How do we know that Person B was infected by Person A, unless Person B went into quarantine directly after interacting with Person A? If they don’t go into quarantine, how do we know that a random Person C didn’t interact with them and cause the infection?
And for the vast majority of infections that can’t be contact traced like the example above, how do we determine the date they were actually infected if the disease can be transmitted asymptomatically?
In: Biology
Sheer numbers and statistics. If someone comes to a hospital with symptoms of a novel virus, and then 7-10 days later 7 hospital workers report similar symptoms, then it’s a safe bet that the patient was the point of infection for all of them. **Statistics** is a field of study which helps us draw concrete conclusions from unreliable data. If only 1 hospital worker were to report symptoms, we wouldn’t have very strong confidence in that 7-10 day incubation period, but as the numbers go up, so does our confidence.
When there is an outbreak somewhere, epidemiologists (scientists who study epidemics) will trace back to the estimated first date of exposure, then watch the numbers of those reporting symptoms. First date of exposure is not very reliable data – it often relies on assumptions (like the hospital example above) or self-reporting (“Well, I was in the airport on Sunday so that was probably it”).
However, it takes far fewer patients than you might imagine to get a strong estimate. For instance, during the 2013 bird flu epidemic, it only took around 250 patients for epidemiologists to get a very strong sense of the incubation period – they were 95% confident that the average was between 3 and 4 days. There are a lot of specialized techniques used, too many to go into detail here – but we know, from experience, that people tend to underestimate exposure; that we should discount anybody who reports a very long or very short exposure time; that exposure times generally fit a bell curve; etc.
When possible, epidemiologists will look for volunteers to help them get better numbers. Known-healthy individuals living in an area with an outbreak can report for daily testing, which allows doctors to know the exact date of infection and the exact date that symptoms appear. This can give exact numbers, but the issue is that generally the number of volunteers / test kits available will be limited, and incubation periods can vary wildly, so often having a huge amount of unreliable data can actually be better than a small amount of very reliable data.
Further reading:
https://academic.oup.com/aje/article/182/8/723/207340
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4301984/
https://jamanetwork.com/journals/jama/fullarticle/2762028#jld200013r4
Latest Answers