Session 2-2:

Data-driven

Healthcare

Date: Thursday 1st October 2020. 16:00-17:30

Scientific chair:

Zoom room: 

Physical room:

16:00-16:20: Clustering a health dataset based on diagnostic co-occurrences. Adrien Wartelle, Farah Chehade, Farouk Yalaoui, David Laplanche and Stéphane Sanchez.
Abstract: The creation and structuration of a coherent territorial health care offer requires an assesment of the patient profiles of the population. Multimorbidity clustering is an effective way of studying this health demand. This paper develops and implements a new multimorbidity patterns clustering algorithm with a new co-occurrence measure to group diagnostic labels. This algorithm is novel and effective to determine multimorbidity patterns. It can be applied on a general population with better insight on the quality of the clusters.The algorithm was tested on an emergency department dataset with 124 943 visits using 153 diagnostic labels as the active variable for clustering. This study has led to 18 clusters representing 92.93% of the population with specific health problems and good measurement on the quality indicators and objective function.
16:30-16:50: An unsupervised learning approach to hospital classification. Jan Chrusciel, Adrien Le Guillou, Adrien Wartelle, David Laplanche, Sandra Steunou, Marie-Caroline Clément and Stéphane Sanchez.
Abstract: Public hospitals have known structural characteristics; however, little is known about the relationship between these characteristics and the mobility of patients between different hospitals. The objective of this study was to classify hospitals according to their characteristics and their role in the hospital network. Similar profiles could be used for epidemiological studies or quality of care evaluation. The interpretation of these results at the hospital level should take into account the particular situation of each hospital.
17:00-17:20: IoT Data Assembly and Recording. Vincent Zalc, Matthias Pideri, Dan Istrate, Carla Taramasco and Gaston Marquez.
Abstract: This work takes place in the context of increasing life expectancy involving a significant percentage of the population over 60 years. To prevent, detect and compensate frailty, we use different existing IoT sensors and self-developed sensors. In this paper, a solution allowing date assembly and recording from sensors using various communication technologies (KNX, IP, among others) is presented. The solution is based on the Constellation platform, which was adapted to our case. This work is a part of the PREDECO project, an inter UT founded project, as well as the BV2 project, an ECOS funded project.