Technology to the rescue of battered children.

According to a study published on Tuesday, an algorithm could help identify child abuse more easily when they are hospitalized.

A team of epidemiologists and forensic doctors propose to rely on the pathologies listed in the databases of hospitals.

This artificial intelligence tool, effective for children under five years old, could make it possible to "identify mistreatment that is not obvious", and thus give the alert, explain these specialists from the University of Dijon, including the work is published in the

Weekly Epidemiological Bulletin

of Public Health France.

A warning device

The system is based on the use of the Program for the medicalization of information systems (PMSI) for hospitals and clinics, where the pathologies and lesions observed in each patient are listed and coded.

In the case of physical violence on young children, "the lesions are not always very specific: they can be, for example, bruises which will not necessarily challenge the professionals", especially since the emergency doctors do not know the antecedents of the young patient, noted Catherine Quantin, epidemiologist and biostatistician and one of the authors of the study, during a press briefing.

However, "if the child has come to the hospital several times", and the device identifies a repetition of "slightly strange lesions", it can give the alert, added Mélanie Loiseau, co-author of the study and specialist in forensic medicine.

Abuse is more likely if the lesions observed are of "different ages", for example if the child has a recent fracture but also sequelae of old fractures.

Encouraging results

The relevance of the algorithm was tested by a team of forensic scientists who studied in detail the files of 170 children “spotted” by artificial intelligence, to check whether indeed it could be thought that they had suffered violence.

The results were found to be approximately 85% reliable in cases where the algorithm identified “highly probable abuse”, and 50% in cases of “suspected abuse”.

The younger the child, the more reliable the result, because "before one year, the baby remains in its cradle, so if it is injured, it is not that it has hurt itself", summarized Mélanie Loiseau.

Our file on abuse

This innovation can be useful to improve the statistical perception of physical violence against children, but also "from an individual point of view" to ensure that suspected cases are "studied by doctors who are used to child abuse “, and thus facilitate reporting to the judicial authority, according to the doctor.

Company

Abused children: A Toulouse child psychiatrist suspended after making a report

Culture

Violence against children: "I no longer consider myself a victim, I am appeased", says Mohamed Bouhafsi

  • Abuse

  • Child

  • Hospital

  • Artificial intelligence

  • Company