In 2020, a study on calls to the SAMU centre 15 of the Gironde region was carried out by the University Hospital Centre (CHU) of Bordeaux and the Inserm research centre Bordeaux Population Health. On March 31, the results of this study were published and the use of artificial intelligence was essential for the automatic classification system set up by the researchers.

A monitoring tool called Transformer was set up by researchers from Inserm and the University of Bordeaux in collaboration with the Bordeaux University Hospital. This tool, used before, during and after the containment period of March, April and May 2020, observed trends in the reason for calls. This tool then allowed, thanks to an artificial intelligence, to automatically classify these patterns.

This was possible by “training” Transformer during its construction: First, the tool was subjected to a huge amount of call reports in order to train it to create these reports, and then it was trained to perform the classification tasks by “editing” it with reports that were correctly classified by a human.

As for the results of this study, it is from February 21, 2020 that calls for flu symptoms increased significantly to reach a first peak on February 28 and a second, more significant, on March 14, 2020, only 3 days before the beginning of the lockdown.

These three key dates, with a 14-day lag, are linked to peaks in daily emergency room admission rates. For stress, anxiety and chest pain, the peak occurred 12 days after the key dates. Finally, one month prior to the lockdown, calls for fainting, unintentional injuries and alcohol poisoning decreased significantly.

Translated from Une étude de l’INSERM et du CHU de Bordeaux utilise l’IA afin d’analyser les motifs d’appel au SAMU