Why QR?

Last updated 07.04.2020


Currently, the apps used for contract tracing rely either on GPS data, cellular networks, wireless or bluetooth low energy (BLE) technologies to passively and continuously track the proximity between users, and (in case of GPS or cellular networks) the location where contact took place [link].

With the current contact tracing implementations, there is a number of major issues to tackle when using apps for contact tracing, both in Switzerland and globally.


User privacy & data protection

Guaranteeing user privacy and data protection when using contact tracing apps is a significant concern. TraceTogether, an app the Singapore government has been using for contact tracing, has already raised significant issues with its application if it were to be used in Europe or North America [link].

At the same time, real-time contact tracing performs an important function in fighting a global pandemic such as the one of COVID-19 [link].

An ongoing collective effort, both local and worldwide, is currently under way to allow real-time contact tracing to effectively take place [link], while at the same time, not violate privacy and data protection laws, often differing between countries.


Data accuracy & false positives

With both GPS and BLE scanning, the data collected for contact tracing (1) has low accuracy and (2) requires significant post-processing efforts to remove false positives.

GPS (incl. cell tower) contact tracing

  • Indoor accuracy - GPS data are either highly inaccurate or unavailable when used indoors. Additionaly, the location information provided by cell towers is only an estimate, usually spanning tens of meters. In this case, proximity to other app users cannot be guaranteed.
  • Outdoor accuracy - Outdoor accuracy is guaranteed at best case to be approximately five meters, on open sky. This can be improved by synchronously using other tracking sources (e.g. cell towers), with additional privacy rights needed to be provided. In some cases, these additional sources are not available (e.g. no reception).

The low accuracy of GPS data cannot guarantee reliability of using the data for contact tracing, creating a lot of false positive tracing chains.

As a very simple example, if a user is on his way to the grocery store and a public transport bus with app users passes by, a tracing chain would falsely be created between them. The same would happen with app users from nearby buildings, among other examples.

To increase data accuracy and remove such occurences, heavy data post-processing would need to take place but is also not guaranteed to remove all false positive instances.

BLE contact tracing

Using bluetooth to determine the distance between users generally works well when using Bluetooth Low Energy.

For users with older phones that do not have hardware BLE capabilities however, the distance estimation can only be an approximation based on RX power. This approximation is affected by many variables (e.g. clothing thickness, antenna orientation and so on). This can impact the data accuracy significantly.

In addition, similar false positive issues also exist with BLE tracking in regards to both indoor and outdoor locations. Again, the only way to remove such false positive occurences is with heavy data post-processing, also not guaranteed to remove all false positive instances.


Data volume

To ensure and further improve the data accuracy of either GPS or BLE, high sampling frequency of the data is necessary. For real-time contact tracing, this leads to a significant data volume, making data processing cumbersome.

A hypothetical situation follows. A person generally walks with an average walking speed of approx. 5m/s. GPS position accuracy is at best 5 meters. The approximate maximum distance for BLE connections is about 30 meters.

To ensure that a person, during his daily walking, he did not come in close contact (less that 4 meters) with an infected person, we would need at minimum one position recording per second with the GPS solution or one distance estimate per six seconds with BLE.

The per user, per day minimum data volume needed for GPS tracing is 675KB (one float for lat/lon respectively) and for the BLE case 56.25KB (one float for location independent distance). For a contact tracing history of fifteen days (as is currently recommended), these values become 9.89MB for the GPS and 843.75KB for the BLE solutions respectively.

Extrapolating this to the population of Switzerland (8.57 million), with an app adoption rate of 25% (only 2.14 million people using an app) we have a total bi-weekly data volume of 20.2TB for the GPS case or 1.68TB for the BLE case.

The actual data volume to be processed will of course be less, as the tracking will happen asynchronously (only when new cases are discovered), but even then, post-processing this volume of data to exclude false positives will prove to be quite cumbersome.


Using QR code scanning instead

Users of spreadit scan one unsafe location per visit. This approach mitigates the issues outlined above.

Location accuracy is instantly guaranteed, with minimal to no post-processing necessary. At the same time, passive, continuous user tracking is not necessary, as the user only needs to scan the relevant locations they visit, once. In addition, tracing data volume is kept at a minimum inherently. Due to this, data post-processing is largely not necessary, allowing for faster chain-identification response times.

