Researchers are studying how artificial intelligence could be used to plan scenarios for future outbreaks.
What if you could more quickly interpret large amounts of health data to predict how long a patient would be hospitalized, or introduce human behavior into an epidemic model to determine the possible scale of a viral outbreak?
These are some of the ways researchers are testing new artificial intelligence (AI) models to better plan for future viral outbreaks such as “disease“, an unknown pathogen that could trigger an epidemic similar to that of Covid-19.
“One of the strengths we see with AI-based approaches to analyzing large data sets is really the ability to identify early signals of potential anomalies in public health”explains to Euronews Next, Alain Labrique, director of the digital health and innovation department at the World Health Organization (WHO).
“I think there are many different ways to use an advanced computing tool like artificial intelligence to improve the way we detect new outbreaks and pandemics, but also to respond to those outbreaks and pandemics,” he adds.
However, he added that to strengthen the models it was important to address bias and provide them with good data, not just from a specific population. This is an area where research is growing, but in practice some of these new models may take time to establish.
Disease severity and hospital capacity planning
Researchers at Yale University in the United States recently published a study which addresses one of the many challenges that have arisen during the Covid-19 pandemic: how to manage saturation in hospital services.
“The number of hospital beds is limited and if you have a pandemic like [Covid-19], you need to prepare yourself. We take a public health perspective. We want to be ready if something happens.”tells Euronews Next, Vasilis Vasiliou, president of the department of environmental health sciences at the Yale School of Public Health.
Their epidemic model uses an AI-powered platform to triage patients by predicting disease severity and length of hospital stay.
It is based on clinical and metabolic biomarkers which, according to researchers, can indicate the progression of the disease.
According to Vasilis Vasiliou, in the event of a viral epidemic, it would be a matter of integrating the first data into an algorithm powered by AI to determine how to better organize the hospital’s resources.
“If something happens very quickly, you have a framework, a model, an algorithm that you immediately feed into [avec] the first data from the first country where this happened. You can then start to develop a new model.”he explains.
According to him, one of the current limitations is the lack of data. “For each AI model, the more data there is, the fewer limits there are”he assures.
Kirill Veselkov, co-author of the study at Imperial College London, said that with an emerging disease, there is a need to find new biomarkers that can influence its severity.
“Current state-of-the-art analytical tools will be able to measure hundreds of thousands of these biomolecules”says Kirill Veselkov.
“If we want to analyze them, human doctors probably won’t be able to do it without resorting to sophisticated mathematical algorithms, and AI is particularly suited for this, to identify the pattern or set of biomarkers and associate them with the disease process and its outcomes”he adds.
But their model will need to be further studied with more populations, taking into account comorbidities and other factors, before it can be generalized for the general public.
Use AI to know when to close doors
For Covid-19, a virus we already have information about, AI can help with hospital scheduling, according to Rachel Dunscombe, a member of the UK AI council and current chief executive of OpenEHR.
“What we have is a set of data on the ground that tells us the real situation and we need to know if we need to intervene, if we need to lock down, if we need to increase the capacity of the systems, if we need to, you know, reduce elective activities to make room”Rachel Dunscombe, who is also the former chief executive of the NHS Digital Academy, tells Euronews Next.
“We can actually use A.I. [dans la planification des soins de santé] to determine the appropriate time to lock down, put masks in place, you know, put additional staff in place to reduce the activity that we do day to day”specifies the former general director.
She said the UK felt better equipped to use models to assess the impact on the ground of certain scenarios, post-Covid-19.
“If given the right data and supervised in the right way, it will give us the likely results.”she added.
Difficult to represent human decision-making
Researchers at Virginia Tech in the US are trying to solve another epidemic modeling problem using AI: how to accurately represent the complexities of human behavior during a viral outbreak?
“In traditional modeling, you have to represent human decision-making in some way, which is difficult to do”, explains Navid Ghaffarzadegan, associate professor at Virginia Tech, to Euronews Next.
“The reason is that human beings are complex. Societies are difficult to predict. With better or different ways of representing humans through AI, you now have the ability to see how they react in different scenarios, and you have models that integrate human behavior”he added.
As part of their studywhich is currently in preprint, researchers modeled an outbreak in a town named Dewberry Hollow with a fictional virus called Catasat, to avoid possible bias when using ChatGPT.
They studied how humans deciding whether or not to stay home could influence the epidemic model by providing a scenario and the personality characteristics of different so-called “agents.”
They found that these humans with generative AI imitated in the simulation “actual behaviors such as quarantining when sick and self-isolating if cases increase”.
The multiple waves of the virus were similar to waves seen in previous pandemics, which resulted in the virus becoming endemic in society.
The main disadvantage of this model is that it is expensive and time-consuming to implement, but it is expected to improve as artificial intelligence develops. Others say their model still needs to be validated.
The future of AI and pandemics
In another article published Last year, Navid Ghaffarzadegan highlighted the difficulties of predicting the trajectory of an epidemic using traditional models and AI models. He found that AI models were not necessarily performing better, but that this was partly due to changes in human behavior.
Some argue that there is still little research evaluating AI performance during the Covid-19 pandemic.
A article published in 2021 in the journal Frontiers in Medicine analyzed 78 studies investigating the use of AI during the pandemic.
The use of AI included AI-assisted diagnosis for Covid-19, epidemic prediction as well as drug development, such as rapid identification of drugs or products that could neutralize the disease.
They concluded that it was a potential tool in the event of an outbreak, but that further research in this area was needed.
Vasilis Veselkov said the AI triage study was in the research and development stage, but it would still take time before these AI models could be used to plan for future outbreaks, like this one. of an unknown pathogen that the WHO calls “disease X”.
“We really need to develop the tools, but also think a lot, especially when it comes to healthcare applications, pandemics, population level applications. We need to think about security and the robustness of the solution, as well as its limits”he concludes.