Authors

Covid-CareBot

Challenge

The coronavirus COVID-19 is spreading rapidly throughout Europe. In order to contain the disease, suspected cases must be identified as early as possible.Communities need an explainable, granular and dynamic view of risk to navigate these times of extreme uncertainty. The health care organes of the government should be provided statistical data per zip code to identify the regions that are most in need of help and warn the rest of the population about the hotspots.  How can we realize a corona tracing app which can be both implemented quickly and which is secure and trustworthy as well? How can we encourage the population to share their data regarding their well being and symptoms in the easiest way possible for the user? How can we help the population in need of medical care and help them manage their symptoms?

Concept

The concept behind our solution is to make an easy to use add-on application on frequently used apps like SBBApp, MeteoSwiss, Alertswiss or as a mini standalone application that would check in with the user daily about their well being and symptoms if they are willing to share these via a push notification. The user is being approached in a friendly conversational manner as opposed to traditional complicated and rather boring surveys. This does not only support the user and keep them aware of their health but also based on the submitted ZIP codes can provide a useful statistical dataset for the swiss government to use. Based on the data gathered one could identify the areas with high infection risk based on the algorithms we use to identify individuals under a suspicion of infection and the workload the health care providers would be up against in near future. Provided enough data from users who get tested, an AI can be trained and so be used more efficiently in the prediction of contamination for each individual. The survey results would function as an early warning for health authorities concerning the areas where needs might be increasing. Simultaneously, the app would encourage users in high-risk groups with early symptoms and other users with symptoms lasting over three days in a row to contact COVID-19 health hotlines through a push of a button if they find their symptoms to be severe and are in need of help. The data provided by the user is collected anonymously so that the individual cannot be identified. The data is iteratively deleted after 21 days further ensuring user data security.

Data Collection

On the first day of usage of the app following settings have to be set up by the user:

  • Year of birth
  • Gender
  • Zip code
  • Whether they have any of the pre-existing conditions (cancer, asthma, chronic diseases etc.)
  • Whether the user smokes or vapes
  • How big their household is and how many individuals have no access to the app (many user's experiences can be logged in from the same device if needed)
  • Whether they have had a physical contact with someone who is infected
  • Whether the individual has tested themselves All of these settings can be changed at a later date (if one moves, gets tested etc.)

Further the user will receive daily push notifications that would ask them about their well being in a caring conversational manner and the way the user is approached varies based on the information obtained from the previous days. If the user is feeling ill, they are asked about their symptoms. This information is further on used in the analysis and is saved for the user in a form of a spreadsheet that they can further use individually. The user is also asked daily whether they have had contact to anyone who has been tested positively for the virus.

Advantages for the user:

  • If an individual in a high risk group experiences symptoms they are directed to a local COVID-19 hotline
  • If an otherwise healthy individual experiences COVID-19 symptoms for 3 consecutive days they are also directed to the COVID-19 hotline for further assistance
  • The app keeps the users self aware of their health and possible symptoms and encourages them to take action
  • The app/add-on is very easy to use and is designed in a way that it is applicable for almost all demographics (especially for the older individuals the font is chosen to be relatively big and legible, the colors are calming)
  • The application can be useful for individuals who live alone as it checks in with the user daily
  • All the information logged in by the user is saved in well ordered spreadsheet on their device that the user has an access to and  can show to their health care provider if need be
  • The user is also notified about in which infection probability group (high, medium, low risk) they are based on their symptoms

Data Analysis

The goals of the data analysis are as follows:

  • Identify definitive and possible infections
  • Classify people into infection probability groups (high, medium, low) based on their symptoms and the information about their contact to infected individuals
  • Order the data based on the zip codes
  • Make predictions about the rest of the population in each area, that is not logging in their data based on gained ratio from the people using the app

With a help of a scientific paper, we have developed an algorithm that would classify people who are experiencing abnormal symptoms in comparison to their regular health state in high, medium and low infection probability groups. For this we mainly look at the occurrence of the symptoms and the contact to infected individuals (logged in by the user). Knowing the number of confirmed (tested positively) cases and the cases that are suspected to be infected based on the results given by the algorithm we can calculate the ratio of these individuals divided by all individuals using the app and apply the ration to the rest of the population in each Zip code covered area (provided the population count in each area). This data sorted by the  ZIP code areas can predict the areas that might be in bigger need of medical assistance.

Future Steps for Data Analysis

  • A neural network architecture is to be built that would take all the information filled in by the user in their primary (age, pre-existing conditions etc) settings and their symptoms that are logged in daily in a specific timeframe (up to 30 days, however this can be raised if needed)
  • After a collection of logs of multiple users, where people have been tracking their symptoms for some time and finally tested positive or negative, the neural machine can be trained on the said dataset
  • This artificial intelligence can be tested with the data of the same type also collected with our app
  • Since for many other diseases the same factors (age, pre-existing conditions etc) and symptoms with possibly small alterations are used to determine whether a person is sick or not, this neural network architecture can be reused in other outbreaks, provided the data is collected in the same way as we propose with this add-on
  • With a tested functioning AI, that might recognize connections in the symptoms and new tendencies a human might not, the infected individuals can be recognized easier and faster

Output of our system

  • Heatmap showing the predicted infections relative to the population of cities based on the algorithms and possibly later AI
  • Customized Reports for Hospitals, Government, other institutions conveying the information about the risk groups and infection spread for each location
  • The data collected from the individual user would also be available to the users themselves in a form of a well-ordered spreadsheet containing all the information about the symptoms they have been experiencing throughout the time of application usage, which the user can show to their healthcare provider

Advantages for health care authorities and government:

  • The data collected can be used to predict where the health care facilities (hospitals) might need additional resources. This will serve as an early warning and make sure that medical capacities of our facilities are not crossed, effectively helping everyone in need of medical assistance
  • The app encourages people with symptoms to reach out and ask for further course of actions, therefore more people in need of help would receive that and people who are not in need of assistance would not be needlessly wasting health care resources
  • Areas with high infection risk can be identified and the population can be notified about it after data collection from a big portion of population
  • In the future steps based on the data collected an infection heat map can be implemented, this would be available to all of the users as well as the government
  • Partners in governmental structures are needed, an integration of the add-on would be especially useful in an app like AlertSwiss, as it can reach many users and be a good data collection medium for the governmental health care departments and an informative medium for the user with the implementation of heat map and warnings about high risk areas
  • This system can also be used in future pandemics/epidemics for symptom tracking

Further Uses of the Add-on

As mentioned before this kind of app is highly dynamic and can be customized to use in any other pandemic or epidemic or even a cold spread for the uses of data collection. The algorithms that are to classify people into infection probability groups are to be adjusted for each disease, however the neural network architecture would need almost no alteration at all, however, it would need new data collected by the app for each disease to be trained on. The way of implementing the heat map and the user spreadsheets would also not need any alteration. The implementation of this kind of add-on might accelerate the way we deal with further epidemics by providing health care authorities with all needed information to distribute their resources in an adequate manner. It is also to be considered that the uses of this application are not limited to any specific country in particular and can be implemented outside of Switzerland as well.

Try It out

Hackathons

Technologies

.net, c#, core, xamarin

Devpost Software Identifier

258502