GOSAIC: Geospatial Platform to Predict Disease Impact

Dynamic prediction of transmission pathways and disease impact on populations through geographic correlation of the environment to human health.

Team lead is T.R. Price, PhD candidate Environmental Economics, Lincoln University, NZ. Masters in Environmental Studies, Sustainable Development and Climate Change. CEO Treetable LLC and Terex Maps.

Environmental degradation, social changes, economic hardship and disease are interrelated pathways that impact health. Environmental, social and economic data can be used in monitoring systems to predict the impact of diseases on populations. Current monitoring systems do not use all available data. Without using this data, our monitoring systems have little predictive power (pixel level) and are largely generalized to administrative units (states). 

Our monitoring systems largely rely on the social determinants of health(e.g., socio-economic, demographic, and genetic conditions). However, this data is spatially coarse, infrequently updated, and costly to measure. Frequently updated, publicly available, fine scale data on environmental determinants of health (temperature, precipitation, air quality) are available but are not extensively used in monitoring systems.

Monitoring systems that have related the environment to health have done so primarily in a single static moment and are not dynamic. Without being dynamic, the monitoring systems don't learn from the data they are monitoring and do not keep up with rates and amounts of change. Without monitoring environmental, social, and economic data and how they are changing, transmission pathways and impact of diseases to populations are unable to be accurately predicted at a fine scale.

According to the World Health Organization, 85 of the 102 major diseases are associated with the environment. A degraded environment disproportionately affects the most vulnerable populations, especially women and children. To ensure global health security and resilience, the environment must be incorporated into prediction models for disease impact. The information gained from GOSAIC will help government agencies, medical institutions, and supply chains (at the local to global level) to make decisions that decrease disease impact to populations. Effective environmental changes could fix at least 25% of the global disease burden. This was exemplified during the COVID-19 lockdown where a 63% reduction in NO2 over air pollution in Wuhan, China prevented an estimated 14,000 deaths.

The global community can increase their understanding of health through the non technical imagery portrayed on GOSAIC’s website. To remove language barriers and promote global accessibility, the written language used on the website is translated to the user’s language. Our global community connections (both current and future) allow us to understand the needs for GOSAIC. We expand the reach of our global community through social features (messaging, forum space), external events (hackathons) and feedback (user surveys and reviews).

Environmental, social, and economic features are all interrelated and together impact disease epidemiology. Human created change (deforestation) and natural change (fires) interact with these features and the disease epidemiology. By improving the accuracy of disease prediction models and learning which key features impact disease, the wellbeing of populations will be increased. The analyzed information from GOSAIC provides decision makers (government officials, individual entities) with the data they need to make more sustainable and equitable decisions. This information is available to everyone, so that they can be in charge of decisions regarding their health. A good portion of the data from GOSAIC is portrayed in imagery. Imagery is not dependent on language, however, it is dependent on senses. Recognizing this, we will work to make this data accessible to the sensory impaired. Throughout GOSAIC, we will try our best to make it globally accessible, fair, reasonable, and non-discriminatory. 

Knowing that our health is dependent on our environment, from the start GOSAIC will support nature-based interventions. Nature-based interventions improve over time, unlike most human created interventions. For example, as a tree grows its ability to filter the air improves whereas a face mask filtering ability degrades. 

Implementing interventions can improve not only our health but also our productivity. The entirety of GOSAIC’s operations is planned to be carbon (CO2) negative. This is done by partnering with organizations to pay people in developing countries to plant and maintain trees. By decreasing CO2 levels (in areas with high carbon dioxide levels), higher levels of productivity (cognitive functions) are able to be experienced. 

GOSAIC’s monitoring system eases the burden for monitoring disease impact on populations. Easing this burden allows for more resources to go towards proactive interventions, such as nature-based interventions (or other data backed interventions), that improve the health and productivity of everyone. 

GOSAIC is especially important for populations (especially women and children) in developing countries, and other vulnerable areas. These areas are disproportionately impacted by diseases, often do not have the resources for monitoring diseases, and are limited in the resources they can put towards interventions.

GOSAIC is designed to scale with data and environmental/health challenges, as and when they happen. In year one GOSAIC’s focus is SARS-CoV-2. Initially one model (MaxEnt) trained on SARS-CoV-2, with environmental data as reference, will be used. In years two to three, GOSAIC will scale monitoring to other emerging global health challenges. The predictive capacity of GOSAIC from year one will increase by using more data and models. In year one GOSAIC is expected to have a minimal accuracy level of 60% at a 30 square meter resolution. By year three the accuracy is expected to improve to at least 80% with a resolution of 15 square meters. 

