In France, one third of emergency room visits are due to trauma. In order to better understand their mechanisms and improve their management, researchers from Inserm and the University of Bordeaux at the Bordeaux Population Health research center, together with teams from the Bordeaux University Hospital, have developed an algorithm capable of classifying emergency room visits due to trauma by analyzing clinical reports through artificial intelligence (GPT). The performance of this project, named TARPON [1] , which reaches 97% accuracy, has been published in the Journal of Medical Internet Research Artificial Intelligence . The results suggest that a national trauma observatory will soon be set up.
Trauma accounts for 9% of mortality in France and often affects young people. More than a third of the 21 million visits to emergency rooms each year are due to trauma. It is therefore a major public health problem that represents a significant health, societal and economic burden, to which scientists are working to provide solutions.
The idea of the TARPON project, led by researchers from Inserm and the University of Bordeaux, in collaboration with the Bordeaux University Hospital, was based on the observation that for each visit to the emergency room, the caregivers write a report. These reports are a mine of information: description of symptoms, condition of the patients, as well as many details on the circumstances of the trauma.
However, these data remain unexploited today, and there are few statistics on victims of accidents in everyday life, violence or attempted suicide. In the field of road accidents, an observatory exists but it is only complete for mortality and most accidents related to cycling, walking or scooter riding are not included. An analysis of anonymized information from emergency room reports would provide the basis for an almost exhaustive trauma surveillance system.
These reports are unstructured texts written with a mixture of common terms but also medical, technical, and abbreviations. In order to extract interesting information from them, without having to read them all, the research teams developed an automatic language processing technique based on artificial neural networks. The researchers adapted the GPT artificial intelligence model and trained it with a sample of more than 500,000 reports from the adult emergency room of the Bordeaux University Hospital [2] . The result is a French clinical language processing tool that complies with RGPD rules.
With the support of the Health Data Hub, Bpifrance, the Nouvelle Aquitaine region, the Agence nationale de sécurité du médicament et des produits de santé (ANSM) and the Délégation à la Sécurité Routière, the researchers were able to finance the purchase of a powerful server, dedicated to artificial intelligence and installed within the hospital. The latter has allowed the implementation of the algorithm developed by the scientists, to automatically classify the traumas according to their types, and this with a surprising accuracy.
Indeed, the method developed by the researchers allows 97% of the reports to be classified correctly (compared to 86% with the old methods), as the researchers detail in their scientific article. Thanks to this first step, the study of the data will be able to begin on the Health Data Hub’s technological platform by the summer.
These results pave the way for the establishment of a national trauma surveillance system, but also for epidemiological analyses of the impact of drug consumption on the risk of accidents, for example. This work should therefore shed new light on important public health issues. In the immediate future, the TARPON project will be extended to some 15 emergency departments throughout France.
[1] TARPON: Automatic Processing of Emergency Room Responses for a National Observatory
[2] This research meets the obligations of the General Data Protection Regulation (GDPR).
