On-going projects as PI or CO-PI
​
Several projects on data mining & healthcare use and cost :
​
"A business process management method to model and predict individuals’ healthcare trajectories"
​
Investigator: Benjamin Dalmas (DGEPS), Alexandra Langford-Avelar (PhD, DGEPS) and Juliette Duc (PhD, DGEPS)
The worldwide aging population raises serious concerns in terms of healthcare organizations as older adults can face several health problems, increasing their healthcare use and hampering their life quality. Research on healthcare trajectories (HTs) could help anticipate patient needs and better organize services around them. However, “hand-made” HTs can be relatively disconnected from reality and difficult to generalize. In a system under pressure with overstretched professionals, it is relevant to seek automated methods based on reliable data.
​
We propose to use Process mining (PM), a recent method used to model processes in various organizations. The benefit of PM in our context is to use data generated by the interaction between individuals and facilities to reflect the field reality, its bottlenecks, deviations and "unnecessary" resource consumption. This could be a step toward a learning system. However, it is still only used to describe one or few processes, within a single facility. We expect this study to support the adaptation of the healthcare system facing the growing demand. Prediction of future HTs will provide additional leverage to guide patients through the healthcare system and thus improve the quality of care and resource allocation
​
%-%-%-%-%-%-%
​
"Needs, access and healthcare utilization for over 65-aged people in Quebec's region:
a machine learning and spatial analyses application"
​
Investigator: Juliette Duc (ESPUM), David Buckeridge (McGill)
Collaborators: Erin Strumpf (McGill), Claire Godard-Sebillotte (McGill), Jean Noel Nikiema (ESPUM)
A world major demographic change is in progress. By 2050, over 1 in 4 people living in OECD countries will be aged 65 and over. This global ageing raises concerns for the healthcare system, as the needs of the ageing population increase and become more complex. It appears challenging to provide adapted services given the disparities in needs and healthcare offer across regions, which influence access to and utilization of services. These findings are even more striking in rural areas such as Quebec.
​
This project aims to describe the interrelations between needs, access, and healthcare utilization for over 65-aged people in Quebec. We will exploit ministerial data of healthcare institutions, as well as data from the TorSaDE (Trajectoire de Soins-Données Enrichies) cohort, which includes participants in the Canadian Community Health Survey, combined with their administrative health data. Using machine learning and geographical tools, we will study how individuals' needs, access and use of healthcare vary according to the administrative health regions, urbanization, deprivation level, and the healthcare offer available in the individual's environment.
This project will then provide a comprehensive description of variations in healthcare access and use, informing decision-makers of the determinants of equitable access to healthcare in Quebec. Finally, it will illustrate the value of combining AA and GIS tools to take advantage of the massive data already collected.
​
%-%-%-%-%-%-%
​
"Monitoring user’s trajectories within the health system to improve service provision :
an application of the process mining methods"
​
Investigator: Alexandra Langford-Avelar, Juliette Duc, Benjamin Dalmas
Health facilities performance is measured using indicators calculated from administrative health data (AHD). Their performance is assessed independently of each other and it’s almost impossible to have a joint view of multiple facilities performance as a whole. The implementation of an automated tool that integrate AHD of all installations and facilities would paint a more accurate portrait of performance, but also target the factors limiting the fluidity of its services and benefit patients’ care.
​
We propose to apply process mining (PM) methods to the AHD collected by the facilities in the CIUSSS de l'Ouest-de-l'Île-de-Montréal (COMTL). The PM methods consists in 1) gathering, cleaning and merging all the data from the different services and facilities of the COMTL, 2) generating an event log of all these activities, and 3) modeling and displaying the patients’ trajectories to measure performance and optimal resource utilization. We are taking advantage of the fact that each contact between a user and the system is registered into AHD, such as an emergency episode, an hospital admission, or a medical consultation.
​
Responding to the COMTL's request to model their users’ trajectory to assess services fluidity, we will establish the first automated tool modeling the processes of multiple healthcare facilities. This will allow to detect deviations mobilizing more resources in order to be able to make rapid adjustments to service provision
​
%-%-%-%-%-%-%
​
"The characteristics and perceived effects of an innovative model of home care: a mixed case study"
​
Co-PI : Anne Bourbonnais (FSI), Jacqueline Rousseau (Médecine)
Investigator: Éveline Gaillardetz, Jérôme Leclerc-Loiselle
Collaborators: Hélène Lefebvre, Saeed Ahmadiani, Laura Blais, Amélia Lamontagne, Jesse D'eramo
​
%-%-%-%-%-%-%
​
"Healtcare use and cost of people aged 65 and over during their last year of life"
​
Investigator: Erin Strumpf, Roxane Borgès Da Silva, Aude Motulsky
Collaborators: Nicolas Sirven, Juliette Duc
​
On-going project as investigator
​
“Evaluation of occupational co-exposures to multiple chemical agents with the Spectrosome approach”
​
Supervised by Philippe Sarazin (IRSST), Vikki Ho (CRCHUM) and Jérôme Lavoué (CRCHUM)
Started as IRSST guest postdoctoral fellow, continue as Investigator.
