Focus on... Guillaume Bottaz-Bosson, Théo Falquès and the treatment compliance in sleep apnea

on the October 15, 2019

The HP2 laboratory from UGA and Inserm studies the sleep apnea. Thanks to the Grenoble Alpes Data Institute’s funding, the team get two students in statistics and data science to work on the clustering of patient compliance trajectories and individual prediction of treatment compliance in sleep apnea from large national databases.
Obstructive sleep apnea (OSA) is a chronic respiratory disease with a high prevalence and is associated with a worsening prognosis, particularly for cardiovascular diseases. The continuous positive airway pressure (CPAP) is an effective treatment for OSA but it is uncomfortable and intrusive in daily life. The patient wears a mask during the night to keep the upper airways open. He/she has to wear it a minimum of 4 hours per night. The compliance with CPAP is a key prognostic factor. The aims are firstly to characterize CPAP compliance behaviors and then to be able to predict an individual compliance behavior for an incident patient case of OSA.
The CPAP homecare equipment monitors the treatment. It provides to the scientists the compliance data of 848 patients with sleep apnea. There are different profiles of CPAP compliance, see some examples on next figure.

The Grenoble Alpes Data Institute funded Guillaume Bottaz-Bosson Master 2 internship in 2018.  He worked on clustering patient compliance trajectories. Data clustering is an explanatory statistical technique used to uncover structure on a dataset by grouping objects in homogeneous groups or clusters. Many clustering methods exists and the focus is made on Hierarchical Agglomerative Clustering (HAC). Guillaume Bottaz-Bosson led a simulation study to compare different configurations of HAC algorithm and validating a clustering process. The application of the configuration which got the best simulation results provides 6 groups on real data. These groups are illustrated on the following chart, by median trajectories, and the number of subject for each group is mentioned at the top of the figure.
Last summer was the turn of Théo Falquès, student in Master 1 of statistics and data science in Université Grenoble Alpes to join the project for his internship with the Grenoble Alpes Data Institute. He used clinical data provided by the CHU Grenoble to try to predict these compliance groups. Results were not satisfying so the study must be continued, including more data such as sociodemographic variables for example.
The final goal is to predict the compliance, to support patients more likely to stop the treatment, and thus to improve the medical care of obstructive sleep apnea patients.
Guillaume Bottaz-Bosson is now continuing the project during his thesis under the supervision of Sebastien Bailly (HP2), Adeline Leclercq-Samson (LJK) and Agnès HAMON (LJK). This thesis is funded by the E-santé chair from the Fondation UGA directed by Pr Jean Louis Pépin.


Published on January 9, 2020