Inspiration

The extent of spread of COVID-19 can only be known through wider testing. Globally, each country has limited testing capabilities and testing resources available. It is critical to use them optimally. We use algorithms developed for global optimization problems to optimize the placement and distribution of testing resources. In the case of SARS-CoV-2, local variables determine the local effectiveness of testing.

What it does

Where2Test offers a platform for the analysis of COVID-19 cases and exploring various testing strategies from a national to a county level. The user will be able to select a country and will be presented a trend history of incidents. The analysis component of the platform also presents model fit for the case history. In future, such analysis can be performed at even smaller scales such as a territory within a country. The testing strategy component of the platform displays an interactive map of the desired country. Relevant data related to testing is presented in a clear form on hovering on a certain area/province/county. We indicate our calculated local testing effectiveness via a heatmap. The testing effectiveness is estimated through optimization. The general optimization algorithm comes from global minimization of a given fit function defined by a constraint number. In this use case, it is the country-wide testing capability. Our algorithm allows for small variations of this capability to better fit a real case scenario and to find a better global optimum. In future, we intend to use local data such as density of population, density of confirmed SARS-CoV-2 cases, density of hospitals, time to reach the testing place, etc. to best place a testing place locally. The algorithm then gives a globally best distribution. We first aim at country-wide best coverage as the data available is usually too coarse to go well below a 10 km level. This often changes when looking at cities, where infrastructure data is available in much finer detail. With higher detail, of course, the optimization becomes trickier, e.g. you do not want to create traffic jams by placing your testing place at the wrong time.

How we built it

The platform includes two components --- the incidence case analyzer and test strategy explorer. The incidence case analyzer component is built using R. The data for testing strategy explorer is obtained from a python code that includes the optimizer.

Challenges we ran into

1) Finding the best local variables for describing the effectiveness of local testing.

2) Available datasets are often not granular enough to evaluate below country level.

3) Defining a local optimization function

4) Getting accustomed to using GIS and other data

5) Validation of our approach

Accomplishments that we are proud of

1) Being part of a great community!

2) Creating a full concept from a rough idea in a relatively short time frame.

3) Stepping out of our field of research and comfort zone and learning everything from the ground up.

What we learned

There is still a lot to do to make this work, so being open for others to help is important. We think a country-wide optimization, for now, is easier than one on a city level (see above), so we are aiming for this first.

What's next for Where2Test

We are at the beginning, but we have a great goal worth working for. We will put more effort into developing Where2Test in the future and are greatly motivated to someday have a full-fledged platform that can be used by governmental agencies to analyze the situation and decide their testing strategies.

Built With

Share this project:

Updates