for sharing knowledge and prototypes, as well as receiving feedback on potential applicability outside of Brazil, Bangladesh and South Africa. Further dissemination/transfer via workshops, seminars and training.

Engagement with governmental agencies and policy makers: ongoing collaborative ties of the PI and partners with global policy makers such as WHO, FAO and UNICEF, as well as with several governmental agencies in particular in Bangladesh (Ministry of Science and Technology, Ministry of Health), will be exploited to maximise impact.

Engagement with industry partners and private sector: by nature of the project, we will liaise with actors active in cloud computing, software development, biotech as well as all the actors gravitating about livestock farming. By showcasing the potential applications and benefits of our research, we will attract industry interest and collaboration, enabling the development of innovative products and solutions.

Academia and Research: publications in peer-reviewed journals and dissemination at renowned conferences. Academic workshops and seminars to engage with fellow researchers, fostering collaboration and knowledge exchange.

Scientific success can be measured as the advancement of knowledge on AMR and as the discovery of new ARGs and of previously unknown correlations between AMR/related diseases and observable variables (markers). Quantitative information about the number and quality of scientific discoveries stemming from the application of the proposed method to selected One Health communities in China can be found in our previous work[6–11,14,15]. Based on previous works[6–11,14,15], we plan to measure the quantity of ARGs and species' abundance, the number of hotspots of transmission (for example between nose and hands of farm workers and animal faeces[8,9]), incidence and prevalence of bacteria and diseases, and correlated social/economical/environmental factors to track the spread of AMR within One Health settings. Markers (observable variables with a demonstrated, strong correlation to AMR/diseases) are in turn the starting point to the development of new early warning solutions, which directly translate into societal impact. Longer term success may be only measurable as a variation of incidence of disease and AMR cases after prolonged application of our proposed solution to the selected One Health communities. Further indirect impact is on economy, in terms of influence on healthcare costs savings, as well as on general welfare.  

Risks and mitigation strategies:

Stakeholder engagement: it might be that specific actors within the communities might become reluctant to the initiative over time, in particular in relation to performing periodic data collection. We will actively engage with relevant stakeholders (farmers, researchers, and clinicians) throughout the research process. By involving stakeholders early on, we can gather valuable insights, receive feedback, and align our research objectives with their requirements. This collaborative approach ensures that our findings are tailored to address the real-world challenges of fighting AMR in both HICs and LMICs, enhancing their potential for exploitation.

Local elections in Brazil that could affect the position of some partners in their current government roles. As we are collecting from multiple provinces and involving multiple actors, we hope this type issue should be mitigated by redundancy.

Importing consumables: Changes in import taxes and increased challenges of importing consumables for sampling and sequencing in Brazil and Bangladesh. We may need to scale down data collection. However, our initial plans involve a quite high number of communities, so reductions should be handled without compromising excessively the number of achievable digital community twins.

We are convinced that Trinity's objectives align closely with our own and the broader global community's priorities. The focus on enhancing health in LMICs by tackling the critical issue of AMR through a data-driven approach is particularly noteworthy. In regions where infrastructures are lacking, and distances are vast, the implementation of widespread genetic and microbiological testing for surveillance becomes impractical due to cost constraints and limitations in healthcare infrastructure. Early detection and rapid response to AMR is the key, but can only achieved through a novel generation of data-driven ideas. We have been working along these lines for a number of years, achieving notable successes with our initial deployment in China, so far primarily in terms of scientific discoveries and publications. We believe that our idea is now mature and ready for application to larger scale in order to produce an actually measurable impact on the population.

Our goal is to connect with a number of key actors.

Academic institutions in Brazil, Bangladesh, South Africa  will be instrumental through their public health/veterinary/epidemiology departments to port our proposed solution to ongoing and future research projects.

Other ideal partner organisations:

• Google: as we plan to develop and deploy our solution using their cloud technologies;
• The Patrick J McGovern Foundation: expertise on how to conceptualize and develop AI solutions could improve ours by providing groundwork on ethical AI and data usage, privacy and stewardship of data, and further innovative ways to apply AI;
• The Institute for Health Metrics and Evaluation: expertise in global health trends, potential of integrating their data with ours to expand the breadth of our datamining.
• Brunswick: expertise in critical issues can help us tell our story and findings in a meaningful way and understand if our solution can be disseminated globally;
• Internews: expertise in sharing health trustworthy news and information will be key for us to communicate our findings globally;
• GSK: expertise in developing new antibiotics could improve our understanding if the genetic elements driving new resistance traits, transmissibility, multidrug resistance, and infection severity could be validated an used for future antibiotics studies.