Artificial intelligence is increasingly used in the field of health. This time, French researchers have used it to fight against child abusers.
- In France, more than 50,000 children and adolescents per year are victims of abuse.
- In 2018, a study showed that in France, a child dies every five days killed by his own family.
It is sometimes possible to miss it, even when you are a health professional. To better identify abused children, scientists have developed a promising new system based on artificial intelligence.
A test at the Dijon University Hospital
“We have created an algorithm using data from the Program for the medicalization of information systems (PMSI) of hospitals and clinics, where the pathologies and lesions observed on each patient are listed and coded, in order to identify hospitalizations secondary to a physical abuse in children under the age of 5”, explain the researchers in Santé Publique France.
“We carried out a monocentric study, aiming to test the relevance of the algorithm”, they continue. In concrete terms, the tool was used to identify among the children hospitalized at the Dijon University Hospital the stays suspects. Two groups were then formed, the first including 54 children most likely abused and the second 102 young people whose injuries left room for doubt. “The cross-referencing of PMSI data and medical records allowed the calculation of the positive predictive value* of the algorithm”, develop researchers.
“A promising tool”
Result: the positive predictive value of the algorithm for group 1 was 85.2% and 50% for group 2. The younger the children, the better was the predictive value, which reached 94.4% for those aged 1 month to 1 year in group 1 and 78.3% for those equally small in group 2. “This algorithm is a promising tool for identifying hospital stays for physical abuse in young children.” conclude the researchers.
*The positive predictive value (PPV) is the probability that the patient, who tests positive, is actually abused or ill.
