OViral-Insights

Find correlations between covid spreading and social & demographic data

Comment

The problem

Have you ever wondered along with politicians, journalists or family and friends why some countries or areas of your home country have been more hit by the Covid19 pandemic than others? Ever tried to assume that was because of: population density, (un)following the state recommendations, percentage of the population tested... and so on. Our team also wondered... but since we are mainly a team of data people we wanted to know more about it and not only refers to assumptions. For us (as it should be for everyone else), assumptions are not reliable sources.

What OViral does

OViral is a website where the data is accessible to everyone, anytime. We wanted to bring the data to the greatest number of people possible and to be understandable. That is why the website has two features: visualise and simulate. The first one allows anyone to access data about the COVID19 in France without further analysis, no agenda here. The second one allows a user to enter criteria (localisation, age, gender, number of member in the household...) and then evaluate the risk to be infected by this category of person.

We do not save any information or simulation.

Our goal is not to discriminate or increase stress for some people, but we want to bring forward new reading keys of the coronavirus spread. We wanted States to acknowledge that some difficult living conditions might be a factor of contagion in case of epidemic. This would be an additional key for changing and developing new urban areas.

OViral Canvas

We also worked on computing the risk of a 2nd wave of contagion in departments: The objective is to modelize the delta of hospitalisation (difference between the number of new entries and the number of leaves) according to the last 5 days deltas. The more the real delta exceeds the prediction, the more risky is the department to face a new wave.

What kind of data are we talking about?

The application is fed in daily period by multi-dimensional data (medical, social, demographic, etc.) to monitor and understand what’s taking place, to identify the apparent dynamic of the pandemic, to prevent different scenarios, to establish preventive campaigns and, at the highest level, to limit the transmission of this virus in the national territory.

Our team gathered all the data available in France about the known COVID19 cases from:

  • the French national institute of economic and statistical information (INSEE)
  • the Directorate for Research, Studies, Evaluation and Statistics (DREES) of the French Ministry of Health and Solidarity
  • the national Public Health Agency (SPF)

The data gathered led us to build a wide range of specific information, namely:

  • The hosting capacity of eldercare facilities
  • The household living conditions during the period of lockdown (accommodation level, individual living alone, population distribution by age, etc.)
  • The gender composition and the socio-professional category of the population, by department and by region
  • The Distribution of the healthcare professionals (doctors, nurses, pharmacists, etc.)
  • Numbers of admissions in the hospital, patients in ICU, aggregated people discharged, and aggregated deaths by age group and gender
  • Numbers of COVID-19 suspected cases in hospital emergency, admissions among the suspected cases, medical acts by ‘SOS Medecins’ (emergency medical service) in the suspicion cases, and total medical acts by ‘SOS Medecins’
  • Data related to the COVID-19 diagnostics tests conducted by city laboratories at national level

The collected data were cleaned, labelled and restored in a proper format.

How we built it

For the website, we build it using Angular JS, D3.js and Bootstrap frameworks, hosted on Github pages. For the data algorithms and API, we used Python and Flask, hosted on Github pages. For the database, we have used Google Firebase Database.

Accomplishments that we are proud of

We had to lay-down our ambition about what we were able to compute and show on our website over the weekend but we are confident that it is a first step and hope this will help in the understanding of previous and future outbreaks. We worked as a team remotely and created a project from scratch in less than 48 hours!

What's next for OViral

We know that we want to add more criteria such as:

  • Employment rates
  • Activity sectors
  • Revenue of the household
  • Mortality rates from previous years • Geolocation data (on an anonymous basis) from a telecom company for instance (namely Orange which is already actively support medical research), to better understand and prevent interactions between exposed individuals and others
  • ...

We also want to make the data feeding mechanism autonomous (move towards a streaming data collection).

The evaluation system will need adjustment before implementing these new variables but it can be done easily. Also the algorithm will be more efficient with data from foreign countries: it will learn from new cases to understand this pandemic and its spreading.

Share this project:

Updates