scientific articles published by the US Library of Medicine were analysed by La mètis to successfully conclude this project.
health solutions for the future are presented using this dynamic display tool.
The data collected explores scientific publications from 1980 to the present to calculate their probability of dissemination between 2015 and 2030 via a predictive analysis algorithm.
men were diagnosed with prostate cancer in France in 2008. This makes it the No 1 cancer affecting men, representing 1% of the French male population.
semantic zoom levels (health solutions, management techniques, trends and main approaches) allow users to progressively delve into this large volume of complex data.
pilots the opening and sharing policy for public data, called 'open data', within the General Secretariat for government modernisation. More than 20,000 free and reusable datasets are currently available on the data.gouv.fr government platform.
Evolution therapeutic solutions
Details of solutions to a particular year
Access to the articles of the American Library of Medicine by selecting a curve
“The decisions that guide our future care system are primarily based on the knowledge of current medical practices.” From this statement arose the need to analyse the medical literature to understand its nature, its main focus and isolated innovations so as to design the most efficient care proposal possible for the future.
This is the work that La Mètis, specialists in strategy design using large volumes of non-structured data, has taken on. In the summer of 2014, La mètis took prostate cancer, the most wide-spread cancer in France among men (53,000 new cases per year), as a case study. Studying the dynamics of innovations relating to the management of prostate cancer between 1980 and 2015 has enabled data scientists, using a predictive analysis algorithm, to calculate the probability of dissemination of these innovations between 2015 and 2030.
To make this analysis meaningful, to make it readable and therefore be able share it with as many people as possible, La Mètis is asking Intactile DESIGN to create a dynamic data visualisation tool that illustrates the changes in volume of scientific articles between 1980 and 2030 relating to prostate cancer, as published by the US Library of Medicine (Pubmed). It has chosen Laboratoire 3ie to develop this application.
This combination of skills must lead to the development of an analysis and exploration tool that will allow patients, doctors, health organisation departments, health industry players and policy managers to explore a large volume of data. This data is now structured as it is the result of a semantic and mathematical analysis of tens of thousands of scientific publications available from the US Library of Medicine.
An iterative and participative design
The first goal of the visualisation tool is to provide hands-on experience for users to become familiar with the quality and abundance of the results obtained during the analysis phase. These results strongly call into question the view of the treatment of pathologies in hospitals from the present to 2030, in particular by promising the development of alternatives to surgery and an increase in targeted therapies. The Big Data approach for the prediction of modes for managing prostate cancer was, as it stood, difficult to understand by data non-specialists because it was presented in the form of raw data, in arrays comprised of thousands of non-hierarchical rows.
Our role as designers consists of making this data understandable.
Differing profiles have been gathered around the participative design table: data scientists from La mètis and the Étalab mission, doctors from the urology department of Hôpital Tenon and the Institut mutualiste Montsouris, patients, representatives from the national association of prostate cancer patients (ANAMACAP) and designers from Intactile DESIGN mandated to co-design and subsequently create the graphic design of the data visualization.
Be they patients, caregivers or administrators, future users have varying expectations and needs with respect to this tool – needs that sometimes converge but that other times diverge. The first session therefore consists of testing the project waters, listening to users, and letting them talk about their situation. The patients present are in need of this. Using a post-it, each participant tells their story. Patients need to understand their illness, administrators want to know the cost of treatments and doctors want to know the availability of treatments.
We therefore choose to put end users at the core of the design in order to understand their expectations in dealing with this new tool.
The designer annotates, summarizes, prioritizes the needs expressed and compares them with possible solutions. Is this data available? Is it reliable? Can it be obtained within the time limit given?
All these requests therefore result in changes to, and sometimes additions to, the specifications – already numerous – provided by La mètis.
Participatory design around a paper mockup
The part played by Intactile DESIGN therefore focuses on the creation of pertinent visualisation tools. How do we format the data? What do we show the user? And by what means?
Two types of visualisation tools quickly come under consideration. The first expresses the volume of publications per year in the form of ''cells''. The links between two cells represent the links between the solutions within the articles analysed. The user moves through time using a timeline.
The second tool presents the change in the volume of annual scientific publications in the form of curves – that are not unlike a network of veins and arteries – ranking the solutions in relation to each other, year after year.
The working group is divided on this issue, with some participants preferring visualisation in the form of curves where the changes are visible without user interaction, while others prefer visualisation in the form of cells that focus on the concept of links.
