In April 2018, the growth charts in France’s child health and immunization record booklets (carnets de santé) were updated according to 1.5 million measurements taken from over 230,000 children. These new charts were created using a Big Data approach combined with mathematical modeling, with the ability to extract and analyze huge medical datasets. It also renders the pediatric anthropometric reference data more reliable and easy to update.
One particularly common way of evaluating a child's healthy development is to use the growth charts contained in their health and immunization record booklet (carnet de santé). When the child’s trajectory strays too far from the reference curve, the cause needs to be investigated – something which is only possible if the reference curves reliably reflect the growth of children today. Yet until recently the reference charts in use were those established in the 1970s based on the measurements of several hundred children born in the 1950s. It is evident that the height and weight of children have changed since then, as the various studies conducted by the Inserm team EARoH* bear out. In addition, the charts recently established by the World Health Organization have proven to be poorly calibrated with the growth of children in France, according to other studies. Therefore, it was necessary to conduct specific research on national data in order to update the reference charts.
On April 1, 2018, France’s Directorate General for Health (DGS) inaugurated a new record booklet with updated growth charts, the preparation of which had been assigned to the researchers from EARoH in collaboration with EPOPé**. For this, they used mathematical modeling, which has a number of advantages, as Martin Chalumeau** explains: "This approach makes it possible to utilize medical databases whose wealth of information offers greater power and representativity than a field study which would be conducted in several thousands or even just several hundreds of people." It is also a method that is relatively quick to deploy once the first model has been developed, making it "quite easy to update data, should this prove necessary".
Thanks to a partnership with the French Association of Ambulatory Pediatrics (AFPA) and the publishing house CompuGroup Medical, the researchers had access to data from the anonymized medical records of 32 AFPA pediatricians and 10 French Society of General Medicine (SFMG) primary care physicians (selected at random from the various users of the same medical software application), representative of the regions and environments of metropolitan France (urban, rural, etc.). This made it possible to extract the data of 238,102 children between 0 and 18 years of age who were monitored by these doctors between 1990 and 2018 – representing around 1.5 million height and weight measurements. After excluding outliers and children suspected of suffering from growth disorders, the researchers modeled the growth of French children, generating curves that are distinctly above the previous ones: for any given age, the weight and height of the latest generation of children are higher than those of the previous generations.
From healthy to sick children
To ensure the reliability of their model, the researchers checked that the new curves were concordant with the findings of a recent national cross-sectional survey by the French Directorate for Research, Studies, Assessment and Statistics (DRESS), in which several thousand children aged 5-6 years, 10-11 years and 14-15 years were measured in school. "We observed that our modeling is perfect in terms of height for the three age groups. Concerning weight, the data are satisfactory for the 5-6 year-olds, but less so for the older children for whom our curves appear to slightly underestimate their weight." This discrepancy could be due to two factors: "It is possible that the children were not weighed in their underwear in the school setting, which can result in a slight overestimation of their weight. It could also be that the children who are overweight or a fortiori obese are less likely to be regularly followed up, implying their underrepresentation in the sample of medical consultations analyzed in relation to the DRESS survey".
In practice, the established height curves are perfectly usable for monitoring the correct development of children. "However, in order to detect overweight and obesity, it is necessary to evaluate body mass index using the International Obesity Task Force curve which is also reproduced in the booklet" insists the pediatrician.
This research shows that the promises of Big Data are significant when it comes to public health. "Now that we have curves to describe the healthy development of children, we would like to model the specific growth of those suffering from various diseases that can affect growth, such as renal insufficiency, inflammatory bowel disease or endocrine disorders. By using new algorithms and databases specific to these diseases, we can envisage developing tools to assist early detection that would inform the primary care doctor should a child under their care specifically deviate from the growth curves."
* EARoH: Early health determinants research team
** EPOPé: Obstetrical, perinatal, and pediatric epidemiology research team
These two teams form part of unit 1153 Inserm/Université de Paris/Université Paris 13/Inra, Center of Research in Epidemiology and Statistics (CRESS)
Source : Heude B, Scherdel P, Werner A et al. A big-data approach to producing descriptive anthropometric references: a feasibility and validation study of paediatric growth charts. The Lancet Digital Health, Decembre 1, 2019. DOI:10.1016/S2589-7500(19)30149-9