Montreal, Quebec, Canada
​
This project aimed to adapt the method developed during my doctorate to occupational hygiene data, in order to describe the co-exposure to multiple chemical agents.
​
The need to take into account the complex multitude of exposures occurring in the workplace is growing. A portrait of co-exposures common in the workplace could lead to better hazard assessments and the development of more tailored prevention strategies. However, the availability of data for characterizing co-exposures at the workplace poses a practical challenge, while the complexity in describing patterns of co-exposures remains a methodological one. Even when data is available, complexity in describing patterns of co-exposures and the large number of potential combinations of agents remains a methodological challenge. We adapted the Spectrosome method and applied it on exposure measurements archived in the IMIS occupational exposure databank in the U.S since 1979. We defined a ‘workplace situation’ (WS) as a job title within a company within a year, and co-exposures were identified by looking at measurements within each WS. The Spectrosome approach offered a visual representation of co-exposures, which facilitates identifying and characterizing combinations of exposures. Our results can be used to flag situations where the calculation of additive exposure indices might be envisaged for agents with similar effects, as well as a source of information to target research on toxicological interactions.
Now, this project is under revision to be financed by IRSST as a bigger project to draw the global picture of the occupational co-exposure in Quebec using different sources of data such as the Quebec databank LIMS and different methods inspired by the big data and financial fields (Frequent Itemset Mining). Results obtained on the LIMS will be enriched and compared with data analysis on Canadian, French and American databases.
​
​
“EXPERIVA Study: Exposures in the Peace River Valley”
Collaborative work with Marc-André Verner (PI) and Élyse Caron-Beaudoin (co-PI);
Montreal, Quebec and Toronto, Ontario, Canada
EXPERIVA: Exposures in the Peace River Valley, explores the health effects associated with hydraulic fracturing. Hydraulic fracturing typically involves the drilling of wells vertically and then horizontally in the natural gas reserve, and the injection of large volume of fracking fluid (water, sand and various chemicals) to fracture the rock formation, freeing the trapped natural gas. Northeastern British Columbia is an area of intensive unconventional natural gas exploitation by hydraulic fracturing, or fracking. The region sits on an important source of natural gas, the Montney Formation. Approximately 30,000 wells have been drilled so far in Northeastern British Columbia. Information on impacts of hydraulic fracturing activity is limited, but recent literature highlighted the risk of environmental contamination. Some chemicals used or associated with hydraulic fracturing may contaminate the soil by accidental spills, leaks, or during disposal of hydraulic fracturing fluids. It is also known that hydraulic fracturing operations can release volatile organic compounds such as benzene, as well as trace elements naturally occurring in the rock formation. Many of these chemicals are known or suspected reproductive and development toxicants, carcinogens, endocrine disruptors and respiratory irritants.
Communities and First Nations in Northeastern British Columbia raised concerns about the health effects associated with this industry. Dr. Caron-Beaudoin and her colleagues developed research projects using diverse methodologies such as exposure assessment and toxicology to provide communities the support needed to address these concerns. Their current study, entitled Exposures In the Peace River Valley (EXPERIVA), uses a multifaceted approach allowing communities living in close proximity to fracking wells to
-
obtain information on levels of contaminants in their air, drinking water, and bodies (through urine, hair and nails measurements);
-
gain access to public health experts with whom they can discuss their exposure and potential health effects;
-
contribute to study design and participate in the dissemination of study results.
From: https://www.utsc.utoronto.ca/healthsociety/experiva-study-exposures-peace-river-valley
On-going project as collaborator
​
“Use of French medico-administrative databases for hypothesis generation
regarding occupational risks of agricultural workers”
Collaborative work with Vincent Bonneterre (EPSP) and Olivier François (BCM)
Grenoble, France ; Montreal, Quebec, Canada
I started on this project as my main project and as a principal investigator (research in a postdoctoral position). I supervised the Ph.D student who is currently working on it and I became a collaborator after my arrival in Montreal.
​
This ambitious project, following a request made by the Scientific Council of the Mutualité Sociale Agricole (MSA), the dedicated health insurance system of the agricultural population, and supported by the National Agency for Sanitary Security of the environment, food and nutrition (ANSES), concerns the entire French agricultural population and represents millions of data. Individuals affiliated with the MSA are farm managers, employees working in farms, foresters, gardeners and so on. On an indicative basis, the population covered but the MSA insurance fund in 2016 represented more than 3 million individuals, including about 1.2 million active workers, the remaining concerning retirees, spouses and children.
​
The main objective of this feasibility study was to evaluate the interest of using massive medico-administrative databases of Social Security systems for vigilance in Public Health. In other words, the aim was to evaluate the possibility of generating health surveillance signals from massive insurance. This project, still ongoing, was planned over the long term (4 to 5 years minimum), bringing together different actors and specialists (modelers, biostatisticians, pharmacists, clinicians, agronomists engineers ...) around different issues.