If the need for increased contact distance granularity using BLE data arises, this can also be achieved, but only at the locations with a QR code, and only for a limited amount of time. In such a case, the core functionality of using QR scanning, ensures that tracing happens accurately even when BLE tracing is not available (e.g. older smartphone).

Is QR the only solution?

No, it is not. A similar system could be set up with numbered tickets, manual or automatic recording of entry to unsafe locations, and a variety of possible ways. It is however one of the most widely available technologies that is (1) straight-forward to implement, (2) reliable to use, and (3) scalable enough for this purpose. We do also envision the possibility of a future combination of QR with BLE or GPS data, facilitating an increase in tracing accuracy and response times.


Functional operation

1 Transmission routes

The tranmission routes of COVID-19 seem to be both aerosol and droplet in nature. Furthermore, it can survive on various surfaces up to a few days [link1][link2][link3][link4][link5]. This means that closed spaces can be characterised as potentially unsafe locations, while open spaces (e.g. pedestrian paths, parks, fields) are safe, with the condition that the population maintains a distance to others, according to government recommended measures [link].

The measures taken by the Swiss [link] and governments worldwide, prohibit (1) the gathering of more than a few people in the same location, (2) the function of non-essential businesses and recommend (3) keeping a safe distance between people (to avoid possible droplet transmission).


2 Contact tracing

2.1 Known contact tracing

Known (close) contact tracing can be easily deduced, either by the government, individually or collectively. Known (close) contact tracing is defined as the tracing of spatio-temporal contacts that are known or close to an individual. This can be people that the individual knows and comes in contact with frequently. These are for example friends, family and work colleagues.

2.2 Unknown contact tracing

Based on the above, and agreeing that the population acts in unison and solidarity towards beating this virus, the only difficulty in identifying infection routes happens via unknown contact tracing.

Unknown contact tracing describes the tracing of spatio-temporal contacts that are unknown to the individual. This can be, for example, a person sharing a bus ride, shopping at the same grocery store, or sharing an elevator in a building.

This type of contact presents the largest difficulty in identifying and containing transmission routes.


3 Unsafe Locations

Given what we know about the virus, the government measures, and trusting in the support of the population to mark them, there is a finite number of locations that unknown contact transmission can occur. Examples of these include

  • Shops - Grocery stores, post offices, pharmacies
  • Public transport - Buses, trams, trains
  • Common spaces - Building hallways, elevators, stairways

These locations all have a risk of infection that depends on a number of parameters. Some of these include

  • The number of people in that location within a time period
  • The size of the enclosed space of the location
  • Frequency people come in contact with objects within that location

For these locations, real-time monitoring of the population is both unnecessary and unreliable.

This is where QR code scanning can help. By tagging each of these locations by a unique QR code, the individual visiting the location can self-record his presense there, when it happens.

All other data (e.g. how he ended up there, how long he was home, etc) are irrelevant to the unknown contact transmission route.

3.1 Location description

Each location gets an individualised QR code that describes

  • Address - The address of the location. This includes a street address and a postcode
  • Type - The type of location (e.g. grocery store, bus, elevator)
  • Size - A categorized location size (small, medium, large)
  • Object handling - A categorised possibility of the frequency of touching objects in this location (low, medium, high)
3.2 QR code: crowd-sourced generation

The QR code generation is crowd-sourced and highly depends on our collective effort.

You are encouraged to create and place a QR code at any location type that is missing it. This code will then be validated against the server and added to the list of locations by the first user that scans it. Subsequent users can then use it normally.

To generate a QR code for a location please visit XXX. You have the possibility to print it directly from the webpage or save it as a PDF.

3.3 Location risk assesment

Based on the location description, a transmission risk assesment is generated to describe the risk each location has to contribute to unknown contact transmission.

This number is also greatly dependent on the concentration of people in that location at any given time. This is why crowdsourcing the number of people that visit each location at any given time by using the app yourself is very important.

3.4 Location risk validation

In the case of detection of a person being infected, the above risk assesment number is subsequently validated and adjusted accordingly via unknown contact tracing for that person, over a time period.

3.5 Location risk measures

If a particular location is found to have a significantly large risk assesment number, then the epidemiologists, together with the government, can take additional measures to reduce the unknown contact transmission route for that location, and not by generalized measures for the population.

Such measures can for example be, increasing the frequency of buses for a particular bus route that was found to be too crowded, or a recommendation to visit the grocery store at particular times to avoid overcrowding.