Artificial intelligence for the evaluation of profiles, trajectories and management of vulnerability in health: COVID-19, frailty of the elderly and cancer
ELCAPA-EDS
Care trajectories of elderly cancer patients and associated clinical-biological profiles
Underway
15/01/2022
-
15/01/2027
Main objectives
This project, supported by the AI-Raclès Chair, is based on linking the ELCAPA clinical cohort with the AP-HP Health Data Warehouse (EDS). It has three main objectives:
- Identifying and characterizing the hospital care trajectories of elderly cancer patients and the associated clinico-biological profiles,
- Assess the potential impact of hospital care trajectories on patient survival, taking into account oncological characteristics and initial management,
- Identify and predict frailty factors using data from Electronic Health Records (EHR).
- Identifying and characterizing the hospital care trajectories of elderly cancer patients and the associated clinico-biological profiles,
- Assess the potential impact of hospital care trajectories on patient survival, taking into account oncological characteristics and initial management,
- Identify and predict frailty factors using data from Electronic Health Records (EHR).
Challenge
More than a third of new cancer cases occur in people aged 70 or over. The physical and mental capacities of these people vary widely, some being comparable to those of younger subjects (robust elderly), while others, known as "frail", present an increased risk of adverse events (falls, fractures, loss of autonomy, unplanned hospitalizations, death). Comorbidities, polymedication, frailty and the lack of evidence-based data make therapeutic decision-making in elderly cancer patients complex.
Methodology/Technology used
Clustering techniques will be implemented to characterize hospital care trajectories. The potential impact of trajectories on survival will be assessed using causal inference methods. Finally, Automatic Language Processing (ALP) algorithms will be developed to identify frailty elements present in an unstructured way in EHRs.
Publications
Contacts
Etienne Audureau () & Florence Canoui-Poitrine () & Thomas Guyet ()& Laurent Le Brusquet ()& Arthur Tenenhaus ()& Charline Jean ()
Members
AP-HP, Centrale Supélec, Inria
Useful links
https://www.bernoulli-lab.fr/project/chaire-ai-racles/
Other projects
Description
Learning a deep representation of patient records for event prediction and patient segmentation
Names of partners involved
AP-HP, Inria