GOSAIC will initially be launched in English, in years two to three GOSAIC will be promoted and translated to different languages. Effectiveness of ads will be tracked by social media and IP addresses accessing the platform. This info will further help target ads and increase the user base. Our revenue will be used to enhance GOSAIC and to implement interventions that have direct impact on populations. This impact can be increased by helping to locate interventions in areas that will directly increase the health of populations.  

For improvement of GOSAIC, we will measure the number of diseases covered, accuracy of disease impact predictions, pixel resolution of analyzed imagery, collaboration on coding, and linked data. In the first year GOSAIC will only cover SARS-CoV-2, expanding to other diseases in following years, at a resolution of 30 meters and at least 60% accuracy. GOSAIC’s code ,used to analyze data for SARS-CoV-2 and other diseases, will improve with new code commits. Our project partner’s code, RasterFrames, currently has 1,406 code commits. Initially, disease data will be at the country level, however, the spatial scale of this data will improve as new publicly available data becomes linked.

For monitoring the user base of GOSAIC, we will measure number of users, location of users, and social media followers. Direct tangible impact of GOSAIC to the global community is measured through money spent on interventions, CO2 sequestered, and eventually economic impact of interventions (through incorporating economic data into GOSAIC). Astraea currently has 1,770 followers on Twitter globally. No money has been spent on interventions and there has been no tracking of CO2 sequestered yet. 

We expect barriers that include data sharing, portraying disease impact, security threats and equitably financing GOSAIC. 

1) Data Sharing In order to overcome the barrier of data sharing, GOSAIC initially only uses open/public data. GOSAIC will work with organizations to educate governments and NGOs on the importance of having their own data more widely available. 

2) Portraying Disease Impact We recognize that GOSAIC’s outputs may be a sensitive subject for certain governing entities. These entities may not want to have the impact of diseases to their own populations identified. By using publicly available data to start, legal risks are reduced. 

3) Security Various security threats (hackers, higher number of users than expected, others) put the system at risk. Technology experts will be hired to maintain security and functionality. In order to handle the amount of users, the amount of computing infrastructure will be adjusted. 

4) Equitably Financing To ensure global access, there is a free trial period. Fees after this period will reflect user location and the cost for maintenance.

Astraea, Market Economics Ltd, Biovista, Treetable LLC, Terex Maps, Lincoln University New Zealand.

The key barrier to our solution is the need for upfront investment. Setting up GOSAIC requires significant data collection, developing tools for data and image processing and classification as well as developing the online platform, and coding. These all are time and resource intensive activities. Our team consists of a group of researchers and professionals that do not have sufficient resources at their disposal for starting up GOSAIC. We would need to hire additional personnel for this purpose. The grant from the Trinity Challenge will help us finance the upfront investment needed during the start-up phase of GOSAIC. 

GOSAIC needs technology experts for maintaining it and making sure it is secure. The founding members of the Trinity Challenge have significant technical expertise that would allow for superb security and maintaining GOSAIC. 

Moreover, we expect to gain new ideas, expand our network and connect with potential GOSAIC stakeholders through the Trinity Challenge platform. We believe these would be beneficial for continuously improving our solution and sustaining it beyond the start-up phase.

Palantir, John Hopkins, Healthmap and ProMed have experience in global disease data aggregation for fine scale administrative boundaries. This data would supplement large administrative unit data. This fine scale data could be supplemented by contact tracing, i.e. Google. Search data from Google and social media data from Facebook could be used in conjunction for early detection of sick populations and possible outbreaks. For transmission of diseases, mobility data from Google, Cuebiq and MapBox would be an important input. Tsinghua has demonstrated expertise in spatial analysis with their Land Use Land Classification classifier. This expertise can be furthered through other open source GIS organizations, such as QGIS. Nextstrain can help us to further specify our models to disease variants. Hyperledger and the Linux Foundation would help to ensure open source standards are met and that we are implementing the latest technology. The infrastructure is currently hosted on Amazon Web Services, but could be migrated to the Microsoft Azure interface. The partnerships formed with these organizations can help in the maintenance and security of GOSAIC.