In terms of major innovations impacts, seven predictive models have been developed based on four state-of-the-art modeling techniques ranging from Compartmental, Agent-based, Discrete-Event to Deep Neural Networks. Particularly. Regarding the EWS, the former extends on the service provisioning scheme of similar epidemic alerting and warning tools, such as the WHO Global Alert and Response platform, by introducing Artificial Intelligence and Deep Learning technologies that identify patterns from historical data and then perform pattern matching on new measurements to identify early signals of high-impact incidents. Examples of these is using historical patterns to predict the impact of an outbreak to the society, by calculating the predicted cases and deaths, based on previous outbreaks, and to the overall health care system, by assessing how the underlying infrastructure (e.g. ICU beds, relevant medical resources) will be impacted by such events. This can be crucial, as early predictions can allow authorities to plan on how to tackle these events, and reallocate resources. In order to produce these predictions a set of ML algorithms have been applied. Such as: Temporal Fusion Transformers and Long Short -Term Memory. Finally, by using ancillary sensing methods, such as the analysis of the viral content in water drains, the EWS can provide additional insights, increase the sensing accuracy and sensitivity and identify blind spots such as the presence of large numbers of asymptomatic cases.
Substantial findings and results have been additionally generated from WP6. Specifically, proprietary bioinformatics pipelines were employed in order to discover novel biomarkers, design new primer sets to cover all the pathogens under investigation.
Moreover, during the period of M01-M12, the identification of a novel strain of SARS-CoV-2 emerged.
While ethical impact is ongoing and expected to be more concretely engaged in the trials, at this stage we have been able to engage collaboratively with end-users and solution developers to identify potential ethical harms and benefits from the project. For instance, for the STAMINA models to provide insights at a granular enough level for the necessary decision-making, the data inputted into them needs to be demographically disaggregated. However, a consequence of that is assigning socio-cultural characteristics to predictions, potentially leading to stigmatization of segments of society.
