Massive Testing Covid-19

Inspiration

As the experiences of Singapore and South Korea testify, carrying out massive tests constitutes a relevant method to slow down the advance of the Coronavirus. However, mass testing is a scarce resource.

What it does

We have implemented a decision system that, given a number of tests and a time horizon, plans their optimal distribution. So, it is returned the optimal number of tests to be performed in each town of a region for each day of the time horizon. The objective is to minimize the total number of individuals infected by coronavirus. Day after day, the system is updated through the data of infected people recently detected and all the parameters of the model are recalibrated, re-planning everything. It is based on mathematical models, combinatorial optimization and artificial intelligence.

An example would be, if a region has 100.000 tests to perform the next 90 days, what number of tests would be optimal to perform each day at each one of its towns? The system provides the optimal solution to minimize the total number of individuals infected with coronavirus for the time horizon. It also provides the 'coronavirus map' of the region and its forecasting, although this is a secondary objective. In addition, other aspects are taken into account such as the number of available testing team, the maximum number of tests that can be performed per day in a location, etc.

The problem is complex. On the one hand, if tests are carried out at an early stage then it is difficult to obtain infected individuals (which can be understood as a waste of the tests). On the other hand, if we apply them at an advanced phase then it is already easy to get infected individuals but it is too late.

A solution usually implemented consists of carrying out sampling proportional to the size of the population or the proportion of infections. In our computed proofs both decisions can be improved. For example, if we carry out 100,000 tests in the towns of the Valencian Community (Spain) by the solution that our decision system offers then we save 122 lifes versus testing by proportional sampling (345 more lifes are saved if we compare our solution with the situation in which no tests are applied).

Nowadays, the point is about how to distribute tests, but later this system can be used for optimizing the distribution of COVID vaccine.

How I built it

mathematical modelling; mathematical programming; AI; Deep learning;c#;

Challenges I ran into

We are working to improve our algorithms in order to save more lives.

Accomplishments that I'm proud of

We have already a Desktop Solution completly functional. We validated our mathematical models and we have saved hundred of lives in our simulations. We are able of mathemacally proving that.

What I learned

We learned about maths, data-science and simulation applied to the spread of COVID-19 and how to stop-it.

What's next for Massive Testing Covid-19

We want to integrate the Desktop Solution with a Web Service Solution to equip the Public Administration with a tool to plan the distribution of massive tests for saving as many lives as possible.