Inspiration

We were inspired through our work at the Berlin-based Voice AI start-up Aaron.ai, which offers a virtual phone assistant for medical practices. A few weeks ago, we noticed a correlation between increased traffic on the phone lines of general practitioners and the rising numbers of new infections reported by the media. A preliminary statistical analysis indicated that, in most cases, phone traffic went up or down about one week before the confirmed COVID-19 infections did - even on a regional level. The idea of an early-warning-system was born!

The problem our project solves

A challenge many public health officials face, is that the time span between new outbreaks and actionable reports of case numbers is 9 days or more. So it is very hard to detect new outbreaks early-on and widespread lockdown became the norm. But it's becoming more and more difficult to enforce strict social distancing measures for a longer time. To phase out lockdowns in good conscience however, local containment needs to become faster.

The solution we bring to the table

The solution consists of 2 components:

  1. An AI-powered virtual assistant developed by Aaron.ai, which helps hundreds of medical practices to cope with overwhelming call peaks. It automatically answers incoming phone calls whenever the team is busy or off-duty. In addition, a web app makes it possible to prioritize and take care of patient requests whenever staff members have a free minute — even from home
  2. An early-warning-system: Whenever patients call about COVID-19-related symptoms, the assistant's response is — without using personal data and thus compliant with the GDPR — registered in a centralized database. Our machine learning models predict shifts in case numbers on a local or regional level by evaluating call traffic data from general physicians' practices and confirmed case numbers reported by public health authorities. The results are shown in a web dashboard.

What we have done during the weekend

  • During the weekend, we have crunched call- and public case data to prove our assumption that we can predict local outbreaks based on call statistics of general practitioners. And we did, as the graphs in our video show!
  • Moreover, we created a prototype version of the prediction dashboard and made the Aaron.ai assistant ready to understand all European languages. You can check our English-speaking demo out. The phone number is in the description of the youtube video.

Our solution’s impact to the crisis

The solution benefits the three most affected stakeholders of the crisis:

  • Public health officials, because they can now predict local outbreaks and proactively decide about preventive social distancing measures, days before they receive statistics on test results.
  • Physicians, who can effectively work from home, prioritize their response to incoming phone calls and thereby protect their team and patient using the virtual assistant
  • Patients, who can now reach a doctor's office 24/7 by phone without any waiting time At scale, our solution can have an immense effect on the lives of millions of Europeans by helping to flatten the curve, reduce restrictions to citizens, and fight the right battles.

The necessities in order to continue our project

  • We currently operate in Germany and Austria. For an effective EU-wide distribution of our product, we need partners with access to doctor practices across Europe. Financial support of these practices to introduce the assistant at no cost would certainly help to facilitate access.
  • To make the prediction dashboard ready to use, the underlying analyses need to be challenged and validated together with a research institution - which we are aiming to do in the next few weeks.
  • Last but not least, we have to invest to convert the dashboard from a hackathon result to a concise, error-free product that supports health officials to make the right decisions.

The value of our solution(s) after the crisis

  • The phone assistant will continue to play a vital role in medical practices, allowing them to serve patients faster, prioritize incoming calls easily, and save time with workflow automation.
  • Additionally, the prediction engine can be adapted to become a general early warning system for any epidemic or pandemic we likely have to face in the future.

Built With

Share this project:

Updates