The Panel selected ten projects which, apart from their scientific quality and benefits in terms of improvement of the healthcare system, will contribute to the first version of the catalogue of data shared by the Health Data Hub.



Identifying the best sarcoma treatment regimens

A highly complex and varied form of cancer, the sarcoma is a category of rare tumours for which the effectiveness of clinical trials has reached its limits. Randomised studies have been carried out for almost 40 years now without any conclusions being reached as to the interest of chemotherapy in treatment of sarcomas, or on the pertinence of using it before or after surgery. Led by Professor Jean Yves Blay, Director-General of the Léon Bérard Centre and Director of the NetSarc network, alongside his teams and experts at the Bordeaux Centre for the Fight against Cancer and the University of Rennes, the Deepsarc project aims to study the impact of various treatments on real-life data with a view to identifying the most appropriate therapeutic regimens. It will make use of clinical data from the NETSARC database, an almost exhaustive reference base on sarcoma in France, including close to 50,000 patients. This data, along with medico-administrative data from the National Health Data System (SNDS), will enable patients to be monitored over time.


Evaluating and improving care pathways following myocardial infarction

Proposed by Doctor Axelle Menu-Branthomme, Head of Department at the Île-de-France Regional Health Agency (ARS), with the support of SAMU 78 and the Health Cooperation Group (GCS) Sesan, the project seeks to characterise care pathways for patients in Île-de-France who have suffered acute myocardial infarction. This frequent and often fatal pathology can result in serious complications (residual heart failure, for example). Adding SNDS medico-administrative data to the eMust register, which provides details of each acute event when it occurs, will enable evaluation of various types of care pathways in terms of quality, safety and pertinence.


Sending prescribers warnings on dangerous drug interactions

Prescription safety tools incorporated into prescription assistance software generate large numbers of warnings due to almost systematic detection of drug interactions with possible adverse effects. The proliferation of such alerts wears down prescribers’ vigilance and they are paid increasingly less attention to. Current gradation of warnings takes account of the seriousness of their potential consequences but not of the frequency of the complications in question. Led by Jean-François Forget, Medical Director of VIDAL, the project aims to make use of SNDS data to estimate the real prevalence of complications resulting from drug interactions, in order to identify warnings that need highlighting due to their impact. It will then be possible to draw physicians’ attention back to the most pertinent warnings, which should lead to reduction in occurrences of serious, frequent and foreseeable complications.


Predicting heart failure crises in patients with pacemakers

Heart failure causes almost 5% of hospitalisations in France, for a total cost of 1.8 billion euros in 2016. Prediction of episodes of aggravation therefore constitutes a real challenge, which Doctor Arnaud Rosier, President of the Implicity startup, means to take up. His project seeks to develop a solution enabling prevention of heart failure crises in patients fitted with pacemakers / defibrillators, via analysis of telemonitoring data by artificial intelligence. The Health Data Hub will enable data from over 8000 devices worn by patients in intermediate care to be cross-referenced with SNDS data with a view to coming up with predictive models without having to label episodes of interest manually.


Predicting individual trajectories of patients with Parkinson’s disease

Jean Christophe Corvol, Professor of Neurology at La Pitié-Salpêtrière Hospital, with support from ICM, INSERM and F-CRIN, is promoting this project, whose main aim is to provide neurologists with a tool for predicting individual trajectories of patients with Parkinson’s disease, so that they can implement appropriate preventive measures. The data provided by the large cohort of 20 000 individuals monitored in Centres Experts Parkinson will be chained with SNDS data in order to enable understanding of patients’ overall treatment. Computational and artificial intelligence approaches will be implemented in order to model the progression of the disease.



Measuring and understanding patients’ real out-of-pocket costs

Promoted by Laurent Borella, Health Director at Malakoff Médéric Humanis, a non-profit mutual insurance company, the ARAC project aims to combine Assurance Maladie (SNDS) data with data on complementary reimbursements in order to calculate patients’ real final out-of-pocket costs in the French social protection model. These days, such OOP costs are either estimated via surveys or imputed. Exact measurement of such costs at individual level will be of great value to health economists. Among other things, they will be able to question the ways in which healthcare expenditures are funded, as well as shed light on questions of purchasing power and making money available for recourse to healthcare, etc.


Quantifying the proportion of patients suffering from undesirable effects

In 2017, almost 82,000 undesirable drug effects were declared for 12,000 medicaments on the market. Declarations of undesirable effects should be interpreted taking account of a given product’s uses as well as numbers of patients treated. Even though it has its limitations, the declaration rate (number of undesirable effects declared compared with the number of patients exposed) is an estimation necessary to the authorities, health professionals and patients alike in order to assess the importance of each declaration. It is currently calculated upon request and “by hand”. Promoted by Patrick Maison, Scientific Advisor at the National Agency of Medicine and Health Products Safety (ANSM), the project’s main goal is to develop a tool to automate calculation. To this end, it will draw on the ANSM’s national pharmacovigilance database, data on reimbursed ambulatory and hospital care prescriptions (SNDS) and the characteristics of products in CODEX (Marketing Authorisation repository).


Mobilising emergency department data in order to improve health monitoring

Promoted by Yann Le Strat, Director of the National Public Health Agency’s Support, Data Processing and Analyses Division (DATA), with the assistance of FEDORU, the Brest Stroke Registry’s teams and the University of Rennes’ REPERES team, the project proposes creation of a database chaining all data from the Oscour® emergency monitoring network with SNDS medico-administrative data. The Oscour® database has been compiling summaries of emergency department admissions for almost fifteen years, with a total of over a hundred and thirty million emergency department visits identified. A cohort of this size will enable examination of a whole range of questions with a view to improving health monitoring in France. An initial experiment is set to be carried out on strokes, the most common cause of death among women and third most common cause among men.


Evaluating artificial intelligence’s contribution to organised breast cancer screening

With an estimated 11,883 deaths in 2017, breast cancer kills more women in France than any other form of cancer. Early screening results in a 21% fall in the mortality rate. Francisco Orchard, Head of the Data Science Unit at Epiconcept, with the support of his team, Occitania’s Regional Centre for Coordination of Cancer Screening (CRCDC) and the Curie Institute, proposes to evaluate the impact of integrating artificial intelligence into the organised breast cancer screening system. To do so, he will be drawing on the e-SIS breast cancer screening database created by the départements of Gard and Lozère, which includes over 250,000 annotated images, adding medico-administrative data with a view to reducing numbers of false negatives. This extensive base should enable creation of successful models of convolutional neural networks on data never previously used for such a purpose.


Measuring the long-term impact of kidney-graft patients’ exposure to immunosuppressive drugs

There is still too little information on the connections between patients’ exposure to drugs and long-term effects, both generally and in cases of organ transplant. In the context of lifelong treatment, knowledge of such relationships would enable optimisation of therapeutic strategies, as well as the dosages and composition of such drugs. Promoted by Doctor Pierre Marquet of Limoges University Hospital, with support from INSERM and Optim’Care, the Rexetris project will be studying the connections between exposure to immunosuppressive drugs and the long-term future of kidney-graft patients and the grafts themselves. Work will make use of the Biomedicine Agency’s Cristal database and the Limoges University Hospital’s ABIS database.

© 2019 Health Data Hub

Hébergé chez