An Inria Chair of Junior Professor at the intersection of statistics and health

Established by the 2020 Research Programming Act and modeled on the Anglo-Saxon tenure track system, junior professor chairs are three- to six-year positions that can lead to tenure as a research director. The Inria Center at the University of Rennes is currently using this system to promote multidisciplinary work that is likely to have implications in the digital field. In 2023, the first chair was launched at the intersection of mathematics and the environment. A second has just been created, this time at the intersection of statistics and health, with the recruitment of Valérie Garès.

After completing a mathematics thesis on the statistical analysis of data related to Alzheimer's disease at Inserm in collaboration with the Toulouse Institute of Mathematics, the researcher completed two post-doctorates. The first was at the University of Sydney's Clinical Trials Center. The second was with the social health inequalities team at Inserm in Toulouse. Recruited as a lecturer by Insa, she arrived in Rennes in 2017. She also joined the statistics team at Irmar, the Rennes Institute for Mathematical Research.

In practice, the research results take the form of new packages that can be directly integrated into R or Python, open source quantitative data analysis software popular with epidemiology researchers.

Quarterly seminar

In order to gain access to real-world problems, upon her arrival in Rennes, the statistician approached epidemiologists and chemists at the School of Advanced Studies in Public Health (EHESP) and the Institute for Research in Health, Environment and Work (Irset). Since her recruitment to this CPJ, she has also launched a quarterly seminar. “Each time, we invite two speakers. One in statistics. The other in health.” These meetings take place at the University of Rennes 2, near the health schools.

Three areas of research

The connections forged over time have led to several collaborations. They have also identified three areas of research. The first is in the context of studying environmental health risks. It focuses on identifying potential pollutants present in the blood, urine, or hair of people studied from childhood to adulthood. Generated by mass spectrometry, the data has characteristics that are difficult to analyze. “To properly take their characteristics into account, we need new statistical methods.”

The second area focuses on data integration, which aims to exploit information from different data sources in order to build a single, richer and more comprehensive database. “An epidemiologist may, for example, have a registry of a few hundred people and want to enrich the registry by adding information about these individuals that is available in the National Health Data System (SNDS).”

But this file preserves anonymity. It is impossible to extract information directly from an identifier. "You have to do some matching work. In other words, compare data to find a similar profile in the SNDS. This introduces uncertainties that need to be taken into account using customized statistical methods. Similarly, researchers may wish to merge two cohorts to form a single one that would thus become richer in terms of individuals.“ But the data is not necessarily recorded in exactly the same way. And here too, ”tools are needed to harmonize the data between the different databases."

The third theme, which is still emerging, questions the ethics of artificial intelligence algorithms. "Do these algorithms make reliable predictions? Does the data contain demographic biases between people of different ages, genders, or socioeconomic statuses ?" Funded by Inserm and led by EHESP, this research draws on SNDS data to predict the risk of adverse events following revascularization after medical intervention.

The junior professorship will last three years. It should enable all these interdisciplinary collaborations to be structured, leading to the creation of an Inria project team composed of statisticians, epidemiologists, and chemists.