On June 20, Health Insurance deployed a free platform accessible to all to view French health expenditure data.
Diabetes, lung cancer or even Parkinson’s disease, Data pathologies focuses on around fifty pathologies in its current form, based on the 1.5 billion invoices that enter the system each year on national territory.
All processed data is returned in the form of interactive infographics. Say goodbye to hard-to-read Excel tables: the ambition of Data pathologies is to reach a wide audience, such as health system players, journalists as well as patient and insured associations.
Researchers are not part of the core target, specifies Damien Vergé, director of strategy, studies and statistics (DSES) at the National Health Insurance Fund (CNAM), at ZDNet. But they can still refer to it, by virtue of the very principle of open data. The department that Damien Vergé heads works in particular for the sharing of health data for research purposes, at the same time as it focuses on the challenges of open data, of which Data pathologies is “the latest flagship” to date.
Aggregated and untraceable data
The data integrated into Data pathologies come from the medical mapping of expenses and pathologies that the Health Insurance has been updating every year since 2015. This analysis work is itself based on data from the national data system. of health (SNDS).
A certain number of parameters can be refined on Data pathologies to show results by region, or even by age group.
The data collected on the platform is untraceable, assures Damien Vergé. “It all starts with billing data. All information received is first pseudonymized and the social security number is irreversibly encrypted three times. It is the aggregated data that is returned. Thus, no identification is possible from this aggregated data. Under a dozen individuals, the data does not appear. »
“Adding value to our heritage in open data”
The last data updated in Data pathologies dates back to 2020. This lag is explained by the time required to classify, cross-check and analyze the gigantic volume of data produced each year.
Analyzes using algorithms are carried out to classify the pathologies and the expenses related to these pathologies. The algorithms used are documented in the Data pathologies methodology page. “We are in a process of continuous improvement. Every year, we re-evaluate these algorithms, ”underlines the manager.
L’Assurance Maladie’s open data approach is not new: “A lot of data is already available from the Studies and Data section of ameli.fr. There is a lot of information on expenses and pathologies. But these data are shared in the form of Excel tables, which is not at all convenient for a non-expert audience”, notes Damien Vergé.
Health Insurance has carried out a first data visualization experiment around health data to monitor vaccination against Covid-19, Data vaccine Covid. “This experience made us want to enhance our heritage in open data in a different way”, says Damien Vergé. “There is a real difference between putting data into a frozen PDF and making an interactive presentation. The reception is not the same. »
Developments based on feedback
Six months were needed to build Data pathologies. L’Assurance Maladie worked with Opendatasoft, a French company that offers data sharing software, to ensure the development of its tool. The data visualization agency WeDoData was also involved in this project to “refine the web design of the site”, specifies Damien Vergé.
The project mobilized around ten people from the CNAM side. L’Assurance Maladie carried out a number of demonstrations to regional health agencies so that they could take ownership of the platform.
If the first returns are “extremely positive” for the time being, comments the director, it is still early to judge the scope of the platform. Evolutions of the platforms are not to be excluded to improve the tool, evokes the director.