Different sub-projects are performed simultaneously to explore the different facets of the medico-administrative data:
-
Unsupervised analysis to estimate the fixed effect between a disease and an occupational activity, taking into account of latent and unknown factors, example of the long-term diseases.
-
Development of an automated algorithm for the retrospective estimation of pesticide used by agricultural workers, based on medico-administrative data and external data about agricultural practices (census, crop-exposure matrices
-
Identification of diseases (other than recognized long-term diseases) based on the health care (medication)
​
After the first two years of set up and exploration, ANSES has renewed the funding for two additional years.
​
​
​
"Perinatal Exposure to Perfluoroalkyl Substances (PFASs) and Polybrominated Diphenyl Ethers (PBDEs)
and Weight Gain Trajectory in Children:
Data from the Maternal-Infant Research on Environmental Chemicals (MIREC) Cohort"
​
Supervised by Maryse Bouchard (Ecole de Santé Publique, UDEM)
Started as part-time postdoctoral fellow, continue as Collaborator
Montreal, Quebec, Canada
​
The objective is to determine whether high concentrations of PFAs and/or PBDEs are associated with anthropometric changes during children growth. We used data from the Maternal-Infant Research on Environmental Chemicals cohort (MIREC), which lists more than 1900 mother-child pairs, for which, among the many questionnaire-related information, blood samples were taken during the first trimester, pregnancy and anthropometric measures taken at birth, at 6 months and during their growth (between 3 and 5 years depending on the children). This project is funded by the CIHR.
Former projects
​
“Observational Surveillance Development”
​
Supervisors: Dominique J. Bicout (EPSP) and Régis de Gaudemaris (EPSP and Grenoble teaching hospital)
Master degree and Ph.D - Ministerial grant
Grenoble, France
​
The National Network of Vigilance and Prevention of Occupational Diseases (RNV3P) is an expert network of occupational physicians and volunteers specialists from teaching hospitals, supported by the ANSES. The network records in a systematic and standardized way cases of Occupational Health Problems (OHP), defined as a diagnosed disease and a set of exposures potentially responsible (chemical, biological, physical). Created in 2001, this experts’ network arose for a need to pool their skills and experiences to improve working conditions and enabled the creation of a national database.
The initial objective of the project was to develop an innovative methodology, centered around the notion of Exposome [Faisandier et al., 2007], in order to optimally take into account the multiplicity of occupational exposures encountered in the career of a individual and potentially responsible for a OHP. The methodology as it was designed consisted of selecting the so-called "significant" observations and restructuring them in the form of a network in order to visualize, quantify and characterize the various links existing between them.
​
It meant, for the RNV3P, to highlight, for a given disease, all the combinations of occupational exposures potentially responsible for the disease as well as the different associations existing between them. Each set of exposures then represented a node of the network, these being connected if two sets shared at least one common exposure. The combinations were then characterized in terms of prevalence (frequency), complexity (how much exposure to different agents within the same set) and temporal monitoring. The results of the various indicators were synthesized on two major graphs called "spectrosome" (the network of connections inspired by the exposome) and the "spectrum", allowing the identification of each combination of identified exposures and their characteristics.
​
The exploration and development of this methodology based on the RNV3P's health data finally led to the development of a generic methodology: the "Observational Surveillance" or "Spectrosome", which can be applied to different databases and different research contexts, such as the Belgian database IDEWE or more recently the IMIS database.
​
​
​
​
​
"Contact pathways between domestic birds and wild birds"
​
Supervisor: Dominique J. Bicout (EPSP)
Grenoble, France
​
This short project (2 months) took place in a global project of the Aviculture Technical Institute, following the avian influenza outbreak in 2006 in France. The wild water birds are considered as ‘reservoirs’ of the virus, but they barely encounter the domestics birds in the farm. We aimed to model the epidemiological risk of the poultry farm with an outdoor access, by describing the existing pathways between wild and domestic birds.
After describing the presence on different area of 82 different species, we clustered them according to their similarities of attendance profile into five different groups: aquatic, “bridge”, forestry, agricultural and opportunistic; and two different specific areas: dry or wet environment.
We were able to defined three different pathways that could lead to the propagation of the virus: 1/ direct contact between domestic birds and “bridge birds”, who are both present near the water area and farms; 2/ indirect contact in wet area by the aquatic, agricultural, forestry or opportunistic birds; and 3/ indirect contact in dry area by the “bridge birds”, agricultural, forestry and opportunistic birds.
​
This work has been presented and published during the 11èmes Journées de la Recherche Avicole et Palmipèdes à Foie Gras, in 2015 (http://www.itavi.asso.fr/content/voies-de-contacts-entre-oiseaux-deau-sauvages-et-parcours-de-volailles-pour-la-transmission).