Algorithm for identifying physical abuse of young children

To identify potential or suspected cases of child abuse, the latest epidemiological bulletin from Public Health France reveals that researchers from the Forensic Unit at Dijon University Hospital have created an algorithm based on data from the Medical Treatment Program Information Systems (PMSI). If validated, this algorithm will provide a monitoring tool for these children.

Underestimated child abuse
Epidemiological data on child maltreatment are inaccurate and underestimated. A 2009 study showed that, on average, 10% of children in high-income countries were abused or neglected. “The National Child Protection Observatory estimates that in 2019 at least 94 minors died within the family due to violence and/or murder,” Can we read the latest epidemiological bulletin of public health in France. PMSI data is an important source of information for documenting cases of physical abuse of children. What emerged from the study is that hospital stays for children after abuse are not coded enough Voluntary physical assault or traumatic injury.

Algorithm for identifying child abuse among children aged 0-5 years
To overcome this underreporting, an algorithm was developed at the Center University Hospital (CHU) in Dijon 13, in order to identify children aged 0 to 5 years, who benefited from hospitalization due to lesions that may have cascaded with physical abuse., in the latest BEH. Public Health France notes that identification in this age group is easier than identification in older children.

Distribution according to the likelihood of abuse
The algorithm classifies children into two groups:

The first group of ‘highly likely abuse’ consisted of children presenting at the time of hospitalization with traumatic lesions described as intentional. The files considered are between 2008 and 2019
– Group 2 “suspected abuse” consists of children who at the time of hospitalization present “traumatic lesions that appear suspicious of physical abuse, by their characteristics or number, without being sufficient to be included in the first group. The files considered are from 2013 to 2019.

Study of traumatic injuries by forensic physicians
The medical records of these children were studied by a forensic pathologist between November 2020 and April 2021. Then, a group of three forensic pathologists studied the injuries of the child and the mechanism of the patient-reported injury. NB as a cause of injury. In each case it was determined whether the alleged mechanism in the medical file was compatible with the lesions presented by the child. This made it possible to determine whether the traumatic lesions of abuse were confirmed, highly suspicious, suspicious or excluded from the medical file as a traumatic etiology.

Comparing the algorithm with the data of forensic scientists
After applying the maltreatment identification algorithm to children, they were divided into two groups: high probability of maltreatment and suspected maltreatment. Forensic specialists also categorized the stay in the Children’s Hospital into two groups (high probability of maltreatment and suspected maltreatment) or exclusion of maltreatment.

Good predictive value of the algorithm to identify highly probable child abuse
The algorithm developed is a promising tool for identifying physical abuse of young children during hospital stays. In fact, the positive predictive value of the algorithm for identifying highly potential cases of abuse appears to be greater than 80%, regardless of the age studied (between 0 and 5 years). The study suggests that the algorithm detected abuse better than the case study of 0-1 year olds, which the researchers say is not surprising because “Traumatic injuries, particularly fractures in a child, before they reach the age of walking (ie around 12-18 months) or able to move on their own (ie around 9 months), are highly suspected of being physically abused.” The authors also determine greater reliability of the algorithm when excluding very young children (<1 month), the age at which “It is very complex to differentiate obstetric injuries from secondary injuries and physical injuries. These two conditions can actually be a cause of traumatic injuries.”

Predictive value of abuse to be refined for suspected cases
The identification of suspected cases of abuse still needs to be refined, even if it appears to be usable for children 1 month to 1 year old.

Babies born prematurely are at greater risk of child abuse
The study found that, in both groups, children with a prior medical history of hospitalization due to maltreatment were overrepresented. In particular, two-thirds of them were born prematurely. This over-representation of sick children especially children born prematurely could explain the disruption in early parent-child interactions, due to increased medical care in the aftermath, separating the infant from his family circle. Furthermore, like disability and disease, prematurity has been identified by the World Health Organization as a risk factor for child abuse.

“A larger-scale study should be conducted to confirm these observations, in addition to assessing sensitivity, in order to see if actions should be considered to improve data collection on physical abuse. All of these steps seem necessary before considering applications in the daily practice of this algorithm for identifying physical abuse. in young children. Described in BEH Public Health France.

Source BEH “Positive predictive value of an algorithm to identify physically abused children, ages 0-5”

Leave a Comment