The latter format will be used less and less, as the links between solutions are difficult to represent semantically (complementarity or competing solutions)
Sketched, different zoom levels of 2 visualizations in competition.
Design & Big Data
The basic challenge of the design work is therefore based on the issue of Big Data. The data to be represented is huge in volume, complex and heterogeneous. How can we help the user understand such a large volume of data? How can the user enter progressively into this flow of information? How do we prioritize data? How can we make it readable? How can we use digital technology to avoid overloading screens? What type of interactions should be implemented to make the user experience flexible and fluid?
The semantic zoom
Data is explored progressively using a semantic zoom. Level 4 illustrates the main approaches in prostate cancer treatments and their respective importance. Prevention and therapeutic education, represented in blue, are the solutions relating to all attitudes or behaviours that tend to prevent recurrence of the cancer and detect it for early management.Precision medicine, represented in red, presents all the appropriate treatment solutions on a case-by-case basis, using an analysis of the patient's history and the results of the patient's diagnostic tests – whether these are biological, genetic or imaging tests.
These colours, chosen to reflect those used in medical schemas, are used while exploring.
At each zoom level, the ''parent'' cluster is broken down into an increasingly more detailed level of information. Level 3 therefore corresponds to 12 underlying trends for prostate cancer treatment. The second level represents the 31 techniques available for managing prostate cancer and, last but not least, the first level shows us all 70 solutions used for managing prostate cancer. At this final level, a system of filters is offered to the user so that only the solutions relative to the cancer stage concerned are displayed.
Regarding scientific advances, the change in the number of new cases each year (incidence) as well as changes in mortality (deaths directly linked to prostate cancer) appear in the lower graphic. Both of these indicators are dropping.
Data and interaction
Every health trend, technique or solution can be searched. Clicking on a curve, for example, displays the total number of publications concerning this curve as well as the detail per year. Information on treatment availability (date of entry on the market, etc.) can be added.
At the most detailed zoom level, each curve becomes a shortcut to the articles in the American Library of Medicine. The display includes a list of the 10 most recent articles for the selected solution.
In addition to being able to search the database by subject across the volume of publications, the user can look at a specific year. Clicking on a year highlights the solutions presented and ranks them in a list – with the highest the one with the most publications and the lowest the least.
For more information, additional messages appear on the visualisation tool.
The graphic design of the display will be refined throughout the design process and during development, because it is only in this phase that we will have possession of the full database.
The usage storyboards as well as the integration specifications delivered (colours, typographies, font weights, etc.) were used in drawing up the specifications given to the requested party for development. Intactile DESIGN then began their integration follow-up mission.
The application was tested by many users at each version. Graphic elements are to be refined. Feedback on certain actions must be taken into account. Small bugs have been fixed.
In this project, our follow-up mission is limited to sending this feedback to Laboratoire 3ie in the form of annotated screenshots or as a list of modifications to be performed. This working method places the responsibility for design integration on the developers and clearly separates the prerogatives of each party. Intactile DESIGN is responsible for the co-design and graphic design of the visualisation tool. Once these two steps are validated, the development team takes over creation of the visualisation tool and design integration. Integration follow-up allows the designers to communicate with the developers. The main advantage of this approach is therefore that it is easy to implement as interaction between the design and development teams is by simple exchange of e-mails.
Drawing on our experience of working with designers and developers, especially in-house, we are implementing a different method in other projects. In these cases, we no longer divide up the work (two teams, two distinct phases) but view all players in the project as part of a single multidisciplinary team working together. With this approach to software development, the role of the designer is not just to deliver specifications, but to work on design integration – which on a Web project, would mean writing style sheets (CSS). This of course requires a bit more organisation – implementing common tools, access and code testing as the development progress… but the time saved in avoiding the usual back-and-forth between designer and developer (design integration by the developer, designer feedback, integration of this feedback, feedback on the integration of this feedback, and so on) is enormous. This method also provides considerable integration flexibility since the designer can modify the graphics himself in accordance with the technical constraints that inevitably arise during a development phase.
An experience worth repeating
If the presentation of this project has given you ideas, if you also have a dataset to make the most of, then whatever the domain, do not hesitate to talk to us. It will be a pleasure for us to design, with you, the visualisation tool most adapted to enhance and share your work. Get in touch!
Introducing visualization on the website data.gouv.fr