It is with great pleasure that the Risk Scoring Symposium Invitational is formally publishing its v2.0 recommendations. The group, comprised of technologists, epidemiologists, data scientists, and public health workers, convened multiple times over the past few months to put these together. It took a major effort across the board to develop these suggestions, and we hope that public health authorities with a GAEN deployment give them a close read and consider implementing them.
This new version came about when the symposium attendees realized that the recommendations we had published a year ago were outdated. Namely, they targeted the transmission characteristics of the ancestral strain and what we knew at the time regarding the technology and people’s behaviors. Since then, we’ve been able to bring in data from a number of jurisdictions and update our models of disease transmission to better match what we know about the Delta variant.
Our goal with these recommendations is to help jurisdictions think through how they set up the characteristics of their exposure notification apps. We again provide two nets, depending on a jurisdiction’s tolerance for false positives or missed notifications, and encourage everyone to take this opportunity to review your settings and move towards one of these two nets.
There are a few changes that I wanted to highlight in the report:
- The symptom onset map has been consolidated into one model. Disease transmission characteristics don’t change based on which version of an app you’re running, and we wanted to reflect that. Therefore all effects of the proverbial size of the net are now included in the threshold and weight settings.
- Our data found that the “Narrow Net v1” had been sending out virtually no notifications. It would only trigger in limited circumstances, when the two people with phones were close by and their phones in optimal locations, such as face-to-face with phones in their front pockets. Even moving the phones to the back pockets caused this net to not trigger a notification, and the group agreed that this was too narrow. Therefore that net has been revised.
- There is a feature, “tiered notifications”, within the GAEN system that has not been getting used much, but has high potential. This allows a PHA to send different messages out based on varying thresholds, and can allow a jurisdiction to issue either multiple types of notifications with different instructions or allow them to send out identical notifications but get additional analytics data back in that can help with refining their recommendations over time. Either way this is a very useful tool and we wanted to encourage everyone to use it.
We also add in very specific recommendations about re-evaluating your setup on a regular basis. This system was never designed to be “set it and forget it”, but should be regularly updated based on the conditions in your jurisdiction. The jurisdictions that have been most successful using Exposure Notifications have been those that have regularly reviewed the data and made adjustments along the way. Whether reviewing the settings within the app or the language used to encourage people to take action, they do not treat this as a static system.
Thank you so much to all the participants within the Risk Score Symposium Invitational. This was a challenging document to compile and I appreciated all the effort each person put into it. As this is an open source project, we do anticipate continuing to update these recommendations moving forward and encourage our readers to submit questions, suggestions, and improvements on the LFPH GitHub. To subscribe to get updates on progress and discussions around these settings, send a blank email to risk-scoring+subscribe@lists.lfph.io.