An algorithm barometer of child abuse

At Disney University Hospital, a team of scientists developed and validated an algorithm capable of observing changes in physical abuse of children based on hospital data.

During first incarceration, hospitalization rate for abuse + 50%

We are in the spring of 2020, during the first captivity. If this measure reduces the number of contamination and deaths caused by Covid-19, another number will increase drastically. 119 “Allô Enfance en Danger” switchboard explodes: From March to May 2020, the number of calls increased by 56%, according to the public interest group Childhood in Danger (GIPED). And indeed, hospital admissions for physical abuse of children aged 0 to 5 have also increased, Catherine Quantin, director of the Biostatistics and Medical Informatics Unit at Disney University Hospital, confirmed. Heard in the Children’s Secretariat office in September 2020 with the French Pediatric Society, their data weighed on the government’s decision to reopen nurseries and schools. “We have seen a 50% increase in the relative frequency of child hospitalization due to physical abuse”, In fact the conclusion of the scientists, in a publication that will appear a year later.

An algorithm that evaluates the likelihood of abuse

To reach this conclusion, the authors developed an algorithm based on data from the Information System Therapy (PMSI) program. This database aggregates all diagnostics and all medical activities performed in the hospital, according to the codes that allow hospitals to pay later. Although PMSI is primarily a budget objective, this database contains valuable information. Thus, if an acute appendicitis is recorded under code K 35.3, many clearly report various types of abuse. T 74.1 is translated as “physical abuse, injured child or infant” and W50 as “hitting, twisting, biting or scratching” by a third party.

One hundred codes were thus selected by a team working at Disney University Hospital to indicate a high probability of poor treatment when they were used in hospitalizing a child aged 0 to 5 years. The algorithm thus classifies children, From anonymous PMSI data, In two groups: “highly probable abuse” or “suspicious abuse”. To assess its relevance, liability for potential or suspected abuse groups was then compared with the diagnosis of field experts: forensic doctors who perform forensic skills. As a result, more than 85% of algorithm assignments in the group of “highly likely abuses” were verified by the medical records and forensic doctors’ skills! A score that justifies using it to monitor the evolution of physical abuse in France.

Abuse barometer

If researchers choose young children, their limited motor skills make it easier to detect abnormal injuries. “Falling off the change table will not cause serious brain damage such as coma, convulsions or brain contractions”, Portrayed Katherine Quantin. “This is a trembling child or further symptom of intentional aggression.” The youngest babies are the least mobile, for whom the algorithm gives the most reliable results – excluding babies under 1 month of age, who sometimes suffer the consequences of trauma during delivery. In children aged 1 month to 1 year, about 95% of the diagnoses of “highly probable abuse” of the algorithm were thus supported by forensic pathologists. “Our purpose is not to immediately trace individual suspected cases, or to determine the exact extent of abuse in France, but to be able to follow its evolution over time, this is a barometer.”, Explains Catherine Quantin. Together with her team, her work, supported by the French Public Health Agency and the National Observatory for Child Protection, will soon be able to assess new trends in child abuse two years after the epidemic.

In clinical practice and outside of statistical observation, this algorithm can be used as a precautionary measure in the end cases. Dr. Melanie Loisiu, a forensic pathologist at Disney University Hospital and who has participated in this work, is a species that she will be able to contribute. Establish future diagnoses of abuse.

Leave a Comment