When we don't know how ~50% of cases acquired COVID-19 in our current lower (though creeping up) prevalence environment in #Ontario, I think it's time to re-examine some bedrock assumptions. 1/n https://twitter.com/IrfanDhalla/status/1303360348028891138

Maybe we should start with what constitutes high vs. low risk contact. For example, AFAIK some asymptomatic household contacts are still not being tested in Ontario on account of this somewhat arbitrary distinction. How many are infected but not identified? 2/n

We have no way of knowing. But, secondary HH infections that go undiagnosed do not have contact tracing completed and are potentially niduses for ongoing transmission prior to being advised to self isolate, or if they don't strictly adhere to self-isolation. 3/n

As this relates to Canada's COVID Alert app, an exposure is currently defined as <2m for >15 minutes, using bluetooth pings q5mins. What are the test characteristics of these somewhat arbitrary thresholds for defining high risk contact/utility of testing? 4/n

Would we identify additional cases at greater distances or lesser times? Within the limitations of bluetooth range, our COVID app (along with PCR sampling in the context of a study) could be configured to answer these questions. 5/n

Unfortunately, AFAIK the app isn't tracking distance and time data as continuous exposure variables (which btw, wouldn't have any substantial privacy implications beyond those already considered in it's design/implementation). 6/n

This strikes me as a missed opportunity to refine our understanding of contact risk as a function of distance and time based on evidence, which would have broad application to the determination of risk for contact management in general. 7/n

It also seems clear at this point that exposure to certain environments is probably equally as important to consider (i.e. driving substantial transmission) as close person-person contact, as seen in several well-documented super-spreading events. 8/n https://twitter.com/DFisman/status/1303809207007928321

I think we might be surprised at the proportion who are not linked to cases by current definitions of high-(or even low-) risk person-person contact, but *would be* linked to environments where transmission occurred, if we were able to identify those environments. 9/n

Perhaps we would be able to link as yet unlinked cases to unknown super-spreading events, or perhaps we would find that isolated or oligo-transmission events >2m or <15 minutes are more common than bedrock assumptions about modes of transmission would have predicted. 10/n

NZ has been facilitating more exhaustive location tracing along these lines with QR codes via their tracing app. It would be so interesting to know the extent to which this provided additional insights during their second outbreak. (Please link if any info!). 11/n

Simple implementation: print a QR code and attach it to entry. People would have to participate by scanning it. The COVID Alert app would have to support this. 12/n

An alternative would be to incorporate this function directly into our COVID Alert app, without the need for QR scanning, by automatically logging entry and exit times from anonymized indoor locations via bluetooth. 13/n

Downside: would require a bluetooth device at entry, but businesses would likely need to have a greeter encouraging the QR code scanning anyway, so their phone could perform this function. 14/n

Wouldn’t it be fascinating (and essential to our control efforts!) to know what proportion of unlinked cases could be linked to certain (probably mostly indoor) environments, while never coming within 2m x 15 minutes of cases? 15/n

Or, whether different thresholds of distance and time offer better exposure risk test characteristics to better inform contact management?

This kind of information is within our grasp, but… 16/n

If you don’t take a temperature, you can’t find a fever. /end