New paper from @baym @BillHanage @LeeKShaffer. Examines benefit of "backward" contact tracing #COVID19 infectors of detected cases. Although basic idea is sound & used by Japan contact tracers, I'm a bit confused by several aspects of ms.
Thread
Thread
Background
In normal "forward" contact tracing (FCT), once you find a case, you identify their contacts who they might have infected & test&quarantine them; if any contacts test +, you trace their contacts and repeat until you don't find any more infected people.
In normal "forward" contact tracing (FCT), once you find a case, you identify their contacts who they might have infected & test&quarantine them; if any contacts test +, you trace their contacts and repeat until you don't find any more infected people.
In "backwards" CT (BCT) you try to find the person (the "index case") who infected the current case and then find the index case's contacts.
Idea for BCT is that for COVID19 & most other pathogens, >50% of cases don't infect anyone whereas small fraction lead to most of infection. 20:80 rule - 20% of cases are responsible for 80%+ of infection.
pnas.org/content/94/1/3…
doi:10.1038/nature04153
wellcomeopenresearch.org/articles/5-67
pnas.org/content/94/1/3…
doi:10.1038/nature04153
wellcomeopenresearch.org/articles/5-67
This means ALL contacts of >50% of cases won't be infected, whereas "index cases" have spread to at least 1 other case (the case that led to them being identified as an index case). The way to find "index cases" is BCT & then you find their contacts.
Their contacts are more likely to be infected than contacts of original case b/c you know they've spread virus to at least 1 person (the original case).
The BCT approach is used by Japan, S Korea & other places, but not always in US.
The BCT approach is used by Japan, S Korea & other places, but not always in US.
In this new paper, authors propose algorithm for what to do when Index case (i.e. a case that has infected another case) is identified: 
First confusion: Criteria #2. Since Index cases are identified BY DEFINITION as those that have infected 1+ of their contacts, then 100% would fit criteria #2. So criteria doesn't make sense.
As a result, their conceptual fig doesn't make sense to me. ALL index cases have 1+ infected contacts (the way they were IDd as an Index case) so every green circle should have at least 1 + contact. But they don't. Why?
Are they suggesting need Index case to have spread to 2+ people to quarantine rest of their contacts (the original case + 1 more of their contacts ID by BCT & FCT)? That seems a bit strange, so I'm pretty sure I'm misunderstanding. But clearly something here doesn't make sense.
(Note: One can also use same idea for normal forward contact tracing & only require quarantine for contacts if one tests + (since we'd know case was infectious). This might make sense if we had list of contacts & initially reached one & they test +. We should then find others!)
Also what about delays? Time needed to find Index case is long: 1st need to find one of their infectees, usually by severe symptoms, which occur ~13d after infection: doi.org/10.1016/S0140-…, then contact index case & then contact their contacts, so >14d after their infection.
Infectious period is usually days -3 to +7 days after symptom onset or 3-13d after infection:
So quarantining contacts of index case not effective in cutting off chains of transmission. They are no longer infectious. Same for isolating severe cases.
So quarantining contacts of index case not effective in cutting off chains of transmission. They are no longer infectious. Same for isolating severe cases.
One could, of course, find contacts of contacts of index case & get them to quarantine & if done quickly could cut off chains that way. So BCT is good way to find new chains of transmission to try to stop, but to do so, need to first BCT & then FCT 2 chains forward. Not discussed
Finally, need to account for fact that that household attack rate for SARS-COV-2 is low (often 10-20%: doi.org/10.1016/S1473-…) & non-household attack rate is v. low (1-2% ). So only small fraction of contacts get infected, even for people that do transmit.
So lots of people end up quarantined when trying to break chains of transmission. If you do BCT & then FCT x 2 from "index case" you identify ~N^2 contacts where N is avg # of contacts of each person. Still good idea, but need to allocate lots of safe space for quarantine.
