The Covid-19 coronavirus is wreaking havoc on the Spanish healthcare system. Technology is playing an important role in this pandemic. In regions such as the Basque Country, a tool based on Artificial Intelligence has been launched, which is capable of forecasting the needs of the ICUs a week in advance. The idea is to specify the number of beds that will be needed in the intensive care units.
Created by the Basque company Sherpa, the software has begun to be used by health services to try to foresee the needs. Something that seems almost magic, but that represents a great advantage in the fight against the coronavirus. Artificial Intelligence is being used very intensively in highly technical countries such as Singapore or South Korea with good results. Why not in Spain?
The «startup» is considered one of the ten most important companies in the field of Artificial Intelligence in the world. Xabi Uribe-Etxeberria, CEO and founder of Sherpa, assures ABC that his software is capable of forecasting seven days in advance the number of beds occupied in the ICU, segmented by hospitals and region. “Considering that Osakidetza-Basque Health Service is capable of setting up an ICU in 5 days, we give them time to prepare and two extra days of margin,” he values.
How does the tool work? «We use data from Italy and, thanks to the fact that they are two weeks ahead of Spain, it allows us to extrapolate what is happening in that country to ours. We look for similar regions and apply mathematical models together with artificial intelligence and machine learning. The results are very good, “promises the manager.
The pandemic has forced many companies to reorient their business plans to exit their production in the face of the health needs that Spain has experienced in recent weeks. Factories and firms in all sectors that have designed products to cover health services. “Medicine is not our field, but if Seat can build respirators, we can adapt and contribute our bit,” he adds.
The Basque company defends, despite the doubts that technology may have, that its predictions are reliable. “We are achieving very good results. We’re nailing it two beds up, two down. It is true that the ICUs in the Basque Country have never reached saturation in other regions, but the data we give them has allowed them to anticipate correctly, “he promises.
Artificial intelligence is as important as medicine. In Singapore or South Korea it has been applied from the beginning, since the diagnosis, which thanks to Alibaba has gone from 15 minutes to 15 seconds in the time necessary to review a tomography, to the production of the vaccine or the detection tests of the Covid-19, reducing a process from months to weeks. When the entire pandemic ends, we will realize the enormous role of Artificial Intelligence in its solution, “he values.
At the moment, the software allows to cross public data to make an estimate, although it has room for improvement. And he adds: “We are also trying to offer data on deceased and infected, although both have a greater uncertainty interval than ICU beds because they are affected by other less predictable variables.” In fact,
Uribe-Etxeberria anticipates that it will
apply the same technology for deconfusion in order to detect risk situations and sources of contagion in advance. In this way, “he says,” it will be possible to evaluate how to safely proceed with unconfinement “because” the alarm must not be lowered at any time “.
The use of user data and information has been a constant in the pandemic. Self-assessment applications and tools that include geolocation functions to monitor people infected with the coronavirus. In this case, this company’s software is limited to your region. The question is: why is it not being applied to the rest of Spain? “The problem is in the data. At the beginning we thought that the data would be more unified, but practically every region and hospital measures practically everything, from infections to the deceased, not to mention the way of exporting them to be able to use them with AI, “he says.
«Practically the data collection has to be done manually. If we had tried to serve another autonomous community, we probably would not have succeeded, we have done the work of three months in practically three weeks. Extracting data without a standard makes the model not scalable, ”he laments. .